Wild Computing

                             Introduction

My subject here is the philosophy of Internet AI engineering: the general 
ideas
that must be understood in order to create intelligent computer programs in 
the
Internet medium. Only a few of these ideas are of my invention; this is a 
body
of knowledge that has emerged from the community of computer scientists,
engineers, complexity scientists and associated thinkers over the last few
decades. But I believe I have filled in a few crucial gaps.
This is not a traditional "AI philosophy" book. I take it for granted here 
that
computer programs can have minds, if they are designed correctly. And I take 
it
for granted that computer programs can be conscious -- ignoring the 
"problem" of
computer consciousness just as I ignore the "problem" of human consciousness
every day, as I interact with other humans. The reader who is interested to 
know
my views on these matters can consult my previous publications.
I firmly believe that, by fleshing out and implementing the ideas given 
here,
over the next decade, we will be able to create Internet-based AI programs 
that
surpass human beings in general intelligence. Some of our expectations about 
AI
will be remain unfulfilled -- for example, the Turing test, which requires a
computer to simulate a human in conversation, is unlikely to be passed until 
we
can engineer human-like bodies for our AI programs. But in other areas, our 
AI
programs will exceed our expectations dramatically. Our understanding of the
collective intelligence of the human race will be greatly enhanced by
interaction with AI systems linked in to the collective data resources of 
the
Internet.
Of course, I am aware that AI optimism has been proved wrong before -- AI
pioneers of the 1950's and 60's predicted that true machine intelligence was
right around the corner. But like optimism about human flight 150 years ago, 
AI
optimism is bound to be proved right eventually. The AI theorists of the 
50's
and 60's lacked many the ideas needed to realize machine intelligence, but 
they
also lacked adequate hardware. If they had had the needed machinery, they
probably would have corrected their ideas in the iterative manner typical of
empirical science. Now we have PC's with 4 gigabytes of RAM, mainframes with
100; and we have high-bandwidth network cable that can link an arbitrary 
number
of computers together into a single computational process. Just barely, we 
have
the firepower needed to implement a mind. The ideas in this book provide the 
key
to making use of this firepower.
Part of the key to thinking about Internet AI correctly is placing AI in the
correct context. Exciting things are happening in computing -- the Internet 
is
more than just a way to send naked pictures of Pamela Andersen to your 
buddies
in Chechnya. There is a new kind of computer science and computer technology
brewing, one that focuses on self-organizing networks and emergent dynamics
rather than algorithms and data structures (though algorithms and data
structures will still be there, to be sure, just as the bit-level constructs 
of
assembly-language were not vanquished by the advent of structured 
programming).
Artificial intelligence is essential here, because humans don't have the 
time or
the ability to deal with the glut of rapidly shifting information that 
modern
networks bring. And the network computing revolution is essential, because 
mind
and brain are necessarily networks. A non-network-based computing paradigm 
can
never adequately support AI.
Java, the premier language of the new paradigm, is maturing into a central 
role
in server-side network apps. Exciting new Java-derivative technologies like
Objectspace Voyager are beginning to take off, and yet others like Jini and
Javaspaces are waiting in the wings. Electronic commerce is beginning to 
become
real, especially in the realm of business-to-business commerce; and online 
AI is
finally picking up speed -- embedded in websites like Excite and
barnesandnoble.com are sophisticated systems for learning and guessing user
profiles. We are just a few years away from the situation where the various
intelligent systems on the Net, like the ones inside Excite and
barnesandnoble.com, are learning from each other rather than existing as 
islands
of intelligence in a sea of inert text and data. The Internet, and the 
computing
environment in general, is poised for the advent of real artificial
intelligence. What is needed to make it happen is understanding on the part 
of
the technical community -- understanding of how the network of mind emerges 
from
the underlying network of computers.
For the last two years, since I founded the Silicon Alley start-up company
, I have been involved in building an internet-based AI system
called Webmind, which exemplifies the methodologies of AI design to be 
discussed
in the following pages. This project has been challenging and exciting, but 
has
left me fairly little time to reflect on the general lessons and principles
underlying the work I'm doing, and even less time to write down these 
lessons
and principles systematically. This book synthesizes some of the ideas that 
I
have had time to write about. Most of the chapters originated as informal
articles, written to clarify ideas to myself and distributed informally to
various e-mail co-workers and acquaintances; several pertain specifically to 
our
work
Chapter 1 reviews the network computing paradigm and its relevance to
intelligence; Chapters 2 and 3 follow up on this by exploring the "mind as
network" theme as it relates to the structure of the brain and the abstract
structure of cognition. Chapter 4 presents a general philosophical framework 
for
understanding Internet information space, beginning from philosopher Kent
Palmer's four ontological categories: Pure Being (data), Process Being
(programs), Hyper Being (network computing) and Wild Being (the next stage,
emergent Internet intelligence; the origin of the term Wild Computing). 
Chapters
5 and 6 discuss Internet AI from two very different perspectives: economics 
and
transpersonal consciousness (the World Wide Brain, encapsulating the 
essential
patterns of the human "collective unconscious"). Chapters 7 and 8 introduce 
and
discuss the Webmind network AI architecture that we are currently building 
at
a specific example of the new network AI paradigm. The
full details of Webmind are not entered into here, but the architecture is
painted in broad strokes, and it is indicated how Webmind has the potential 
to
fulfill the promise of next-phase, emergent AI. Finally, Chapter 9 describes
some ideas as to how a new agents communication protocol could allow a 
diverse
population of intelligent network entities, including Webminds, to
synergetically interact -- and hence form the practical infrastructure for a
global brain.
This is a diverse body of thinking, and I have not attempted to force a 
false
unity upon it. There is a real underlying coherence here, and I am too aware 
of
its depth and profundity to want to obscure it under a glib, premature
systematization. (As Nietzsche said, "The will to a system is a lack of
integrity.") Instead, , the book mirrors its subject matter, in that the 
topics
presented are not locked together into a rigid linear series, but represent
rather nodes in a network, interrelating with each other in numerous complex
ways. As the new computing develops, the pattern of interrelation between 
its
various aspects will become clearer and more aesthetic. We are all part of 
this
clarifying process.

                            Wild Computing

                              Chapter 1:
               The Network is the Computer is the Mind




1. Past and Future History of the Internet
The evolution of the Internet up till now can be divided, roughly speaking, 
into
three phases:
    Pre-Web. Direct, immediate interchange of small bits of text, via e-mail 
and
    Usenet. Indirect, delayed interchange of large amounts of text, visual
    images and computer programs, via ftp.
    Web. Direct, immediate interchange of images, sounds, and large amounts 
of
    text. Online publishing of articles, books, art and music. Interchange 
of
    computer programs, via ftp, is still delayed, indirect, and
    architecture-dependent.
    Network Computing. Direct, immediate interchange of animations and 
computer
    programs as well as large texts, images and sounds. Enabled by languages
    such as Java, the Internet becomes a real-time software resource.
    Intelligent agents traverse the web carrying out specialized intelligent
    operations for their owners.
The third phase, the Network Computing phase, is still in a relatively early
stage of development, driven the dissemination and development of the Java
programming language. However, there is an emerging consensus across the
computer industry as to what the ultimate outcome of this phase will be. For
many applications, people will be able to run small software "applets" from 
Web
pages, instead of running large, multipurpose programs based in their own
computers' hard disk drives. The general-purpose search engines of the Web 
phase
will evolve into more specialized and intelligent individualized Web 
exploration
agents. In short, the Web will be transformed from a "global book" into a
massively parallel, self-organizing software program of unprecented size and
complexity.
But, exciting as the Network Computing phase is, it should not be considered 
as
the end-point of Web evolution. I believe it is important to look at least 
one
step further. What comes after Network Computing, I propose, is the 
autopoietic,
emergently structured and emergently intelligent Web -- or, to put it
metaphorically, the World Wide Brain. The Network Computing environment is a
community of programs, texts, images, sounds and intelligent agents, 
interacting
and serving their own ends. The World Wide Brain is what happens when the
diverse population making up the Active Web locks into a global attractor,
displaying emergent memory and thought-oriented structures and dynamics not
programmed into any particular part of the Web. Traditional ideas from
psychology or computer science are of only peripheral use in understanding 
and
engineering a World Wide Brain. As we shall see in the following chapters,
however, ideas from complex systems science are considerably more relevant.
It may seem hasty to be talking about a fourth phase of Web evolution -- a 
World
Wide Brain -- when the third phase, network computing, has only just begun, 
and
even the second phase, the Web itself, has not yet reached maturity. But if 
any
one quality characterizes the rise of the Web, it is rapidity. The Web took 
only
a few years to dominate the Internet; and Java, piggybacking on the spread 
of
the Web, has spread more quickly than any programming language in history. 
Thus,
it seems reasonable to expect the fourth phase of Internet evolution to come
upon us rather rapidly, certainly within a matter of years rather than 
decades.
2. Network Computing
The history of networks as a computing paradigm is at first glance simple. 
There
were mainframes and terminals, then PC's and local-area networks ... and now
large-scale, integrated network computing environments, providing the 
benefits
of both mainframes and PC's and new benefits besides. This is a true story, 
but
far from the whole story. This view underestimates the radical nature of the
network computing paradigm. What network computing truly represents is a 
return
to the cybernetic, self-organization-oriented origins of computer science. 
It
goes a long way toward correcting the fundamental error committed in the 
1940's
and 1950's, when the world decided to go with a serial, von-Neumann style
computer architecture, to the almost total exclusion of more parallel,
distributed, brain-like architectures.
The move to network computing is not merely a matter of evolving engineering
solutions, it is also a matter of changing visions of computational
intelligence. Mainframes and PC's mesh naturally with the symbolic, 
logic-based
approach to intelligence; network computing environments, on the other hand,
mesh with a view of the mind as a network of intercommunicating, 
intercreating
processes. The important point is that the latter view of intelligence is 
the
correct one. From computing frameworks supporting simplistic and 
fundamentally
inadequate models of intelligence, one is suddenly moving to a computing
framework supporting the real structures and dynamics of mind.
For mind and brain are fundamentally network-based. The mind, viewed
system-theoretically, is far more like a network computing system than like 
a
mainframe-based or PC-based system. It is not based on a central system that
services dumb peripheral client systems, nor it is based on a huge host of
small, independent, barely communicating systems. Instead it is a large,
heterogenous collection of systems, some of which service smart peripheral
systems, all of which are intensely involved in inter-communication. In 
short,
by moving to a network computing framework, we are automatically supplying 
our
computer systems with many elements of the structure and dynamics of mind.
This does not mean that network computer systems will necessarily be
intelligent. But it suggests that they will inherently be more intelligent 
than
their mainframe-based or PC counterparts. And it suggests that researchers 
and
developers concerned with implementing AI systems will do far better if they
work with the network computing environment in mind. The network computing
environment, supplied with an appropriate operating system, can do half 
their
job for them -- allowing them to focus on the other half, which is
inter-networking intelligent agents in such a way as to give rise to the
large-scale emergent structures of mind. Far more than any previous 
development
in hardware, network computing gives us real hope that the dream of 
artificial
intelligence might be turned into a reality.
3. The Network Becomes the Computer
The history of computing in the late 20'th century is familiar, but bears
repeating and re-interpretation from a cybernetic, network-centric point of
view.
As everyone knows, over the past twenty years, we have seen the mainframes 
and
terminals of early computing replaced with personal computers. With as much
memory and processing power as early mainframes, and far better interfaces, 
PC's
have opened up computing to small businesses, homes and schools. But now, in 
the
late 90's, we are seeing a move away from PC's. The PC, many computer 
pundits
say, is on its way out, to be replaced with smart terminals called "network
computers" or NC's, which work by downloading their programs from central
servers.
An NC lacks a floppy drive; word processing and spreadsheet files, like
software, are stored centrally. Instead of everyone having their own copy 
of,
say, a huge word processor like MicroSoft Word, the word processor can 
reside on
the central server, and individuals can download those parts of the word
processor that they need. Routine computing is done by the NC; particularly
intensive tasks can be taken over by the central computer.
At the present time NC's are not a viable solution for home computer users, 
as
the bandwidth of modem or even ISDN connections is not sufficient for rapid
transmission of complex computer programs. For businesses, however, there 
are
many advantages to the network computing approach. Internal bandwidth, as
exploited by local-area networks, is often already high, and by providing
employees with NC's instead of PC's, an employer saves on maintenance and 
gains
in control.
And once cable modem or equivalently powerful technology becomes common, 
NC's
will be equally viable for the home user. The individual PC, with its 
limited
array of software, will quickly come to seem sterile and limiting. Why pay a 
lot
of money for a new game which you may not even like, when you can simply
download a game in half a minute, and pay for it on an as-you-play basis? 
When
it is almost as fast to download software from the Net as it is to extract 
it
from the hard drive of a PC, the advantages of PC's over NC's will quickly
become negative. The entire network becomes a kind of virtual hard drive, 
from
which programs or pieces thereof can be extracted on command.
These developments are widely perceived as ironic. After all, the move to 
PC's
was greeted with relief by almost everyone -- everyone but IBM and the other
large corporations who made their money from the mainframe market. What a 
relief
it was to be able to get a print-out on your desk, instead of waiting two 
hours
and then running down the the basement printer room. How wonderful to be 
able to
individualize one's computing environment. How exciting to compute at home. 
What
is ironic is that, in an NC context, these mainframes -- or newer machines 
of
even greater "grunt power" -- are becoming fashionable again.
The catch is, of course, that the context is different. In a modern network
computing environment, consisting of a central server and a large group of
peripheral NC's, no one has to wait two hours for their print-out to come 
out in
the basement. No one has to use a line printer interface, or even a text
interface on a screen. High-bandwidth communications conspire with advances 
in
hardware, operating system and graphical interface design to make the modern
network computing environment something entirely new: something 
incorporating
the advantages of the old mainframe approach and the advantages of PC's, 
along
with other advantages that have no precedent.
The grunt is there, the raw processing power of the central server; and so 
is
the efficiency of storing only one central copy of each program. But the
convenience of PC's is there too, provided by the modern possibility of 
spinning
off new copies of centrally-stored programs on the fly, and downloading them
into super-smart terminals -- terminals so smart that they are no longer 
merely
terminals but network computers.
From the point of view of the NC user, the network is like a PC with a giant
hard drive. The fact that this hard drive happens to be shared by other 
people
is barely even relevant. From the point of view of the central server 
itself, on
the other hand, the network is almost like an old-style mainframe system, 
the
only difference lying in the semantics of the input/ouput messages sent and
received, and in the speed with which these messages have to be sent. 
Instead of
just exchanging verbal or numerical messages with dumb terminals operated by
humans, the central server is now exchanging active programs with less 
powerful
but still quite capable satellite computers.
Finally, what is entirely new in modern network computing is the 
interconnection
of various central servers into their own large-scale network. This occurs 
in
the Internet, and within large organizations, it also occurs internally. 
This is
important because it means that no single server has to hold all the 
information
potentially required by its client NC's. The server is responsible for 
feeding
information to its NC's, but it may in some cases derive this information 
from
elsewhere, and simply serve as a "middleman." In other words, a higher-level
network of information is overlaid on a lower-level network of control. This
opens up the scope of information and applications available to computer 
users,
to a degree never before seen.
All this is exciting, tremendously exciting, but to computer industry 
insiders,
it is becoming almost hackneyed before it has even come to pass. "The 
Network is
the Computer" is a late 90's mantra and marketing slogan. And it is 
accurate;
the network should be, and increasingly really is the computer. What is now
inside the PC -- memory, processing, information -- will in the network
computing environment be stored all over the place. The overall computation
process is distributed rather than centralized, even for basic operations 
like
word processing or spreadsheeting.
What is not observed very often, however, is the relation between network
computing and the mind. As it turns out the mainframe approach to computing 
and
the PC approach to computing embody two different incorrect theories of
intelligence. The network computing approach embodies a more sophisticated
approach to intelligence, which, although still simplistic compared to the
structure of the human brain, is fundamentally correct according to 
principles
of cybernetics and cognitive science. As we move toward a worldwide network
computing environment, we are automatically moving toward computational
intelligence, merely by virtue of the structure of our computer systems, of 
the
logic by which our computer systems exchange and process information. In 
other
words, not only is the network is the computer, but it is the mind as well. 
This
leads to the new, improved slogan with which I have titled this chapter: The
network is the computer is the mind.
What does this slogan mean? It stands for a series of statements, of 
increasing
daring. At the most conservative end, one may say that the network computing
environment provides an ideal context within which to program true 
artificial
intelligence. At the most radical end, on the other hand, one might 
plausibly
argue that the widespread adoption of network computing will lead 
automatically
to the emergence of computational intelligence. The truth almost certainly 
lies
somewhere between these two extremes. Network computing provides a certain
degree of intelligence on its own, and it provides a natural substrate 
within
which to implement programs that display yet greater degrees of 
intelligence.
The network is the computer is the mind.
4. AI and the History of Hardware
It's a pleasant, familiar story: mainframes to PC's to network computer 
systems.
The whole thing has the inevitability of history about it. In fact, though,
there is more to these developments than is commonly discussed. Network
computing is, I believe, an inevitable occurrence -- but only in the 
abstract
sense of "computing with self-organizing networks of intercommunicating
processes." Networks are the best way to do artificial intelligence, and 
they
are also the best way to solve a huge variety of other problems in 
computing.
The particular form that network computing is assuming at the present time,
however -- computing with networks of interconnected serial digital 
computers --
is merely a consequence of the evolutionary path that computer hardware has
taken over the past half-century. This evolutionary path toward network
computing has in fact been highly ironic. Of all the directions it could
possibly have taken, computer technology has taken the one most antithetical 
to
artificial intelligence and self-organizing network dynamics. Even so, 
however,
the "network" archetype has proved irresistable, and is emerging in an
unforeseen way, out of its arch-enemy, the standard serial, digital computer
architecture.
During the last century, we have seen, there have emerged two sharply 
divergent
approaches to the problem of artificial intelligence: the cybernetic 
approach,
based on emulating the brain and its complex processes of self-organization, 
and
the symbolic approach, based on emulating the logical and linguistic 
operations
of the conscious mind.
Neither of these approaches is perfect. It has become increasingly clear in
recent years, however, that whereas the symbolic approach is essentially
sterile, the cybernetic approach is fertile. The reason for this is quite
simple: the mind/brain really is a large, specially-structured, 
self-organizing
network of processes; it really is not a rule-based logical reasoning 
system.
We do, of course, carry out behaviors that can be described as logical 
reasoning
-- but what we are doing in these cases, while it can be described in terms 
of
rule-following to some degree of accuracy, is not really rule-following.
Ignoring the self-organizing dynamics underlying apparent rule-following --
taking logical reasoning out of the context of intelligently perceived
real-world situations, and out of the context of illogical, unconscious 
mental
dynamics -- results in an endless variety of philosophical paradoxes and
practical obstacles. The result is that symbolic AI systems are inflexible 
and
uncreative -- not much more "intelligent," really, than a program like
Mathematica which does difficult algebra and calculus problems by applying
subtle mathematical methods.
The cybernetic approach to AI, for all its flaws, has a far clearer path to
success. Building larger and larger, more and more intelligently structured
self-organizing networks, we can gradually build up to more and more 
mind-like
structures. Neural networks are one of the more popular implementations of 
the
cybernetic approach, but not the only one. One can also think in terms of
genetic algorithms, or more abstract systems of "software agents" -- the 
point
is the emphasis on creative, complex network dynamics, rather than 
deterministic
interactions between systems of logical, rational rules.
In the beginning, back in the 1940's and 50's, these two different 
approaches to
AI were tied in with different approaches to computer hardware design: 
parallel,
distributed analog design versus serial, digital design. Each theory of
"computational mind" matched up naturally with a certain approach to
"computational brain."
In the parallel, distributed, analog computer, many things happen at each 
point
in time, at many different points in space. Memory is stored all over, and
problem-solving is done all over. Memory is dynamic and is not fundamentally
distinct from input/output and processing. Furthermore, the basic units of
information are not all-or-nothing switches, but rather continuous signals, 
with
values that range over some interval of real numbers. This is how things 
happen
in the brain: the brain is an analog system, with billions of things 
occurring
simultaneously, and with all its different processes occurring in an 
intricately
interconnected way.
On the other hand, in the serial, digital computer, commonly known as the 
von
Neumann architecture, there is a central processor which does one thing at 
each
time, and there is a separate, inert memory to which the central processor
refers. On a hardware level, the von Neumann design won for practical
engineering reasons (and not for philosophical reasons: von Neumann himself 
was
a champion of neural-net-like models of the mind/brain). By now it is 
ingrained
in the hardware and software industries, just as thoroughly as, say, 
internal
combustion engines are ingrained in the automobile industry. Most likely, 
every
computer you have ever seen or heard of has been made according to the von
Neumann methodology.
The cybernetic approach to artificial intelligence, however, has survived 
the
dominance of the von Neumann architecture, moving to a methodology based
primarily on serial digital simulations of parallel distributed analog 
systems.
The fact that cybernetic AI has survived even in a totally hostile computing
hardware environment is a tribute to the fundamental soundness of its 
underlying
ideas. One doubts very much, on the other hand, whether symbolic AI would 
have
ever become dominant or even significant in a computing environment 
dominated by
neural net hardware. In such an environment, there would have been strong
pressure to ground symbolic representation in underlying network dynamics. 
The
whole project of computing with logic, symbolism and language in a formal,
disembodied way, might never have gotten started.
Today, many of us feel that the choice of the von Neumann architecture may 
have
been a mistake -- that computing would be far better off if we had settled 
on a
more brain-like, cybernetics-inspired hardware model back in the 1940's. The
initial engineering problems might have been greater, but they could have 
been
overcome with moderate effort, and the billions of dollars of money spent on
computer R&D in the past decades would have been spent on brainlike 
computers
rather than on the relatively sterile, digital serial machines we have 
today. In
practice, however, no alternate approach to computer hardware has yet come 
close
to the success of the von Neumann design. All attempts to break the von 
Neumann
hegemony have met with embarrassing defeat.
Numerous parallel-processing digital computers have been constructed, from 
the
restricted and not very brainlike "vector processors" inside Cray
supercomputers, to the more flexible and AI-oriented "massively parallel"
Connection Machines manufactured by Thinking Machines, Inc.. The Cray 
machines
can do many things at a each time step, but they all must be of the same 
nature.
This approach is called SIMD, "single-instruction, multiple dataset": it is
efficient for scientific computation, and some simple neural network models, 
but
not for sophisticated AI applications. The Thinking Machines computers, on 
the
other hand, consist of truly independent processors, each of which can do 
its
own thing at each time, using its own memory and exchanging information with
other processors at its leisure. This is MIMD, "multiple instruction, 
multiple
dataset"; it is far more reminiscent of brain structure. The brain, at each
time, has billions of "instructions" and billions of "data sets"!
These parallel digital machines are exciting, but, for a combination of
technical and economic reasons, they have not proved as cost-effective as
networks of von Neumann computers. They are used almost exclusively for
academic, military and financial research, and even their value in these 
domains
has been dubious. Thinking Machines Inc. has gone bankrupt, and is trying to
re-invent itself as a software company; their flagship product, GlobalWorks, 
is
a piece of low-level software that allows networks of Sun workstations to 
behave
as if they were Connection Machines (Sun workstations are high-end 
engineering
computers, running the Unix operating system and implementing, like all 
other
standard contemporary machines, the serial von Neuman model).
With GlobalWorks, all the software tools developed for use with the 
Connection
Machines can now be used in a network computing environment instead. There 
is a
serious loss of efficiency here: instead of a network of processors 
hard-wired
together inside a single machine, one is dealing with a network of 
processors
wired together by long cables, communicating through complex software 
protocols.
However, the economies of scale involved in manufacturing engineering
workstations means that it is actually be more cost-effective to use the 
network
approach rather than the parallel-machine approach, even though the latter 
is
better from a pure engineering point of view.
Of even greater interest than the massively parallel digital Connection 
Machine
is the analog, neural net based hardware being produced by several 
contemporary
firms, -- radical, non-binary computer hardware that is parallel and 
distributed
in nature, mixing up multiple streams of memory, input/output and processing 
at
every step of time. For instance, the Australian company Formulab 
Neuronetics,
founded in the mid-80's by industrial psychologist Tony Richter, 
manufactures
analog neural network hardware modeled fairly closely on brain structure. 
The
Neuronetics design makes the Connection Machine seem positively 
conservative.
Eschewing traditional computer engineering altogether, it is a a hexagonal
lattice of "neuronal cells," each one exchanging information with its 
neighbors.
There are perceptual neurons, action neurons, and cognitive neurons, each 
with
their own particular properties, and with a connection structure loosely
modelled on brain structure. This technology has proved itself in a variety 
of
process control applications, such as voice mail systems and internal 
automotive
computers, but it has not yet made a splash in the mainstream computer 
industry.
By relying on process control applications for their bread and butter,
Neuronetics will hopefully avoid the fate of Thinking Machines. But the 
ultimate
ambition of the company is the same: to build an ultra-high-end 
supercomputer
that, by virtue of its size and its brainlike structure, will achieve
unprecedented feats of intelligence.
As of now, this kind of neural net hardware is merely a specialty product. 
But I
suspect that, as PC's fade into history, these analog machines will come to 
play
a larger and larger role in the world. In the short run, we might see
special-purpose analog hardware used in the central servers of computer
networks, to help deal with the task of distributing information amongst 
various
elements of a network computing environment. In the long run, one might see
neurocomputers joining digital computers in the worldwide computer network, 
each
contributing their own particular talents to the overall knowledge and
processing pool.
The history of AI and computer hardware up till now, then, is a somewhat sad
one, with an ironic and optimistic twist at the end. The dominant von 
Neumann
architecture is patently ill-suited for artificial intelligence. Whether it 
is
truly superior from the point of view of practical engineering is difficult 
to
say, because of the vast amount of intelligence and resources that has been
devoted to it, as compared to the competitors. But it has incredible 
momentum --
it has economies of scale on its side, and it has whole industries, with 
massive
collective brainpower, devoted to making it work better and better. The 
result
of this momentum is that alternate, more cybernetically sophisticated and
AI-friendly visions of computing are systematically squelched. The 
Connection
Machine was abandoned, and the Neuronetics hardware is being forced to earn 
its
keep in process control. This is the sad part. As usual in engineering, 
science,
politics, and other human endeavors, once a certain point of view has 
achieved
dominance, it is terribly difficult for anything else to gain a foothold.
The ironic and possibly optimistic part, however, comes now and in the near
future. Until now, brainlike parallel architectures have been squelched by
serial von Neumann machines -- but the trend toward network computing is an
unexpected and unintentional reversal of this pattern. Network computing is
boldly cybernetic -- it is brainlike computer architecture emerging out of 
von
Neumann computer architecture. It embodies a basic principle of Oriental 
martial
arts: when your enemy swings at you, don't block him, but rather position
yourself in such a way that his own force causes him to flip over.
The real lesson, on a philosophical level, may be that the structure of 
brain
and intelligence is irresistable for computing. We took a turn away from it 
way
back in the 1940's, rightly or wrongly, but now we are returning to it in a
subtle and unforeseen way. The way to do artificial intelligence and other
sophisticated computing tasks is with self-organizing networks of
intercommunicating processes -- and so, having settled on computer hardware
solutions that do not embody self-organization and intercommunication, we 
are
impelled to link our computers together into networks that do.
5. Issues of Scale in Artificial Intelligence
Another way to look at these issues is to observe that, historically, the
largest obstacle to progress in AI has always been scale. Put simply, our 
best
computers are nowhere near as powerful as a chicken's brain, let alone a 
human
brain. One is always implementing AI programs on computers that, in spite of
special-purpose competencies, are overall far less computationally able than 
one
really needs them to be. As a consequence, one is always presenting one's AI
systems with problems that are far, far simpler than those confronting human
beings in the course ordinary life. When an AI project succeeds, there is 
always
the question of whether the methods used will "scale-up" to problems of more
realistic scope. And when an AI project fails, there is always the question 
of
whether it would have succeeded, if only implemented on a more realistic 
scale.
In fact, one may argue on solid mathematical grounds that intelligent 
systems
should be subject to "threshold effects," whereby processes that are 
inefficient
in systems below a certain size threshold, become vastly more efficient once 
the
size threshold is passed.
Some rough numerical estimates may be useful. The brain has somewhere in the
vicinity of 100,000,000,000 -10,000,000,000,000 neurons, each one of which 
is
itself a complex dynamical system. There is as yet no consensus on how much 
of
the internal dynamics of the neuron is psychologically relevant. Accurate,
real-time models of the single neuron are somewhat computationally 
intensive,
requiring about the computational power of a low-end Unix workstation. On 
the
other hand, a standard "formal neural-network" model of the neuron as a 
logic
gate, or simple nonlinear recursion, is far less intensive. A typical
workstation can simulate a network of hundreds of formal neurons, evolving 
at a
reasonable rate.
Clearly, whatever the cognitive status of the internal processes of the 
neuron,
no single computer that exists today can come anywhere near to emulating the
computational power of the human brain. One can imagine building a 
tremendous
supercomputer that would approximate this goal. However, recent history 
teaches
that such efforts are plagued with problems. A simple example will 
illustrate
this point. Suppose one sets out, in 1995, to build a massively parallel AI
machine by wiring together 100,000 top-of-the-line chips. Suppose the 
process of
design, construction, testing and debugging takes three years. Then, given 
the
current rate of improvement of computer chip technology (speed doubles 
around
every eighteen months), by the time one has finished building one's machine 
in
1998, its computational power will be the equivalent of only 25,000
top-of-the-line chips. By 2001, the figure will be down to around 6,500.
Instead of building a supercomputer that is guaranteed to be obsolete by the
time it is constructed, it makes more sense to utilize an architecture which
allows the continuous incorporation of technology improvements. One requires 
a
highly flexible computer architecture, which allows continual upgrading of
components, and relatively trouble-free incorporation of new components, 
which
may be constructed according to entirely new designs. Such an architecture 
may
seem too much to ask for, but the fact is that it already exists, at least 
in
potential form. The Web has the potential to transform the world's 
collective
computer power into a massive, distributed AI supercomputer.
Once one steps beyond the single-machine, single-program paradigm, and views 
the
whole Web as a network of applets, able to be interconnected in various 
ways, it
becomes clear that, in fact, the Web itself is an outstanding AI 
supercomputer.
Each Web page, equipped with Java code or something similar, is potentially 
a
"neuron" in a world-wide brain. Each link between one Web page and another 
is
potentially a "synaptic link" between two neurons. The neuron-and-synapse
metaphor need not be taken too literally; a more appropriate metaphor for 
the
role of a Web page in the Web might be the neuronal group (Edelman, 1988). 
But
the point is that Java, in principle, opens up the possibility for the Web 
to
act as a dynamic, distributed cognitive system. The Web presents an
unprecedentedly powerful environment for the construction of large-scale
intelligent systems. As the Web expands, it will allow us to implement more 
and
more intelligent World Wide Brain, leading quite plausibly, in the not too 
far
future, to a global AI brain exceeding the human brain in raw computational
power.
6. Hardware-Software Synergy
What is the relation between computer system designs and approaches to
artificial intelligence? We have already observed that all the mainstream
computing designs of the last half-century, from mainframes to PC's to 
network
computing, support symbolic AI over cybernetic AI: they are serial and 
digital
rather than analog and parallel. But there is more to the story than this. 
On a
more detailed level than the symbolic/cybernetic dichotomy, different 
computer
architectures embody different visions of computational intelligence, 
different
models of mind.
The vision of mind suggested by the mainframe computer is quite different 
from
the vision of mind suggested by the PC, which is in turn quite different 
from
the vision of mind suggested by the modern network computing system. All the
different digital computing frameworks are "universal computers," so that 
one
can, in principle, run any kind of AI algorithm on either a mainframe, a PC 
or a
network computing system. A computing system can be used to run AI 
algorithms
embodying a philosophy of intelligence entirely different from the 
philosophy of
intelligence embodied in its structure -- this is what happens when, for 
example
neural networks are simulated on PC's or mainframes. But when one does this,
when one works with AI algorithms that are "conceptually mismatched" to 
their
hardware, one pays a price. One loses efficiency, ease of experimentation, 
and
conceptual integrity.
When one's algorithm conceptually matches one's hardware, one finds that
"following the hardware" leads to all sorts of new twists and innovations. 
When
there is a mismatch, on the other hand, the suggestions implicitly made by 
the
hardware will more than often be wrong, or at least will push one in 
directions
alien to the ideas that one started out with. This may be seen in many of 
the
recent innovations to neural network AI, derived from experimenting with 
neural
nets via simulations on serial digital machines. The neural net learning
algorithms that work best in serial digital simulations may not be the same 
ones
that would work best in a more neural-net oriented, parallel analog hardware
setting. In fact, many recent neural network techniques make no sense 
whatsoever
as massively parallel, self-organizing analog algorithms: they are merely
neural-net-inspired serial computing algorithms, consequences of the 
adaptation
of neural networks to the unnatural environment of digital serial machines. 
This
kind of development, spurred on directly by hardware/algorithm mismatch, is
useful from a short-term perspective, but it does not add much to the 
long-term
understanding of intelligence and its computer implementation. If
experimentation were taking place on Neuronetics or similar technology, most 
of
the recent neural net algorithms would not exist; instead there would be 
other
new approaches, ones fitting in better with parallel analog computing and 
hence
with the underlying vision of the neural network approach.
A good analogy to this hardware-software synergy in AI may be found in 
music.
The hardware is the instrument and the software is the song played on the
instrument. One can play any song on any instrument, but some song go more
naturally with some instruments. Playing a punk rock song on a guitar, one 
finds
that the "hardware" naturally enhances the "software" -- playing around with 
the
song on the guitar, one is led to different timings, slight modifications 
here
and there, maybe even new, related melodies. On the other hand, playing the 
same
song on a violin, one can hit all the notes perfectly accurately, but one 
may
find that one has turned something inherently very simple into something 
complex
and awkward to play -- "efficiency" has degraded. Also, one will find that 
the
physically natural variations and improvisations on the melody no longer 
sound
so musically natural -- they are no longer punk tunes, because the violin is 
not
a punk instrument. Ease of experimentation has decreased.
Orchestration, the mapping of complex compositions onto sets of instruments, 
is
essential in classical music, and it is even subtler in jazz, because many 
of
the instruments must improvise around the melodies and rhythms of the song, 
and
their improvisations must be simultaneously physically logical in terms of 
the
structure of the instrument and musically logical in the context of the song 
as
a whole. Designing a complex AI system on a complex system of computer 
hardware
is a somewhat similar process, the difference being that one is 
orchestrating,
not human beings producing sounds, but mechanical devices producing various
types of information, and monitored and modified by human beings.
The mainframe computer, with its central control unit and its dumb 
peripherals,
embodies a simple and powerful model of mind: mind as a monolithic cognitive
center, carrying out thinking and reasoning processes, collecting raw data 
from
unintelligent peripheral processes and providing raw data to unintelligent 
motor
processes. This is the kind of AI program that is most naturally programmed 
on
mainframes. Of course, one could always deviate from this implicit 
suggestion of
the computer architecture, and program complex, sophisticated perceptual and
motor processes, together with cognitive processes. But this would be an
unmotivated move, an "artificial" gesture with regard to the system one is
working within.
A network computing environment, on the other hand, embodies a vision of 
mind as
a large, heterogenous system of interacting agents, sending messages to each
other and processing information in a variety of different ways. The modern
network computing environment is not ideal for artificial intelligence -- it
would be much better if our PC's and engineering workstations were powered 
by
Neuronetics or similar hardware instead of serial digital hardware. But even 
so,
there is reason to believe that the overall structure of mind/brain is more
important than the microscopic details by which individual processes are 
carried
out. The point is that network computing, unlike previously popular 
computing
methodologies, embodies a vision of mind as a network of autonomous, 
interacting
agents, which is precisely the right vision of mind to have.
Less crucial than the hardware/software synergy, but still important, is the
software/operating-system synergy. In particular, in order to turn network
computing into network AI, one needs an appropriate operating system (OS) -- 
a
cybernetically wise, intelligence-friendly system for communicating between
intelligent mind-network-embodying software, and the network computing
environment.
The operating system is the low-level computer software that mediates
communications between hardware and other computer software. Even in the 
realm
of ordinary serial digital computers, there is a variety of operating 
systems;
and two different OS's, running on the same system, can give the feeling of 
two
different machines. For instance, two of the most popular OS's today are the
Microsoft operating systems, DOS and Windows, and the Unix operating system,
used on engineering workstations. These competing systems do essentially the
same things, and are now invading each other's territory: Windows NT is a
version of Windows that runs on engineering workstations, and Linux is a 
version
of Unix that runs on PC's. But even so, they have drastically different
"flavors."
What one needs, in order to exploit the potential of network computing -- 
for AI
in particular and for other sophisticated applications as well -- is to have 
a
kind of network operating system. Not an OS for any particular hardware, but 
an
OS for networks of diverse, intercommunicating hardware. This is different 
from
the conventional concept of an "operating system," but it is well within the
common-language meaning of the phrase. What one needs is a system by which 
AI
programs (and other programs) can operate the computer network as a network 
of
agents, a network of processes, continually changing and communicating with 
each
other -- rather than as a collection of von Neumann machines, each with
particular properties to be taken into account, and sending messages of
unrelated types that all need to be treated differently. One needs a network 
OS
that leaves one blissfully blind to the details of managing different types 
of
computers, and communicating different types of messages, leaving one in a
mindlike world of abstract entities communicating abstract information and
modifying themselves accordingly.
Java goes a fair way toward this goal of an agent-oriented network OS. Java
allows one to write network- aware computer programs that will run on any 
modern
computer, and it streamlines the processes of communication and interaction 
to
an unprecedented extent. However, it falls a fair way short of providing the
tools required to truly bring process-network-based AI and network computing
together. Much creative engineering on the part of Java and other software
developers is needed, if the potential embodied in "The network is the 
computer
is the mind" is to be brought to fruition.

                              Chapter 2:

                    Mind as Network and Emergence

1. Introduction
Mind consists of emergent pattern -- but emergent pattern is necessarily
emergent from something. At present, the intelligence we know in the 
greatest
detail is our own, which emerges from the human brain. The emergence of
intelligence from the Internet will not mimic the emergence of intelligence 
from
the brain in every detail, but there will be many commonalities. In this 
chapter
I will discuss the brain as a self-organizing network, and the mind as a
self-organizing network, and how the latter network emerges from the former.
This serves us two purposes in our study of Internet intelligence: it gives 
us a
coherent model of emergent intelligence, to be applied to the Net; and it 
gives
us an example of emergent intelligence coming out of a particular physical
system.
2. Neural Networks
The premier computer models of the brain are "neural network" models, which 
bear
many interesting similarities to modern computer networks -- they are
asychronous systems carrying out massively distributed processing -- and so 
are
highly germane to the topics of this book. Mind as emergent from neural 
networks
is different from mind as emergent from computer networks, but there are 
more
similarities between the two than one would expect to find, say, between 
mind as
emergent from neural networks and mind as emergent from an intelligent ocean 
or
an intelligent cloud of gas. Network-ness carries numerous specialized
properties that are important for the emergence of mind.
The "neural" in "neural network" refers to the nerve cell or neuron -- a 
type of
cell which exists throughout the body for the purpose of registering 
sensations,
but exists in greatest bulk within the brain. The purpose of most neurons in 
the
brain is not to directly register sensations from the outside world, but to
register sensations from other neurons in the brain. I.e., the brain is a 
huge
sense organ whose goal is to sense itself via its neurons. Neurons are not 
the
only brain cells; in fact, they are greatly outnumbered by glia. However, 
many
neuroscientists (Edelman, 1987; Rose and Dobson, 1985) believe that the key 
to
mental process lies in the large-scale behavior of networks of neurons; and 
I do
not disagree.
A neuron consists of a cell body with a long, narrow axon emerging from one 
end,
and a large number of branches called dendrites snaking out in all 
directions.
The dendrites are inputs _ they receive electrical signals from other 
neurons.
The cell body periodically generates _ "fires" _ a new electrical impulse 
based
on these input signals. After it fires, it needs to "recover" for a while 
before
it can fire again; and during this period of recovery it basically ignores 
its
input.
The axon carries the electrical impulse from the cell body to the dendrites 
and
cell bodies of other neurons. The points at which signals pass from one 
neuron
to another are called synapses, and they come in two different forms _
excitatory and inhibitory. When an impulse arrives through an excitatory
synapse, it encourages the receiving neuronto fire. When an impulse arrives
through an inhibitory synapse, it discourages the receiving neuron from 
firing.
Each synapse has a certain conductance or "weight" which affects the 
intensity
of the signals passing through it. For example, suppose excitatory synapse A 
has
a larger "weight" than excitatory synapse B, and the same signal passes 
through
both synapses. The signal will be more intense at the end of A than at the 
end
of B.
Roughly speaking, a recovered neuron fires if, within the recent past, it 
has
received enough excitatory input and not too much inhibitory input. The 
amount
of the past which is relevant to the decision whether or not to fire is 
called
the period of latent addition. How much excitation is "enough," and how much
inhibition is "too much," depends upon the threshold of the neuron. If the
threshold is minimal, the neuron will always fire when its recovery period 
is
over. If the threshold is very high, the neuron will only fire when nearly 
all
of its excitatory synapses and virtually none of its inhibitory synapses are
active.
Mathematically, the rule that tells a neuron when to fire can be modeled 
roughly
as follows. Let wi be the weight of the i-th synapse which inputs to the 
neuron,
where positive weights denote excitatory connections and negative weights 
denote
inhibitory connections. Let xi(t) denote the signal coming into the neuron
through the i-th synapse, at time t, where the time variable is assumed to 
be
discrete. Let P be the period of latent addition, and let R be the recovery
period. Then the total relevant input to the neuron at time t is the sum, 
over
all s so that
max{t-P,r+R} < s < t, of wi(s)xi(s), where r is the time at which the neuron
last fired. Where T is the threshold of the neuron, a neuron fires at time t 
if
its total relevant input exceeds T. When a neuron fires, its output is 1; 
when
it does not fire, its output is 0.
A "neural network," then, is a network of interconnected neurons firing
according to this rule. This is a greatly oversimplified model of the brain: 
it
gives no role to other brain cells such as glia, and it completely ignores 
all
the chemistry that mediates actual neural interaction. In the brain the 
passage
of an electrical signal from one neuron to another is not exactly analogous 
to
the passage of electricity across a wire. This is because most neurons that 
are
"connected" do not actually touch. What usually happens when a signal passes
from neuron A to neuron B is that the dendrites of neuron A build up a 
charge
which causes certain chemicals called neuro-transmitters to carry that 
charge to
the dendrites of neuron B. The neural network model ignores all the 
subtleties
of this process.
So, to consider the brain as a neural network is an oversimplification. The
"neural networks" which are now so popular in computer science and 
electrical
engineering (Garrido,1990; Kawato et al., 1987; Goldberg et al., 1988; 
Hopfield
and Tank, 1985) are usually simplified even further. It is generally assumed
that the period of latent addition is 1 time step, and the recovery period 
is 0
time steps.
Such simplified "neural networks" have proven themselves effective at a 
number
of difficult practical problems _ combinatorial optimization, associative
memory, (Hopfield and Tank, 1980), pattern recognition (Grossberg, 1987) and
robotic control (Goldberg et al., 1988), to name a few. Mathematically, they 
are
similar to the physicists' spin glasses (Garrido, 1990). Everyone realizes 
that
these networks are mediocre brain models, but the connection with 
neuroscience
is tantalizing nonetheless.
For example, the well known Hopfield network (Hopfield, 1980) uses 
first-order
neurons to minimize functions. To explain the idea behind this network, let 
us
define a state of a network of n neurons as a binary sequence a1a2...an. A 
state
is periodic with period p if whenever, xi(t)=ai for i=1,...,n, xi(t+p)=ai 
for
i=1,...,n. If a state is periodic with period 1, then it is an equilibrium.
Given a certain function f from binary sequences to real numbers, Hopfield's
approach is to define a network whose equilibrium states are local minima of 
f,
and which has no periodic points besides its equilibria. Then one may set 
the
state of the network at time zero equal to any random binary sequence, and
eventually the values xi(t) will settle into one of the equilibria. Of 
course,
if a function has many local minima, the corresponding network will have 
many
equilibria.
Formal neural networks like the Hopfield net, and feedforward 
backpropagation
nets, are useful for computer and behavioral scientists, but unsatisfying to
biological theorists. In recent years, several inventive biologists have 
sought
to bridge thelarge gap between formal neural networks and actual brains. In 
my
opinion, the most impressive of these efforts is Edelman's (1987) theory of
neuronal group selection, or "Neural Darwinism."
The starting point of Neural Darwinism is the observation that neuronal 
dynamics
may be analyzed in terms of the behavior of neuronal groups. The strongest
evidence in favor of this conjecture is physiological: many of the neurons 
of
the neocortex are organized in clusters, each one containing say 10,000 to
50,000 neurons each.
Once one has committed oneself to looking at groups, the next step is to ask 
how
these groups are organized. A map, in Edelman's terminology, is a connected 
set
of groups with the property that when one of the inter-group connections in 
the
map is active, others will often tend to be active as well. Maps are not 
fixed
over the life of an organism. They may be formed and destroyed in a very 
simple
way: the connection between two neuronal groups may be "strengthened" by
increasing the weights of the neurons connecting the one group with the 
other,
and "weakened" by decreasing the weights of the neurons connecting the two
groups.
Formally, we may consider the set of neural groups as the vertices of a 
graph,
and draw an edge between two vertices whenever a significant proportion of 
the
neurons of the two corresponding groups directly interact. Then a map is a
connected subgraph of this graph, and the maps A and B are connected if 
there is
an edge between some element of A and some element of B. (If for "map" one 
reads
"program," and for "neural group" one reads "subroutine," then we have a 
process
dependency graph as drawn in theoretical computer science.)
This is the set-up, the context in which Edelman's theory works. The meat of 
the
theory is the following hypothesis: the large-scale dynamics of the brain is
dominated by the natural selection of maps. Those maps which are active when
good results are obtained are strengthened, those maps which are active when 
bad
results are obtained are weakened. And maps are continually mutated by the
natural chaos of neural dynamics, thus providing new fodder for the 
selection
process. By use of computer simulations, Edelman and his colleague Reeke 
have
shown that formal neural networks obeying this rule can carry out fairly
complicated acts of perception.
3. The Psynet Model of Mind
Supposing that we accept, provisionally, the neural network as a model of 
the
brain. We may then ask: how does the mind emerge from the brain? In fact, 
even
if neural networks are fundamentally flawed as brain models, they may still 
be
useful in inspiring us to address the question of emergence in interesting 
ways.
The main point is -- the mind is a network of one kind, and the physical
substrate underlying the mind, i.e. the brain, is a network of another kind. 
The
emergence of mind from reality is the emergence of a network from a network.
In a series of publications over the past half-decade (Goertzel, 1993, 
1993a,
1994, 1997), I have constructed a novel complex systems model of mind, which 
I
now call the psynet model, aimed at answering precisely this question. The
psynet model envisions the mind as an agent system, involving special kinds 
of
agents called "magicians", which are distinguished from "agents" as 
generally
discussed in computer science (and discussed in the following chapter) by 
the
fact that their main activity is transforming other magicians. Of course, 
all
agents that interact with other agents can be thought of as magicians; the
distinction between magicians and run-of-the-mill agents is fuzzy, just as 
is
the distinction between, say, intelligent agents and run-of-the-mill agents. 
The
question, in determining whether a system of agents is a "magician system," 
is
how transformative the interactions are.
So, a magician system consists, quite simply, of a collection of entities 
called
"magicians" which, by acting on one another, have the power to cooperatively
create new magicians. Certain magicians are paired with "antimagicians,"
magicians which have the power to annihilate them. According to the psynet
model, mind is a pattern/process magician system. It is a magician system 
whose
magicians are concerned mainly with recognizing and forming patterns.
Crucial to this picture is the phenomenon of mutual intercreation, or
autopoiesis. Systems of magicians can interproduce. For instance, a can 
produce
a, while b produces a. Or a and b can combine to produce c, while b and c
combine to produce a, and a and c combine to produce b. The number of 
possible
systems of this sort is truly incomprehensible. But the point is that, if a
system of magicians is mutually interproducing in this way, then it is 
likely to
survive the continual flux of magician interaction dynamics. Even though 
each
magician will quickly perish, it will just as quickly be re-created by its
co-conspirators. Autopoiesis creates self-perpetuating order amidst flux.
The magician system view of mind is merely one small step beyond Neural
Darwinism. It says: let us take as the fundamental entities, the states of
neuronal modules. Let us envision these states, patterns localized in space 
and
time, as interacting with one another, transforming each other into 
different
states. Each state lasts for a brief period of time, giving rise to other
states, and may then arise again as a result of this activity. The biologist
does not wish to think in terms of states, as he prefers to deal with 
tangible,
persistent structures, but it is states that make up the experienced mind, 
and
it is the dynamics of states that we must come to grips with if we wish to
understand the mind -- and the parallels between the human mind and the
Internet. The Net does not have neuronal modules, in any exact sense, but 
yet
the self-organizing system of states that emerges from the Net is 
structurally
and dynamically very similar to the system of states emerging from the human
brain. There is a parallel on the level of magician systems.
Some systems of magicians might be unstable; they might fall apart as soon 
as
some external magicians start to interfere with them. But others will be 
robust;
they will survive in spite of external perturbations. These robust magician
systems are what I call autopoietic systems. The psynet model posits that
thoughts, feelings and beliefs are autopoietic. They are stable systems of
interproducing pattern/processes. In Chaotic Logic, autopoietic 
pattern/process
magician systems are called structural conspiracies, a term which reflects 
the
mutual, conspiratorial nature of autopoiesis, and also the basis of
psychological autopoiesis in pattern (i.e. structure) recognition. A 
structural
conspiracy is an autopoietic magician system whose component processes are
pattern/processes.
But structural conspiracy is not the end of the story. The really remarkable
thing is that, in psychological systems, there seems to be a global order to
these autopoietic subsystems. The central claim of the psynet model is that, 
in
order to form a functional mind, these structures must spontaneously
self-organize into larger autopoietic superstructures. And perhaps the most
important such superstructure is a sort of "monster attractor" called the 
dual
network.
The dual network, as its name suggests, is a network of pattern/processes 
that
is simultaneously structured in two ways. The first kind of structure is
hierarchical. Simple structures build up to form more complex structures, 
which
build up to form yet more complex structures, and so forth; and the more 
complex
structures explicitly or implicitly govern the formation of their component
structures. The second kind of structure is heterarchical: different 
structures
connect to those other structures which are related to them by a sufficient
number of pattern/processes. Psychologically speaking, the hierarchical 
network
may be identified with command-structured perception/control, and
theheterarchical network may be identified with associatively structured 
memory.
(While the dual network is, intuitively speaking, a fairly simple thing, to 
give
a rigorous definition requires some complex constructions and arbitrary
decisions. One approach among many is given in From Complexity to 
Creativity)
A psynet, then, is a magician system which has evolved into a dual network
structure. Or, to place the emphasis on structure rather than dynamics, it 
is a
dual network whose component processes are magicians. The central idea of 
the
psynet model is that the psynet is necessary and sufficient for mind. And 
this
idea rests on the crucial assumption that the dual network is autopoietic 
for
pattern/process magician dynamics.
The psychological meaning of these structures has been elaborated in detail 
in
previous publications and so will not be dwelt on here. Suffice to say that 
the
hierarchical network is prevalent in visual perception, and is notable in 
the
structure of the visual cortex as well as via perceptual psychology. It is 
also
present, though not quite as blatantly, in such areas as the composition of
motor actions, and language processing. The heterarchical network, on the 
other
hand, is prevalent in the processing of taste and smell, and in the 
functioning
of associative semantic and episodic memory.
4. Evolution and Autopoiesis in the Network of Mind
The dynamics of the dual network may be understood as a balance of two 
forces.
There is the evolutionary force, which creates new forms, and moves things 
into
new locations. And there is the autopoietic force, which retains things in 
their
present form. If either one of the two forces is allowed to become overly
dominant, the dual network will break down, and become excessively unstable, 
or
excessively static and unresponsive.
Of course, each of these two "forces" is just a different way of looking at 
the
basic magician system dynamic. Autopoiesis is implicit in all attractors of
magician dynamics, and evolutionary dynamics is a special case of magician
dynamics, which involves long transients before convergence, and the 
possibility
of complex strange attractors.
To fully understand the role of evolution here, one must go beyond the 
standard
Darwinian notion of "fitness", and measure fitness in terms of emergent 
pattern.
In The Evolving Mind, I define the structural fitness of an organism as the 
size
of the set of patterns which synergetically emerge when the organism and its
environment are considered jointly. If there are patterns arising through 
the
combination of the organism with its environment, which are not patterns in 
the
organism or the environment individually, then the structural fitness is 
large.
Perhaps the easiest illustration is camouflage -- there the appearance of 
the
organism resembles the appearance of the environment, generating the 
simplest
possible kind of emergent pattern: repetition. But symbiosis is an even more
convincing example. The functions of two symbiotic organisms match each 
other so
effectively that it is easy to predict the nature of either one from the 
nature
of the other.
The "environment" of a process in the psynet is simply its neighbors in the
network. So the structural fitness of a process in the psynet is the amount 
of
pattern that emerges between itself and its neighbors. But, at any given 
time,
the probability of a process not being moved in the network is positively
correlated with its degree of "fit" in the associative memory. This degree 
of
fit is precisely the thestructural fitness! So, survival in current position 
is
correlated with structural fitness with respect to immediate environment; 
and
thus, the psynet evolves by natural selection. Furthermore, by the same 
logic,
clusters of magicians may also be understood to evolve by natural selection.
This observation leads up to a sense in which the psynet's evolutionary 
logic is
different from that which one sees in ecosystems or immune systems. Namely, 
in
the psynet, every time a process or cluster is moved in accordance with 
natural
selection, certain processes on higher levels are being crossed over and/or
mutated.
The issue of evolution is highly relevant to the question of the creativity 
of
mental process networks. The genetic algorithm (Goldberg, 1988; to be 
discussed
in detail in a later chapter) demonstrates the creative potential of the
evolutionary process in a wide variety of computational contexts. And the GA 
is
approximated by the activity of subnetworks of the dual network. Subnetworks 
are
constantly mutating as their component processes change. And they are 
constantly
"crossing over," as individual component interactions change in such a way 
as to
cause sub-subnetworks to shift their allegiance from one subnetwork to 
another.
And, finally, it must be observed that this genetic-algorithm-type 
creativity,
in the hierarchical network, and the evolutionary reorganization of the
heterarchical network, are one and the same! When memory items move around 
from
one place to another, seeking a "fitter" home, they are automatically
reorganizing the hierarchical network -- causing subnetworks (mental 
"programs")
to cross over and mutate. On the other hand, when processes switch their
allegiance from one subnetwork to another, in a crossover-type process, 
their
changing pattern of interaction consitutes a changing environment, which 
changes
their fitness within the heterarchical network. Because the two networks are
one, the two kinds of evolution are one. GA-style evolution and ecology are
bound together very tightly, much more tightly than in the case of the 
evolution
of species.
Ecology tells us that what is evolving in the mind is not arbitrary forms 
but
self-preserving, autopoietic systems. In order to achieve the full psynet 
model,
one must envision the dual network, not simply as an hierarchy/heterarchy of
mental processes, but also as an hierarchy/heterarchy of evolving 
autopoietic
process systems, where each such systems is considered to consist of a 
"cluster"
of associatively related ideas/processes. Each system may relate to each 
other
system in one of three different ways: it may contain that other system, it 
may
be contained in that other system, or it may coexist side-by-side with that
other system. The dual network itself is the "grand-dad" of all these
autopoietic systems.
Autopoiesis is then seen to play an essential role in the dynamics of the 
dual
network, in that it permits thoughts (beliefs, memories, feelings, etc.) to
persist even when the original stimulus which elicited them is gone. Thus a
collection of thoughts may survive in the dual network for two reasons:
    a usefulness relative to the hierarchical control structure, i.e. a
    usefulness for the current goals of the organism
    a strong ability for self-preservation (autopoiesis), mediated primarily 
by
    heterarchical connections in complex feedback loops
This is the logic of the mental network, emergent from the underlying 
network of
brain. It is the logic of the set of states of neuronal modules -- and, I 
claim,
it is also the logic of the set of states of agents in the newly complex,
increasingly intelligent Internet. But this leads us on....


                              Chapter 3
                       The Psynet Model of Mind


"The Law of Mind is one, and but one..."
– Charles S. Peirce


1. What is the Psynet Model?
The psynet model is a conceptual model of the mind, created not only for AI
purposes, but for the analysis of human thought processes as well (see 
Goertzel,
1993, 1993a, 1994, 1997 for detailed expositions of various aspects of the
model, as it developed over time). It aims to capture the abstract 
structures
and dynamics of intelligence, under the hypothesis that these are 
independent of
the underlying physical implementation. The model can be cast in 
mathematical
language in various ways, and validated as an empirical scientific theory in
various respects; but these perspectives will not be presented here. The 
goal
here is simply to get across enough of the model to provide a conceptual
grounding for various ideas in the chapters to follow. We will begin with a
highly general, almost philosophical treatment of the psynet model, then 
show
how a particular realization of this model leads to a computational model of 
the
mind in terms of nodes, links and agents -- the Webmind architecture, to be
outlined in more detail in Chapter 8.
The essential ideas of the psynet model are simple. A capsule summary is:
    A mind is a system of agents (called "magicians") which are able to
    transform, create & destroy other agents
    Many of these magicians act by recognizing patterns in the world, or in
    other agents
    Thoughts, feelings and other mental entities are self-reinforcing,
    self-producing systems of magicians
    These self-producing mental subsystems build up into a complex network 
of
    attractors, meta-attractors, etc.
    This network of subsystems & associated attractors is "dual network" in
    structure, i.e. it is structured according to at least two principles:
    associativity (similarity) and hierarchy (categorization and 
category-based
    control).
    Magicians pass attention ("active force") to other magicians to which 
they
    are related
    Because of finite memory capacity, mind must contain magicians able to 
deal
    with "ungrounded" patterns, i.e. magicians which were formed from
    now-forgotten magicians, or which were learned from other minds rather 
than
    at first hand -- this is called "reasoning"
    A mind possesses magicians whose goal is to recognize the mind as a 
whole as
    a pattern -- these are "self"
The use of the word "magician" here is perhaps worth comment, and 
interrelation
with conventional computer science terminology. Gul Agha, in his book 
Actors: A
Model of Concurrent Computation in Distributed Systems, defines an actor as
follows:


Actors are computational agents which map each incoming communication into a
3-tuple consisting of:
    a finite set of communications sent to other actorsa computational agent
    that can communicate with other actors,
    a new behavior (which will govern the response to the next communication
    processed); and,
    a finite set of new actors created


Magicians are more general than actors, in that they are not restricted to
digital computers: in a quantum computing context, for example, magicians 
could
be stochastic, and hence send an infinite set of possible communications
determined by quantum-level randomness, violating Agha's first condition. 
Since
the psynet model of mind is intended to apply to human brains as well as 
digital
computers, and since human brains may well be quantum systems, the concept 
of
"actor" is not adequate for the general psynet model. However, it is
nevertheless true that, in a digital-computing context, a magician is a 
special
kind of actor. In particular, it is an actor whose behaviors explicitly 
include
    The ability to destroy itself
    The ability to carry on complex dynamics within its own behavior space, 
in
    the absence of external communications
The psynet model therefore coincides well with the modern tradition of
distributed computing. What the psynet model does, which nothing else in
contemporary computing or cognitive science does, is to give detailed plan 
for
how a large community of computational agents should be set up in order that 
the
community should, as a collective, evolve highly intelligent behavior. The 
right
mix of agents is required, as well as the right kind of "operating system" 
for
mediating agent interactions.
The Webmind AI system provides a general "agents operating system" for 
managing
systems of software magicians that share meaning amongst each other, 
transform
each other, and interact in various ways; the magicians may live in the RAM 
of a
single machine, may be run on multiple processors, and may live across many
machines connected by high-bandwidth cable. It provides mechanisms for 
magicians
to represent patterns they have recognized in other magicians, and patterns 
they
have recognized in the overall magician system of Webmind. And it also 
provides
an appropriate assemblage of magicians, specialized for such things as
    text processing
    numerical data processing
    text-numerical interrelation
    supervised and unsupervised concept formation
    reasoning
    social interaction
    language processing
These magicians have their own localized behaviors but achieve their true
intelligence only in the context of the whole magician system that is 
Webmind.
This article describes the psynet model in general in a clear and concise 
way,
and then explains how the particular architecture of the Webmind system 
serves
to realize this model in an effective way on modern computer networks.
2. The Psynet Model of Mind in 37 Easy Lessons
There are many ways to construct the psynet model. This section will give a
construction that is centered around the concept of "meaning" -- an approach
that is particularly useful in a Webmind context, because the goal of 
Webmind as
a product is provide meaningful answers to human questions. The model will 
will
be presented here as a series of 37 Observations about the nature of mind. 
No
attempt, in this exposition, will be made to determine the extent to which 
the
observations are axiomatic as opposed to derived from other observations,
assumptions, etc. These are interesting questions, but beyond the scope of a
document whose main focus is the use of the Psynet model and its general
implications for Internet AI. In Chapter 8, we will return to these 37 
questions
and specifically indicate how each one is realized in the Webmind 
architecture.
Observation 1. Mind is a set of patterns, where a pattern is defined as "a
representation as something simpler".
To understand what a "pattern" is one must understand what "representation" 
and
"simpler" are. In order to define simplicity one requires a "valuation" -- a 
way
of assigning values to entities. In mathematical terms, a valuation is a 
mapping
from some set of entities into some partially ordered domain, e.g. into the
domain of numbers. A representation, on the other hand, is an entity that
"stands for" another entity. To talk about representation one must have 
three
entities in mind: the entity A being represented, the entity B doing the
representing, and the entity C that recognizes B as a representation of A. 
The
recognition activity of C is a kind of transformation; we may write C(B)=A.
Thus, putting simplicity and representation, together, the conclusion is
Observation 2: To be able to have patterns, one must have entities that can
transform entities into numbers (or some other ordered domain, to give
simplicity judgements), and one must have entities that transform entities 
into
other entities (so as to enable representation)
The space of entities in which patterns exist must be a space of entities 
that
can be considered as transformations, mapping entities into other entities. 
The
optimal name for entities of this sort is not clear; in some past writings 
on
the psynet model these entities have been called "magicians," in others they
have been called "agents" or "actors." Here we will stick with the term
"magician", which is a whimsical terminology intended to evoke images of
intertransformation: each magician, each entity in the mind, has the ability 
to
transform other magicians by a variety of methods ("magic spells"). The mind 
is
envisionable as a community of magicians constantly magicking each other 
into
different magicianly forms.
Observation 3: Mind is intrinsically dynamical. The transformation C(B)=A, 
in
the definition of pattern, implies a "change": C changes B into A
Observation 4: Magicians can be combined in a way other than transformation;
they may be combined in space. The result of joining A and B in space may be
denoted A#B.
Observation 5: Spatial combination gives rise to the possibility of emergent
pattern: patterns that are there in A#B but not in A or B individually.
Observation 6: The meaning of an entity may be defined as the set of all
patterns associated with that entity -- where a pattern P may be associated 
with
an entity A in several ways: P may be a pattern in A, P may be an emergent
pattern in the combination A # B. or P may close to A in spacetime (P and A 
may
have occurred in about the same place at about the same time)
Observation 7: Pattern and meaning are subjective, in that they depend on 
who is
measuring simplicity, and who is defining the set of permissible 
transformations
Observation 8: Meaning is dynamic as well as static. The patterns in an 
entity
include patterns in how that entity acts, and interacts, over time.
Observation 9: In any given mind, at any given time some magicians are given
more attention than others. Attention means that a magician is allowed to 
carry
out transformations on other magicians.
Observation 10: A mind always possesses some degree of randomness (i.e.,
dynamics which has no patterns that it can detect). Randomness occurs 
wherever
attention occurs.
Observation 11: The basic law of mind dynamics is: A magician passes some of 
its
attention to other magicians with whom it shares meaning
This dynamical law has a long history in philosophy; it was most clearly
enunciated by the American philosopher Charles S. Peirce toward the end of 
the
last century. The relation to neural network dynamics is clear and will be
elaborated below: in a neural net, a neuron passes some of its activation to
other neurons that it is connected to. If one thinks of a magician as a 
neuronal
module, and posits that two modules share meaning if they are strongly
interconnected, then Observation 11 fits in perfectly with neural net 
dynamics.
However, it is important from an AI point of view that we adopt a certain
dynamic because it has a psychological importance rather than because it 
loosely
models some aspect of the human brain.
Observation 12: Sharing of meaning may take many forms. Primally, meaning
sharing may be of three different kinds:
symmetric
asymmetric
emergent.
The first three types of meaning sharing are all "structural" in nature, and 
may
be formalized as follows. Denote the meaning of magician A by the fuzzy set 
m(A)
consisting of patterns in A or emergent between Aand other entities. 
Symmetric
meaning sharing is gauged by the formula
[ m(A) intersect m(B) ] / [m(A) union m( B)]
Asymmetric meaning sharing is given by
m(A)/[ m(A) union m(B) ]
Emergent meaning sharing is given by
[m(A#B) - m(A) - m(B)] / [m(A) union m(B)]
Meaning sharing incorporates temporal and spatial reality to the extent that 
the
meaning m(A) of A includes entities that occurred close to A in spacetime.
The above observations pertain to mind but do not directly address the 
concept
of "intelligence." Intelligence, however, can be approached in a similar 
way:
Observation 13: Intelligence may be defined as the ability to achieve 
complex
goals in complex environments
Observation 14: The complexity of an entity may be defined as the total 
amount
of pattern in that entity, or equivalently, the amount of meaning in the 
entity.
Thus, intelligence is the ability to achieve meaningful goals in meaningful
environments.
To compute the amount of meaning means to take all X in the meaning m(A) of 
and
entity A, and add up the degree of membership of X in this fuzzy set. The 
catch
is that one must take not an ordinary sum but a "non-overlapping sum," not
counting twice two patterns that express essentially the same thing. The 
correct
formulation of this non-overlapping sum is a problem in algorithmic 
information
theory.
Note that complexity is defined in terms of pattern, which is defined in 
terms
of simplicity. In this sense this definition of complexity is not entirely
satisfactory, from a philosophical view. However, in any formal system, one 
must
take some things as basic and undefined. I have chosen to take simplicity 
and
transformation as basic quantities, and derive others, such as pattern,
complexity and intelligence, from these.
This is a subjective rather than objective definition of intelligence, in 
that
it relies on the subjective identification of what is and is not a pattern. 
If
dolphins are bad at solving goals that we think are meaningful, and 
operating in
environments that we think are meaning-laden, this means that they are not
intelligent with respect to our own subjective simplicity measures, but they 
may
be highly intelligent with respect to some other simplicity measure, e.g. 
their
own. The upshot is that this definition of intelligence is pragmatically
valuable only in comparing different entities of like kind -- i.e., 
different
entities sharing largely the same goals, and comfortable in largely the same
environments.
These definitions lead to the following observation:
Observation 15: In order to achieve complex goals in complex environments --
i.e., to be intelligent -- a complex mind is required
This is important from an engineering perspective, because it tells us that 
the
dream of a mind in 100 lines of code is unachievable. This point comes up 
again
in the discussion of specialization among magicians, below. It is also 
important
in pointing out problems that can occur with complex systems engineering --
which only reflexive intelligence can solve:
Observation 16: A complex mind, implemented in a physical medium, will 
require
continual modification of its internal parameters to assure steady 
intelligent
functioning. This modification must be done intelligently in some cases, and 
so
there must exist certain magicians with a special feedback relation to the
physical medium determining the parameters of mental action.
A complex environment is one with a lot of patterns; in order to recognize a
complex web of patterns in an environment, however, a long & deep exposure 
to
this environment is required. This tells us that an intelligent system is
necessarily hooked up to a "rich" environment via perceptual sensors, "rich"
meaning rich in pattern. Furthermore it must be able to proactively search 
for
pattern:
Observation 17: Pattern recognition in a complex environment is best done by 
a
combination of perception, cognition (internal transformation of 
perceptions),
and action
Observation 18: A substantial amount of a mind's attention must often be
allocated to recognizing pattern in its environment, i.e. to this threefold
"perceptual/cognitive/active loop."
A mere collection of patterns recognized in an environment, however, is 
never
going to be a very intelligent mind. Mind is characterized by certain 
universal,
"archetypal" structures.
Observation 19: A "magician system" is a collection of magicians that is
self-producing, in the sense that any magician in the system can be produced 
by
the combination of some other magicians in the system. Minds are magician
systems, at least to within a high degree of approximation.
This is similar to the idea that minds are autopoietic systems, in the sense 
of
Maturana and Varela.
A terminological question arises here: When do we want to call a collection 
of
patterns a mind? Is every collection of patterns a mind, or is intelligence
required? Does a mind have to be a magician system, or not? These are not 
very
important questions, in that they just pertain to the definitions of words. 
A
sound practice is to refer to a mind as the set of patterns in an 
intelligent
system. Since the definition of intelligence is fuzzy, the definition of 
mind is
fuzzy as well, and the conclusion is that everything is mind-ful, but some
things are more mind-ful than others.
Observation 20: Highly intelligent minds are characterized by hierarchical
structures. The definition of hierarchy in this context is: A relates to 
{B1,
B2,...,Bk} hierarchically if each of the B[i] asymmetrically shares much 
meaning
with A. The process of creating hierarchical structure is called 
"clustering" or
"categorization."
Observation 21: Highly intelligent minds are characterized by heterarchical
structures, large connected networks of symmetrically meaning sharing 
entities
Observation 22: In a highly intelligent system, the hierarchical and
heterarchical structures of mind are aligned, so that in many cases, when A
relates to {B[1],...,B[k]} hierarchically, each B[i] relates to a number of
other B[i] symmetrically
This alignment of hierarchy and heterarchy has sometimes been called the 
"dual
network" of mind.
Observation 23: Minds are finite, so that if they live long enough, they 
must
forget. They will run into situations where they lose the B involved in a
representation C(A)=B, but retain the pattern A that was recognized.
Forgetting has profound consequences for mind. It means that, for example, a
mind can retain the datum that birds fly, without retaining much of the 
specific
evidence that led it to this conclusion. The generalization "birds fly" is a
pattern A in a large collection of observations B is retained, but the
observations B are not.
Observation 24: A mind's intelligence will be enhanced if it forgets
strategically, i.e., forgets those items which are the least intense 
patterns
Observation 25: A system which is creating new magicians, and then 
forgetting
magicians based on relative uselessness, is evolving by natural selection. 
This
evolution is the creative force opposing the conservative force of
self-production.
Observation 26: A pattern A is "grounded" to the extent that the mind 
contains
entities in which A is in fact a pattern
For instance, the pattern "birds fly" is grounded to the extent that the 
mind
contains specific memories of birds flying. Few concepts are completely 
grounded
in the mind, because of the need for drastic forgetting of particular
experiences.
Observation 27: "Reason" is a system of transformations specialized for
producing incompletely grounded patterns from incompletely grounded 
patterns.
Consider, for example, the reasoning "Birds fly, flying objects can fall, so
birds can fall." Given extremely complete groundings for the observations 
"birds
fly" and "flying objects can fall", the reasoning would be unnecessary --
because the mind would contain specific instances of birds falling, and 
could
therefore get to the conclusion "birds can fall" directly without going 
through
two ancillary observations. But, if specific memories of birds falling do 
not
exist in the mind, because they have been forgotten or because they have 
never
been observed in the mind's incomplete experience, then reasoning must be 
relied
upon to yield the conclusion.
So far this is a highly general theory of the nature of mind. Large aspects 
of
the human mind, however, are not general at all, and deal only with specific
things such as recognizing visual forms, moving arms, etc. This is not a
peculiarity of humans but a general feature of intelligence.
Observation 28: The specialization of a transformation may be defined as the
variety of possible entities that it can act on. The magicians in a mind 
will
have a spectrum of degrees of specialization, frequently with more 
specialized
magicians residing lower in the hierarchy.
The necessity for forgetting is particularly intense at the lower levels of 
the
system. In particular, most of the patterns picked up by the
perceptual-cognitive-active loop are of ephemeral interest only and are not
worthy of long-term retention in a resource-bounded system. The fact that 
most
of the information coming into the system is going to be quickly discarded,
however, means that the emergent information contained in perceptual input
should be mined as rapidly as possible, which gives rise to the phenomenon 
of
"short-term memory."
Observation 29: A mind must contain magicians specialized for mining 
emergent
information recently obtained via perception. This is "short term memory." 
It
must be strictly bounded in size to avoid combinatorial explosion; the 
number of
combinations (possible grounds for emergence) of N items being exponential 
in N.

The short-term memory is a space within the mind devoted to looking at a 
small
set of things from every possible angle. The bound of short-term-memory size 
in
humans and animals seems to be 7+/-2. For combinatorial reasons, it seems 
likely
that any physical system of similar scale will have a similar bound.
Observation 30: The short-term memory may be used for tasks other than
perceptual processing, wherever concentrated attention on all possible views 
of
a small number of things is required
One of the things that magicians specialize for is communication. Linguistic
communication is carried out by stringing together symbols over time. It is
hierarchically based in that the symbols are grouped into categories, and 
many
of the properties of language may be understood by studying these 
categories.
Observation 31: Syntax is a collection of categories, and "syntactic
transformations" mapping sequences of categories into categories. Parsing is 
the
repeated application of syntactic transformations; language production is 
the
reverse process, in which categories are progressively expanded into 
sequences
of categories.
Observation 32: Semantics is a collection of categories, and "semantic
transformations" mapping: categories into categories, category elements into
category elements, transformations into categories, and semantic 
transformations
into semantic transformations.
Observation 33: A key psychological role of syntax is to transfer semantic
knowledge from strongly grounded patterns to weakly grounded or entirely
ungrounded patterns.
Observation 34: Language is useful for producing magicians specialized for
social interaction. Syntax in particular is crucial for social interaction,
because another intelligence's observations are in general ungrounded in 
one's
own experience.
Language is for communication with others, and is tied up with sociality; 
but
the structures used in language are also essential for purely internal 
purposes.
Observation 35: The most intelligent minds have selves, where a "self" S is 
a
pattern which a mind recognizes in the world, with the property that, 
according
to the mind's reasoning, the substitution of S for the mind itself would 
produce
few changes. I.e., the self asymmetrically shares meaning with the entire 
mind.
Observation 36: The "self" of a mind is a poorly grounded pattern in the 
mind's
own past. In order to have a nontrivial self, a mind must possess, not only 
the
capacity for reasoning, but a sophisticated reasoning-based tool (such as
syntax) for transferring knowledge from strongly grounded to poorly grounded
domains.
Observation 37: The existence of a society of similar minds makes the 
learning
of self vastly easier
The self is useful for guiding the perceptual-cognitive-active
information-gathering loop in productive directions. Knowing its own 
holistic
strengths and weaknesses, a mind can do better at recognizing patterns and 
using
these to achieve goals. The presence of other similar beings is of 
inestimable
use in recognizing the self -- one models one's self on a combination of: 
what
one perceives internally, the effects of oneself that one sees in the
environment, and the structures one perceives in other similar beings. It 
would
be possible to have self without society, but society makes it vastly 
easier, by
leading to syntax with its facility at mapping grounded domains into 
ungrounded
domains, and by providing an analogue for inference of the self.
3. Psynet AI
The psynet model of mind, as developed above, is clearly far too abstract to
lead directly to any particular software program or engineering hardware
design.. It can be implemented in many different ways. But, even without 
further
specialization, it does say something about AI. It dictates, for example,
    that an AI system must be a dynamical system, consisting of entities
    (magicians) which are able to act on each other (transform each other) 
in a
    variety of ways, and some of which are able to evaluate simplicity (and
    hence recognize pattern).
    that this dynamical system must be sufficiently flexible to enable the
    crystallization of a dual network structure, with emergent, synergetic
    hierarchical and heterarchical subnets
    that this dynamical system must contain a mechanism for the spreading of
    attention in directions of shared meaning
    that this dynamical system must have access to a rich stream of 
perceptual
    data, so as to be able to build up a decent-sized pool of grounded 
patterns,
    leading ultimately to the recognition of the self
    that this dynamical system must contain entities that can reason 
(transfer
    information from grounded to ungrounded patterns)
    that this dynamical system must be contain entities that can manipulate
    categories (hierarchical subnets) and transformations involving 
categories
    in a sophisticated way, so as to enable syntax and semantics
    that this dynamical system must recognize symmetric, asymmetric and 
emergent
    meaning sharing, and build meanings using temporal and spatial 
relatedness,
    as well as relatedness of internal structure, and relatedness in the 
context
    of the system as a whole
    that this dynamical system must have a specific mechanism for paying 
extra
    attention to recently perceived data ("short-term memory")
    that this dynamical system must be embedded in a community of similar
    dynamical systems, so as to be able to properly understand itself
It is interesting to note that these criteria, while simple, are not met by 
any
previously designed AI system, let alone any existing working program.

                              Chapter 4:

                A Fourfold Model of Information Space


1. The Four Levels of Being
I have posited a fairly close structural and dynamic mapping between the
Internet, the human brain, and mind in general. To make this mapping 
explicit, I
will now venture into the domain of philosophy, and present a "fourfold 
model of
information space", applying to digital and neural information equally, and
derived substantially from the four "levels of being" described by the
contemporary philosopher Kent Palmer (1997): pure being, process being, 
hyper
being and wild being ... the latter category being the source of the name 
for
this book. Internet intelligence is bringing us a new kind of computing 
which
spans all philosophical levels.
A brief note on the role of philosophy in this kind of inquiry may be
appropriate here. Cognitive science is generally defined as an 
interdisciplinary
study involving a number of fields, including psychology, computer science,
linguistics and philosophy. However, transmission of ideas from the 
technical
disciplines to philosophy is generally much more common than transmission in 
the
reverse direction. I believe that cognitive science, and perhaps science in
general, would benefit from paying more serious attention to philosophical
concepts. Of course, philosophy cannot solve scientific problems, but it can
point science in interesting and useful directions, saving time that would
otherwise be spent exploring conceptually barren "dead ends." To make a
moderately loose analogy, one might say that the relation between philosophy 
and
science is like the relationship between design and coding in software
engineering. Design doesn't tell you how to code, but it gives you the 
structure
according to which the coding will proceed. While coding, you will often
discover new things that force you to go back and re-design, but this 
doesn't
mean that the design process was useless. Similarly, philosophy doesn't tell 
you
how to do science, it just gives structures according to which science can
proceed; and when the empirical or mathematical results feed back and modify 
the
philosophy, this is merely part of the form-generation process.
Palmer's four levels of being are:
    static being, that which is present, existent
    process being, that which is dynamic, changing, becoming
    hyper being, emergent, autopoietic being, that which achieves 
semi-stasis
    emergent from dynamics
    wild being, a meta-field of overlapping fields of stasis and dynamics,
    involving interaction of numerous emergent beings
These are extremely general concepts which, nevertheless, can be quite 
useful
for conceptualizing particular situations, such as the human brain, and the
Internet.
Traditional science focuses on static being and process being. Differential
equations, as used in physics, represent the most refined possible 
understanding
of process being. Chaos and complexity science, on the other hand, represent 
an
understanding of hyper being. The notion of an attractor, a structure 
emergent
from complex or chaotic dynamics, is moving toward hyper being -- and the 
notion
of a class of attractors, a type of attractor, is hyper being par 
excellence.
The word "hyper" here evokes the fact that this is neither just being, nor 
just
becoming, but is rather being overlaid on becoming -- becoming becoming 
being! A
given type of attractor -- e.g. a spiral, say -- may come out of an 
unlimited
number of dynamical systems; it is a form, a definite entity, which exists
tenuously as an emergent layer atop an underlying reality of process. An
autopoietic system is also an example of hyper being; and it is no 
coincidence
that autopoietic systems are full of archetypal attractor structures -- this 
is
the compositionality of hyper being, the tendency of emergent forms to 
combine
to form yet greater emergent forms, a tendency which culminates in the dual
network structure of the mind.
Wild being, finally, is not confronted by modern science hardly at all, 
though
it is discussed extensively, under various names, in existentialist, 
postmodern
and Eastern philosophy. Every empirical scientist has had numerous 
interactions
with this type of being, in the form of raw data. Raw data, from a fairly
uncontrolled experiment, will have all sorts of interesting patterns popping 
out
of it, but none of them quite strongly enough to be ascertained as 
definitely
existent. In order to do away with this field of possibilities, the 
scientist
focuses the experiment, and thus achieves hyper being -- a definite trend, a
pattern, emergent from the dynamical uncertainties of the actual phenomenon
being studied.
In the mind, one may say that wild being plays around the edges of the dual
network, arising as different parts of the dual network transform each other 
and
reinforce or decrease one anothers' structural rigidity. Wild being plays 
with
autopoietic structures as process being plays with rigid, static reality.
2. Internet Information Space
In this section I will use these philosophical ideas to give a fairly 
concrete
model of Net information space, based on the intuitive concept of the Net as 
an
self-organizing agent system. The four levels of being, we shall see, 
provide an
abstraction of the Internet on a par with the neural net abstraction of the
human brain, avoiding engineering specifics, but acknowledging the 
asynchronous
nature of Internet communication, and the heterogeneity of the population of
computational processes comprising the Net. The ideas of this section owe a
great deal to the "agent-data-event" model of Internet information space
developed by John Pritchard, and the presentation "Thinking Through 
Cyberspace"
given (and placed on the Web) by Kent Palmer in 1997.
In an Internet context, the role of static being is fulfilled by nodes and
connections -- by the physical substrate of the Net, being machines with
processors and RAM, with cables running between them. This is what is 
present,
what is given. Of course, it is also changing, but it is changing on a much
slower scale than the information within the Net, so from the point of view 
of
an initial conceptual model, it may be approximated as static.
The feedback between the static realm and the dynamic realm existing within 
it
is interesting, but is different in nature from the faster feedback occuring
within the static realm. It seems to be the case on general, 
system-theoretic
grounds that, for interesting things to evolve and grow in a world (real or
virtual), there must be a part of the world which is not allowed to evolve 
and
grow, but is fixed.
The dynamic, process aspect of Net reality comes in two forms: the 
information
flowing back and forth across inter-computer connections, and the 
computational
processes carried out at individual nodes of the network. Hyper being, on 
the
other hand, is represented by agents living within the Internet. Agents 
"surf"
on the dynamics of node-resident processes and information flow. Their
computation does not reside entirely at any particular node, and would not 
exist
without the continual flow of data, both within individual machines and 
among
different machines.
The distinction between an agent and a computational process resident at a 
node
is fuzzy rather than crisp, but is nevertheless important (much like the
distinction between ``living'' and ``nonliving'' matter in biology). An
``agent'' is, in essence, a computational process
    whose main role is to help in carrying out computations that are 
distributed
    over many nodes
    that is largely concerned with interacting with other agents.
It almost goes without saying that these ``agents'' may be either human or
artificial; this distinction will not be an important one here, as we will 
seek
a framework that allows us analyze characterize human and artificial agents 
in a
uniform way, as parts of an overall matrix of agent-event determination. The
distinction between human and artificial agents is a fuzzy one anyway, as 
all
artificial agents are at present designed by humans and hence reflect human
goals and biases; and all human interactions with the Internet occur by the
medium of artificial computing ``agents'' such as Web browsers and search
engines, which increase in their intelligence and autonomy each year. Rather
than distinguishing between human and artificial agents, a better approach 
would
be to say that each agent is characterized by a certain amount of
``human-directedness''; but for the present, this concept will not play a 
role
in our considerations either. An agent is defined behaviorally: in terms of 
the
activities that it carries out at each time, based on the information it has
received in the past.
Agents do things, they are the nexus of intelligence in information space;
nodes, on the other hand, are static and dynamic data resources. Network 
data is
information available via network activity. One might call a static resource 
a
store node and a dynamic resource a processing node; some nodes can be both
store nodes and processing nodes. An agent retrieves information by calling 
on a
node. Recursive retrieval occurs when a processing node calls on other nodes 
for
processing or storage retrieval; when recursive retrieval becomes 
sufficiently
complex it achieves the level of hyper being rather than dynamic being.
Currently, recursive retrieval plays only a minor role in the Internet 
economy,
but this can be expected to change: in the future, the Website that accepts 
your
money and provides you with your service may not actually do any of the work
that you've paid for -- it may subcontract out the work to the highest 
bidder
from a number of artificial agents representing other machines.
This terminology may easily be related to more standard Internet 
terminology.
Events, in the Internet context, are information transactions, carried out 
by
agents or by processes resident at individual nodes. When an event has an 
effect
on data form or content one may say that it is a transformative event, and 
that
the agent effecting the event is an actor. When an event has no effect on 
the
state of network information except to provide an agent with a local copy of
some data, then the agent has effected a retrieval event. Transformative 
events
have effects on network data subsequently available as processor or agent
events, and so we may say that agents behave as distributed processors in 
their
effects on nodes (remembering the fuzziness of the distinction between
node-resident processing and agent action).
Examples of events propagated by agents are replies to email or additions of
hyperlinks in Web pages. Much of the activity carried out by artificial 
agents
is currently retrieval-oriented, but as the Internet develops, 
transformative
events will become more and more common, and the distinction between the two
types of events may ultimately vanish. For instance, to an intelligent 
search
engine, each query represents a piece of information regarding the
humanly-perceived structure of the Web, and thus potentially leads to
modifications in the search engine database; a prototypical retrieval event,
searching for a Web page, then becomes a transformative event as well.
It is interesting to observe that the dynamics of a network of agents in
information space is essentially nondeterministic, in the sense that there 
is no
way to determine the overall state of the system of agents at any given 
time.
One might think that, since we are dealing with digital systems at this 
point in
the evolution of computer technology, the dynamics of an Internet agent 
system
could be understood by experimentation -- by manipulating an active network 
of
data, involving continuous agent activity and continuous external 
modification
of data. Such experimentation would require the collection of a periodic 
global
state image, however, and the collection of this state image, in practice, 
would
accidently change the definition of the network by interfering with system
operations (much as, in quantum physics, measuring a system inevitably 
disturbs
its state). Halting the network of agents to measure the overall state is 
not a
viable option either -- this is difficult without building a system that
synchronizes all activity; but synchronization must not be used in agent
processing activity for communication or retrieval if the definition of the
network is to remain the asynchronous, packet-switching network. While the
asynchronous network employs lazy synchronization in the form of collection 
or
set ordering for message encapsulation over packets, many such connections 
exist
simultaneously which implies that node reading, processing and writing
operations (effects) occur in random order relative to one another. In 
short,
the very nature of the Internet makes global state measurement difficult, 
and
results in a system that is in practice globally nondeterministic.
What we have, then, is a vision of the Net as complex system of agents 
passing
messages around a graph of nodes, interacting with the computational 
processes
and the static data stored at nodes, and giving rise to dynamical patterns 
that
can be understood only partially by any particular observer. Understanding 
and
utilizing these subtle dynamical patterns becomes an essential part of 
software
practice.
And here we get to the crux of the matter. Up till now, the focus of writers 
and
thinkers and programmers in the area of Internet intelligence has been on 
the
intelligence of individual agents. But, this is a limited approach. What is 
more
interesting and ultimately more fruitful is to think of the intelligence of 
the
overall dynamic graph that is the Internet -- of the patterns of emergent
computation binding together the different agents and computational 
processes on
the net. And this brings us beyond the level of hyper being into the more
intriguing territory of wild being. Wild being is what happens when a 
population
of intelligent agents interact within a richly fluctuating network of
information. It is intrinsically social, in the sense that it is made of
interacting subjective perspectives; and in the same sense that the 
individual
human brain is intrinsically social, being made up of the interacting 
subjective
perspectives of the numerous sophisticated and specialized modules within 
it.
3. Net Versus Brain
It is simple enough to see that the Net, as a whole, has the very same 
network
structure that modern AI theorists, with their neural nets and semantic
networks, have simulated within individual serial computers for the purpose 
of
modeling brain and mind processes. The Internet's nodes are more complex 
than
individual neurons, having more of the computational power of neuronal 
modules.
And, the packets sent around the Net are more complex than the jolts of 
charge
sent around the brain -- they are more similar, perhaps, to coherent, 
nonlinear
spatially-distributed electromagnetic waves passing through the brain. But, 
in
essence, they are quite similar systems: they are self-organizing networks, 
in
which information is carried and computation is carried out both locally and 
in
a global, distributed fashion. While the early cyberneticists set out to 
build a
brain based on the neuron level of abstraction, what has happened is that we
have built a brain based on the neuronal module level of abstraction. 
Adapting
our mathematics and design methodologies to this fact is not a nontrivial 
task,
but nor is it impossible. "The network is the computer is the mind" is the
important thing -- and not so much what kind of computer or what kind of 
network
happens to be involved.
Analysis in terms of the levels of being allows us to take this analogy 
little
deeper. In the brain, static being corresponds to the wiring of the brain, 
the
layout of neurons, given at birth via evolution and pre-natal 
self-organization,
and evolved to a lesser degree during the first years of life (Edelman, 
1988).
Dynamic being corresponds to electricity coursing through the brain, from 
neuron
to neuron; and to the flow of neurotransmitters and other associated 
chemicals
through the brain, and from the brain to the body. Hyper being corresponds 
to
neural attractors, and classes of neural attractors -- structures that 
emerge
from the dynamics of neural subnetworks of the brain. Different neural
subsystems have different attractors, and communicate with each other by 
methods
such as frequency synchronization (Koch, 1994), which depend sensitively on
attractor structure. Furthermore, though this is as yet little understood, 
there
are mechanisms for shifting attractors from one part of the brain to 
another.
For instance, when an animal loses part of its visual cortex, another part 
of
its brain may reshape itself into an image of the lost region. Or, in some
cases, the processes enabling knowledge of a skill may move from one part of 
the
brain to another as other skills accumulate, taking up room.
And wild being, finally, corresponds to the overall dynamics of the brain, 
which
is a matter of overlapping subnetworks, each one imposing its own view on 
the
remainder of the brain via complex, organized activity. Monitoring the 
brain's
global activity is very difficult precisely because it is "wild" in this 
sense.
Each EEG wave has numerous possible generators, and is usually a nonlinear
superposition of different generators. The activity in any one part of the 
brain
is not comprehensible in terms of that part alone, but only in terms of the
views of that part of the brain held by the other parts of the brain that
interact with it.
We see, then, that there is a harmony between the brain and the Internet on 
a
fairly deep philosophical level. In both cases, we have a fixed underlying
substrate consisting of nodes and links -- neurons and synapses on the one 
hand,
computers and cable on the other. The level of granularity is smaller in the
brain; but on the other hand, if one accepts the "Neural Darwinist" 
postulate of
brain modularization, then the basic module of brain structure is not the 
neuron
but the neuronal group, which is more closely similar in size to a single 
PC. In
both cases, one has two kinds of dynamics constituting the process level of
being -- dynamics within the individual neuron (as modeled by the 
Hodgkin-Huxley
equation) or neuronal module, together with electrical and chemical flow 
between
neurons; and computation within the individual computer, together with flow 
of
sound and light along cables between computers.
On the hyper being level, in both cases, one has complex entities that span
numerous nodes and links to carry out their work. Also, in both cases, many 
of
these entities have a certain fixed location around which their distributed
operation centers -- e.g. the home node of a Web spider; the physical 
location
of a neuronal group with a certain attractor. There is a greater potential
variety to the types of emergent entities -- computational agents -- in the
Internet than in the brain. On the other hand, precisely because the range 
of
emergent entities is more restricted in the brain, the crucial
"compositionality" by which emergent entities build together to form more 
and
more complex emergent entities may come more easily in the brain than in the
Net.
In both cases, one has the potential for the emergence of wild being from 
the
interpenetration of numerous emergent entities. The challenge for Internet
intelligent engineers is to replicate the property of the brain whereby a
population of emergent "hyper being" entities arise which are
    sufficiently different from one another to produce a huge spectrum of
    structures and behaviors when acting in combination
    sufficiently similar to one another to act effectively in diverse
    combinations
This is an issue which will recur throughout this book. The WebMind
architecture, to be described below, represents a solution to this problem 
via a
proprietary software architecture. The Agents Interaction Protocol, also
described below, represents a sketch of a public-domain solution to the 
problem,
a solution which is neutral in regards to specific implementation of agents,
consisting of a general "glue" language making it easier for different kinds 
of
agents to act cooperatively as composites, thus contributing to the leap 
from
hyper being to wild being.
Finally, in both cases, one has the potential for a "looping back" by which 
the
wild dynamics of the whole loops back and affects the most basic, static 
level
of being. However, the feedback works quite differently in the two examples. 
In
the case of the Net, such feedback occurs when high-volume regions of the
computer network get more cable, more computers, etc. It also occurs when
systems break down due to excessive demand. In the case of the brain, on the
other hand, this happens in youth when stimulation of brain regions causes
growth in those regions; and in adulthood when drug use causes brain 
connections
to atrophy. This is an area where the Net excels over the brain: the human 
brain
has hardware which grows in youth, and then decays throughout adulthood, 
whereas
the Net will be able to continue growing on the static level even as it 
develops
through successive stages of maturity on the higher levels of being. This is 
a
capability that will have to be managed carefully, but that will doubtless 
have
tremendous consequences.
3. The Psynet in Biological and Computer Networks
In the psynet model of mind, the two main processes of mind are said to be
evolution and autopoiesis. Autopoiesis, we have said, is closely aligned 
with
hyper being -- it is the emergence of "semi-stable" structures from 
underlying
fluctuations. Ecological evolution, acting on the field of autopoietic
subsystems of the dual network, is both process being and wild being. It is 
a
process creating new forms toward specific purposes, and it is also an
aim-transcendent generative process, producing archetypal and incidental 
forms
which are retained in the mind and serve as seeds for future 
self-organization
of autopoietic structures.
The Internet itself has a naturally hierarchical structure, expressed in the
domain name system and in the patterns of network connectivity. Individual
machines naturally cluster with the other machines on their intranet, then 
with
the other machines in their nation or continent, lastly with other machines
around the world. This is a consequence of network bandwidth which may 
disappear
as engineering improves, but it is a present fact, and a fact which may help
rather than hinder the development of intelligence. The relative but not
complete isolation of intranets provides an analogue of the clustering of
neuronal modules into meta-clusters in the brain, and (as we shall see later 
in
the discussion of Webmind) provides a lower-level impetus for the formation 
of
autopoietic agent systems on the intranet level.
And, of course, there is an heterarchical structure to the Net as well. This 
is
the obvious sprawling hyperlink structure of the Web, which is taken on by 
Web
agents such as spiders, e-mail agents, e-commerce agents, etc. The 
heterarchy is
indeed aligned with the hierarchy, in that machines within the same 
hierarchical
domain are likely to deal with similar topics and experience similar 
phenomena,
thus being strongly heterarchically interlinked. One has, in the Net of 
today, a
germinal dual network. The primitive Internet agents currently in action, do
their thing in the context of this dual network. And, when these agents 
become
sufficiently numerous and sophisticated and intercoordinated to lock into a
pattern of globally emergent intelligence, this dual network structure will 
no
doubt be a part of it.
What the psynet model claims, in philosophical terms, is that the dual 
network
structure is an incomparably effective means for giving rise to hyper being 
and
wild being. The dual network is a compositional structure of hyper beings --
emergences upon emergences upon emergences -- which lends itself naturally 
to
evolution, in both its directed and free-flowing, generative, wild aspects. 
That
the brain and the Net both manifest dual network structures on the static 
being
level reflects feedback processes taking place over a long time scale -- it
reflects the fact that both of these systems have physically evolved in 
order to
be more intelligent, one over millennia of biological advance and the other 
over
years of engineering advance. The brain is the only current example of Wild
Computing; but we are about to have another.

                              Chapter 5:

               The Internet Economy as a Complex System

1. Introduction
One of the main factors catalyzing the explosive growth of the Internet has 
been
the widely perceived prospect of making money via networked computers. This 
is a
prospect which is rapidly becoming a reality, especially in the the areas of
intranet-based knowledge management and business-to-business e-commerce.
Profit via the Internet takes many forms. There is the profit that goes to
Internet service providers, telecommunications and backbone providers, 
Internet
software companies, and to Internet media companies via online advertising. 
And
then, more interestingly, there is electronic commerce (e-commerce) --
commercial transactions taking place over the Net, a rapidly growing segment 
of
the economy, which is expected to increase yet more rapidly over the next 
few
years (Varian, 1996). Current Internet commercial transactions are estimated 
at
hundreds of millions and are projected for billions within the decade.
Though many consumers are still reluctant to deal financially over the Net,
businesses have shown no such reticence, and business-to-business e-commerce 
is
booming. Furthermore, one expects that the growth of consumer transactions 
over
the Net will be energized with Visa and Mastercard's release of secure 
software
standards for their cardmembers' internet transactions, and the acceptance 
of
standards for micropricing. All in all, it is reasonable to project that, in 
a
decade's time, the vast majority of economic transactions will involve a
significant electronic component.
Given these trends, it seems clear that, over the next few decades, dealing 
with
electronic commerce is going to be one of the most important jobs of
self-organizing Internet AI software. Software dealing with the Net is going 
to
have to understand Internet economics. However, this is more of a challenge 
than
it might sound like at first, because no one understands Internet economics!
Traditional economic theory gives us, to say the least, fairly poor guidance 
in
understanding this emerging economic domain. A new conceptual framework is
required. And the framework needed for understanding Internet economics is, 
I
believe, the same as the framework needed for understanding Internet AI:
self-organization, emergence, complexity. In this chapter I will sketch a 
simple
model of Internet commerce based on dynamical systems theory and the theory 
of
autonomous agents.
Starting from a discussion of the software market and the particular 
economic
efficiences provided by globally networked communication, I will model 
economic
transactions as agent interactions taking place on a graph of 
store/processor
nodes. In this context, I will turn to the dynamics of economies in general,
summarizing an emerging body of evidence which suggests that economic 
dynamics
is self-organizing and chaotic, with complex, overlying emergent patterns.
Finally, putting these ideas together, I will argue that the increased
communication provided by the Internet is highly likely to make the 
electronic
economy more pronouncedly complex than the traditional economy. This 
suggests
that, in the context of e-commerce, the ability to manipulate information
regarding markets and their dynamics will be even more important than it is
today. Artificial intelligence and powerful data-mining and data 
visualization
tools will allow human and computational agents to successfully navigate 
this
new, newly complex and dynamic economy. The main conclusion is that emerging
information-based economy, a major role will necessarily be played by
    artificially intelligent economic agents (AI)
    tools for guiding intuitive human exploration of market dynamics (IA,
    intelligence augmentation)
As we move from an economy focussed on exchange of goods to an economy 
focussed
on communication of information, the ability to elicit information from 
complex,
chaotic market dynamics via statistical pattern recognition will assume 
premium
value.
2. General Features of the Emerging Internet Economy
The emerging Internet economy is a complex beast, with many different 
facets.
All the features of traditional economies are present, but with a different
slant, caused by the increased role of communication and the increased ease 
of
information collection.
One can now use the Net to buy products of all sorts -- in the consumer 
arena,
books, software, computer hardware, CD's and travel reservations are 
high-volume
items at present; the business-to-business market is extremely diverse with
particularly high volume in electronic and mechanical components of various
sorts. One also can now use the Net to buy data, a fact which is exemplified 
by
Lexis-Nexis and other private storehouses of text, and by for-a-price 
Internet
finance services such as PC Trader. The next stage, which is just getting 
off
the ground, will be buying computational processes that act on one's own 
data.
Some Internet services which are currently free are expected to become
commercial in the near future. For instance, Web search engines, electronic
magazines and Web classified ad sites are now free services, but with the 
advent
of "micropricing," it will become practical to charge users prices as low as 
a
thousandth of a cent for individual transactions such as accessing Websites 
or
performing searchings. The software for micropricing already exists 
(Glassmann
et al, 1995), it has merely to become widely accepted. The transformation of
currently free services into micropriced services can be expected to lead to 
a
substantial increase in the quality of services offered; and it may well 
also
result in the transformation of some currently commercial services, such as 
Net
access itself, to free services.
The Internet offers an unparalleled mechanism for market efficiency. No 
longer
will a consumer have to ``shop around'' in tedious or time-consuming 
fashion;
instead there will be services that list the prices available for comparable
items from different vendors. An initial experiment in this regard is 
PriceWeb
(Zabeh, 1995), which is a Website listing prices offered for computer
peripherals by various hardware vendors. If PriceWeb becomes commonly used 
it
will have a tremendous impact on the way peripherals are purchased. The 
effect
will be to push vendors into charging a single, uniform price, with 
deviations
from the price only being possible if the vendor offers extra services of
significant and obvious value.
Not all commodities are as standardized as computer peripherals, but the 
basic
idea of PriceWeb can obviously be extended throughout the Internet economy 
in
general. One will have artificial and human agents offering goods at various
prices; other agents monitoring the prices being offered by various agents 
and
posting this information to databases available for free or at a price; and
other agents using the posted information to purchase from or negotiate with 
the
selling agents.
In spite of advancing commercialization, however, the economy for free
information is not likely to disappear. Free transactions form the backbone 
of
the Internet currently and will surely remain, both in the academic and 
hobbyist
communities, and within the commercial Net community as a necessary tool for
attracting interest in goods and services. Free transactions take on the 
role of
advertising in non-interactive communications, and extend the simple relay 
of
image and brand with in-depth information for various classes of users.
Interactivity between consumers and producers enables producers to respond 
to
users' interests in differentiating their products.
In terms of the fourfold model of information space, we may observe that
Internet commerce involves a complex system of agents carrying out a large
variety of transactions -- i.e., it lives on the level of hyper being, with 
a
potential for entering into wild being. We have seen that the dynamics of a
system of agents acting on a graph of static and dynamic nodes will be 
extremely
complex, and the special case of economic agent interactions is no 
exception. In
the following section we will review evidence for chaotic and otherwise 
complex
dynamical behavior in certain economic systems: the evidence, as will become
clear, very strongly suggests that the Net economy will give rise to 
structured
chaos, and complex emergent dynamical patterns of various kinds. Making use 
of
these patterns becomes an essential aspect of doing business.
For example, a producer's data nodes are organized into an interactive web 
which
classifies users by interests, but cannot be seen as exclusively classifying
users except in the simplest case. Agents who explore one node web will 
often
explore others and so are classifiable with many interest properties. This
classification constitutes a macroscopic perspective on the product's or
products' differentiation from others' for the user -- while the user has
similarly explored other products with purpose or feature correlations. In 
the
information space economy, there is an enhanced ability for producers to
identify small niche markets, and consequently, an increased necessity for 
them
to do so, using the vast amount of consumer behavior data at their disposal, 
and
the most sophisticated available categorization techniques. Perceiving 
patterns
amidst the complexity and chaos of the electronic market becomes imperative.
The Software Economy
One important and interesting aspect of e-commerce is the software economy 
--
the economics of the ``computational processes'' resident at particular 
nodes in
information space. Software economies are both free and priced. Currently
emerging pricing structures include the common shrink-wrap model, and the 
usage
model.
In the shrink-wrap model, a user pays a one time license fee for free usage.
This market for software is defined with one class of users and one 
producer.
Multiple producers' products may be interoperable, codependent or 
competitive.
For example, a word processor comonly requires a particular operating system 
to
run, so the user must have the operating system before using the word 
processor.
On the other hand, if the desktop operating system provides special
interoperability characteristics for the word processor software, then the
dependence of the word processor on the operating system is enhanced by
interoperabilty features which may have independent utility for the user.
In the usage model, on the other hand, users pay for a particular 
transaction
which employs software and computing resources of time and memory space.
Computing resources have opportunity cost across users. The usage model 
prices
in economies that are at least predominantly internet contained. The usage 
model
also serves a variety of users from disjoint utilities. A consumer and a 
worker
requiring search engine processing time may have very different demands as 
the
worker's time has opportunity or compensation costs for the firm. With the
advent of Java applets and network-based computing environments, it is
anticipated that the usage model will increasingly supplant the shrink-wrap
model, potentially until the latter becomes obsolete.
A notable aspect of the software economy is the high cost of producing the 
first
unit of a product, as compared to the minimal cost of producing subsequent
copies. This is an exaggeration of the economics of the traditional 
publishing
industry (e.g. it is estimated that about 70% of the cost of producing an
academic journal goes into the cost of producing the first issue). What this
implies is that there is no easy way of determining the appropriate price 
for a
software product; pricing strategies depend in a very fundamental way of
estimates of future demand. Often the ``beta'' version of a software product 
is
given away for free, to whet demand. Also, differential pricing is common, 
with
inferior versions of a product offered for lower prices, even though there 
would
be no additional cost in providing everyone with the fully featured 
software.
Hal Varian (1996) has studied these issues in a thoughtful way, and 
presented
economic mechanisms which provide for effective differential pricing in an
e-commerce context.
The combination of differential pricing with usage-based pricing may lead to 
a
new, complex and intriguing software market in the near future. Consider a
situation in which, instead of purchasing a large shrink-wrapped program, 
each
user dynamically assembles software according to their needs, from 
components
(applets) available on the Net. The utility of a given applet to a given
consumer will depend on the particular configuration of other applets within
which they want to use it.
One then has a situation where agents are required in order to assemble
computational (node-resident) processes. Each user will send out an agent to
find the applets that he needs in a given context. The agent will survey the
different applets and assess how well each one will fit the user's 
particular
needs. Given its assessment of the utility of each applet, it will place 
bids,
which the selling agents will then accept or reject based on their own
predictively motivated differential pricing strategy. The selling agents 
will
alter their pricing scheme based on perceived trends in demand, and,
potentially, buying agents may alter their purchasing patterns in ways
determined to have particular effects on selling agents' pricing schemes.
Information versus Goods
We have said that the information economy will be complex and chaotic and
nondeterministic, and we have suggested how this complexity might manifest
itself in the software market of the near future. Now let us look at the 
issues
from a more general perspective. What, fundamentally, is the difference 
between
an economy based on agents in information space and an economy based in the
``real world''?
The difference, we suggest, is both quantitative and qualitative. The basic
quantitative difference is the ease of communication, of information
dissemination.In the information space economy, each agent is potentially
``directly connected'' to each other agent, and the fixed cost of exchanging
information is very low. Obviously the ease of communication between 
economic
agents has been increasing steadily through history, with the gradual
introduction of superior communication and transportation technology. But 
with
the advent of Internet commerce, we are seeing the largest and most sudden
change ever in ease of communication of economic information. This change, 
it
seems highly likely, is going to push the global economy into a new 
dynamical
regime -- a new mode of being. The quantitative difference in ease of
communication is leading to a qualitative change in focus -- from a focus on
goods and their exchange to a focus on information and its communication.
Exchange of real-world goods is still an important part of e-commerce, along
with exchange of information goods (software programs, databases, etc.); but
what is new is that, in an information-space economy, every economic agent 
must
continually be acutely aware of the information base from which its 
decisions
emanate, and the effect that its actions have on the information bases of 
other
agents in the market. When economic agents spend as much time ``thinking'' 
about
the informational causes and effects of their actions, as about the actions
themselves, then we have moved into a different order of economy -- an
information-based economy rather than a goods-based economy. The fuzziness 
of
this distinction should not be glossed over -- all economies involve both
information exchange and goods exchange -- but the point is that the change 
in
emphasis is likely to be sudden rather than gradual.
3. Chaos and Complexity in Economic Dynamics
Having discussed some of the particularities of electronic commerce, and the
general structure and dynamics of information space, we will take a step 
back
and review some general properties of economies, which happen to be 
particularly
relevant to the dynamics of heterogeneous economic Internet agents.
Economic systems, ordinary or digital, are far more complex that traditional
economic theory would like to admit. Most notably, the notion of an 
``invisible
hand'' which drives prices toward an equilibrium balance of supply and 
demand
does not bear up to scrutiny, either mathematically or empirically. The 
reality
seems to be that prices fluctuate according to noisy chaotic dynamics, with
overlying long-range statistical predictability, and with perhaps some
``windows'' of unusually high short-term predictability. In short, economic
systems are no more and no less predictable or equilibrium-oriented than 
other
complex, self-organizing systems -- e.g. brains, ecosystems, or the Internet 
as
a whole.
This conclusion is particularly important for electronic commerce, as it 
tells
us what kind of dynamics to expect for the emerging Internet economy. It 
tells
us that e-commerce agents, in order to be successful, will have to be 
skilled in
the art of statistical pattern-recognition -- will have to be capable of
recognizing and acting on short-, medium- and long-term trends in noisy,
statistically patterned chaotic systems. In the real economy, the noisy,
patterned chaos of price dynamics is dealt with by human intuition. In the
Internet economy, on the other hand, things will happen too fast and in too
distributed a way for human intuition to intervene every time a decision has 
to
be made. The complexity of economic dynamics thus leads directly to the need 
for
intelligent e-commerce agents.
Chaotic Price Dynamics
As shown in detail by (Saari, 1994), the chaotic nature of economic dynamics 
can
be derived directly from basic price theory. A brief review of Saari's 
arguments
may be useful here. Consider, for simplicity, an economy without production,
consisting of agents exchanging n types of goods according to positive 
prices.
Let pj be the price per unit of the jth commodity; then the cost of xj > 0 
units
is pj xj. The price of a commodity bundle x = (x1, ... , xn) is given by the
inner product (p, x), where the vector p represent the prices of all
commodities. In the absence of production, what the kth agent can afford is
based on what he can sell-- his "initial endowment" wk.
How does each agent determine what to buy, in this simplified set-up, given 
a
particular price vector? There is no natural ordering on n-dimensional 
space,
so, following typical practice in economic theory, we may impose one by 
defining
a "utility function" uk: Rn+ --> R for each agent, defined so that uk( y) > 
uk(
x) means the k'th agent prefers bundle y to x. For convenience, Saari 
assumes
that utility functions are strictly convex, and that no good is undesirable 
to
any agent, i.e. more of each good is better than less from each agent's
perspective. These assumptions are not realistic, but we are showing that
economic dynamics are chaotic; and it is plain from the mathematics that 
lifting
these restrictions will not make the dynamics any less chaotic.
Within this framework, one can use traditional Lagrange multiplier methods 
to
determine the k'th agent's demand at price p, and the k'th agent's "excess
demand" xk( p), being the difference between what is demanded and what is
supplied (wk). One can then construct the "aggregate excess demand function"
X(p) as the sum over k of the Xk( p), and study its properties, called 
Walras'
laws.
It is not hard to show that equilibria do exist in this model economy; i.e.,
there is a price vector p* for which the excess demand x(p*) = 0, and supply
equals demand. But do prices tend toward these equilibria? In the classical
picture, increased demand leads to increased prices, and so we have, for
discrete- time dynamics,
pn+1 = pn + h X(pn)
for some positive constant h. The mathematical fact, however, is that this
classical view virtually never holds. As shown by a theorem of Sonneschein
(1991), most trajectories of the dynamical system of economic agents are 
going
to be chaotic, rather than converging to equilibria (this holds in both the
continuous time and discrete time cases). Furthermore, the only way to 
restrict
attention to the equilibrium-convergent trajectories is to make completely
unreasonable assumptions regarding the preferences of the agents. In short, 
as
Saari puts it, ``with at least as many agents as commodities, anything can
happen!''
In fact, Saari's conclusions are even more dire. He notes that the 
application
of dynamical systems theory to economic equations `` not only causes worry 
about
the invisible hand story, but it forces us to question those tacit
assumptions--assumptions basic to several tools from economics--about how 
the
aggregate excess demand function for one commodity set relates to that of
others. One might argue (and this is a common reaction during a colloquium
lecture, particularly in a department of economics) that there may exist
conditions imposing strong relationships. Yes, but it is obvious from the
theorem that such constraints cannot be based upon the aggregate excess 
demand
function (as is a common practice). Instead they appear to require imposing
unrealistically harsh global restrictions on the agents' preferences....''
The basic point is that individual rationality on the part of individual
economic agents does not necessarily lead to simple, orderly, convergent and
transparently ``rational'' behavior on the market-wide level. A community of
agents, each acting to fulfill their own utility functions, can lead to all
manner of complex or chaotic dynamics. If one adds yet more realism, and 
makes
the utility functions of the agents vary based on the agents' predictions of 
the
future behaviors of the other agents, then things become yet more complex. 
The
overwhelming point is that economic theory should be a theory of intelligent
though not perfectly rational agents, navigating their way through a very
complex dynamical landscape -- not moving toward or arriving at equilibria, 
but
rather sculpting the chaos in which they live into temporarily beneficial 
forms.

Complexity in Economic Time Series Data
One may well wonder how this conclusion, derived through mathematical 
theory, is
borne out in by the detailed analysis of economic time-series data. The 
answer
here is that real economic data is even less ``orderly'' than mathematical
chaos. The incredible complexity of economic systems, combined with the
relatively short time series involved, makes convincing data analysis 
difficult,
but the clear message from the work that has been done is that economies are
complex and predictable only with great intelligence.
The case for chaos in economic time series is still ambiguous at this point
(leBaron, 1994), but the problem is not one of equilibrium-seeking versus 
chaos,
rather one of chaos versus pseudorandom behavior generated by exogenous 
shocks.
A real economy receives with a steady stream of new products and new
information, as well as intensive influence from politics, mass psychology 
and
so forth, and these factors combine to jolt economic time series away from
easily- identifiable chaotic attractors. Much as with the analysis of time
series in behavioral science, one finds that it is more productive to seek
statistically predictive models than to seek to identify attractors. The
disentanglement of chaos from external ``noise'' in extremely complex 
systems
like minds and economies is a formidable problem, and, ultimately, a problem 
of
little practical use. The key point is that there is no simple structure, 
such
as that posited by classical economic theory; instead there is a complex
dynamical structure which can only be appreciated by methods of statistical
pattern recognition.
This conclusion has, in essence, been realized by observers of the financial
world for a long time. For instance, ``technical trading rules,'' which were
once dismissed as useless, are now being accepted by economic theorists, and
appear to be connected with the chaotic nature of financial time series
(leBaron, 1994). Many technical trading rules have to do with moving 
averages;
they recommend that a trader buy when the price is above a long-term moving
average, and to sell when the price is below (Levich, 1994). The particular 
form
of the average is the subject of much scrutiny, as is the dependence of the
urgency of selling on the difference of the price from the average. This is
exactly the type of rule that one would expect to be useful in dealing with 
a
system that displays ``structured chaos'' -- chaos with overlying 
statistical
patterns. Technical trading rules are nothing but statistical prediction
schemes, which ignore low-level chaos and noise, and focus on high-level
regularities. In particular, rules involving buying when prices are above a
long-range moving average are tied in with the ``fractional Brownian noise''
model of time series, a particular variety of statistically structured 
chaos.
Simulations of Economic Complexity
Recognizing the complex, chaotic nature of economic markets, a number of
researchers have thought to simulate economic dynamics with stripped-down
``complex systems'' models. For instance, David Lane and Roberta Vescovini, 
in a
paper called ``When Optimization Isn't Optimal' (1995),report on simulations 
in
which agents choose between one product and another over and over gain, on a
series of trials, the decision at each time being based on information 
gathered
from agents who have previously adopted the different products. The big 
result
here is that following the optimal strategy suggested by probability theory 
--
the ``rational'' strategy -- is not actually maximally effective. What is
optimal in terms of probability theory, applied to a single agent making
decisions in a fixed environment, is not optimal in an environment 
consisting of
other agents who are also making their own adaptive decisions. They find,
furthermore, that in some cases an increased access to product information 
can
lead to a decreased market share for the superior products.
Along similar lines, Michael Youssefmir and Bernardo Huberman (1995) have
experimented with large distributed multiagent systems, involving agents 
that
continually switch their strategies in order to find optimal utility. They
report that they ``have analyzed the fluctuations around equilibrium that 
arise
from strategy switching and discovered the existence of a new phenomenon. It
consists of the appearance of sudden bursts of activity that punctuate the 
fixed
point, and is due to an effective random walk consistent with overall 
stability.
This clustered volatility is followed by relaxation to the fixed point but 
with
different strategy mixes from the previous one.'' In their particular 
system,
then, the notion of equilibrium is a reasonable approximation to reality, 
but
there are many different price vectors that give approximate equilibrium. 
One
near-equilibrium is reached, only to be disturbed by a bout of random
fluctuation that disrupts the system temporarily. Things then return to a
different near-equilibrium, and then start up again....
These simulations, like the mathematical calculations of Saari and 
Sonnenschein,
involve approximations to economic reality. However, the phenomena that they
describe have an undeniably ``real-world'' feel to them. These simulations
capture some of the fascinating richness of real economic dynamics. And they 
are
particularly interesting from the point of view of e-commerce, because, 
after
all, electronic commerce will also involve the interaction of computational
agents. One expects that the dynamics of e-commerce will resemble these
simulations as much as it resembles the dynamics of current markets.
4. Intelligence and E-commerce
The nature of Internet commerce is that information exchange becomes as
important as or more important than goods exchange. This feature, it seems
clear, can only intensify the complex, chaotic nature of economic 
interaction.
It makes it easier for a small effect to propagate widely. It makes changes
happen more quickly, with less time for intuition to acclimate to the new
circumstances. It makes markets more efficient, but it also makes more 
complex
differential pricing schemes viable; and because it supports products with 
high
initial production cost and low replication cost, it leads to pricing 
schemes
that depend intimately on predictions of future demand.
In order to deal with the complexity of the electronic economy, it will be
necessary for artificial intelligence and intelligence augmentation to be 
built
into standard e-commerce agents. Selling agents will have to be able to 
predict
the future demand for their products, in order to enforce intelligent 
pricing
strategies. Buying agents will have to be able to intelligently assess the
utilities of various products on offer (e.g. various applets for use in 
software
applications). This kind of prediction, because of the complex, chaotic 
nature
of economic dynamics, will require sophisticated statistical pattern 
recognition
techniques, including such tools as neural networks, genetic algorithms and
Markov chains. It will also require that complex network dynamics be made
accessible to human intuition, by the provision of sophisticated 
visualization
tools, based on a general model of information space such as the fourfold 
model
given above.

                              Chapter 6:

                      The Emerging Global Brain

1. Introduction
Over the next decade or two, we will see the evolution of the Net into a
full-fledged, largely autonomous, globally distributed intelligent system. 
And
as this occurs, we will see this Internet AI network wend itself further and
further into human affairs, yielding a synergetic, symbiotic global 
intelligent
system, incorporating machine and human intelligence into a single continuum 
of
thought, a human-digital global brain.
This sounds like science fiction, and indeed, it is closely related to 
several
science-fictional visions of the digital future. In this chapter I will
unapologetically pursue this angle, and indulge in some speculations on the
future of Internet intelligence, which will seem more or less farfetched to 
the
reader depending on their inclination and taste. The more conservative 
reader
may skip ahead to the following chapter which begins the more technical 
section
of the book. However, I believe that both speculation and concrete technical
work are valid and important; indeed, the two are inseparable. Without
speculations as to the future, we would have no way of deciding what to do 
in
the present.
When, at the end of the book, we visit the question of emergent natural 
language
processing, we will see that the philosophical ideas of this chapter become
strikingly relevant. The collective construction of meanings, as it turns 
out,
is a notion that strikes right at the heart of modern philosophy, and brings 
to
the fore all sorts of difficult ideas such as the nature of reality and the
collective unconscious.
2. Possible Futures
Many futurists have envisioned a future Net populated by artificially
intelligent entities, interacting in virtual worlds. This vision was 
portrayed
most memorably by William Gibson in his entertaining and influential novel
Neuromancer (1994). While this is a reasonable idea, and it does not 
contradict
my own thinking in any way, it is different from what I am projecting here,
which is that the Net itself will become a global intelligent system -- a 
World
Wide Brain.
Contra Gibson, a number of others thinkers -- most famously, roboticist Hans
Moravec (1990) -- have envisioned that humans will eventually "download"
themselves into computers, and lead bodiless digital lives. This is also 
related
to, but different from the idea of a global Web mind. The two ideas 
synergize in
a fascinating way.
Initially, I suggest, the global Web mind will exist as an entity physically
separate from human beings: we will interact with it via our computer 
terminals,
and perhaps more creative interface devices such as VR headsets. This first
phase is in store maybe five to twenty years down the line; and it is the 
main
focus of my thinking at present. The increasing integration of human 
activity
with World Wide Brain operations may ultimately occur via body-modifying or
body-obsoleting technologies a la Moravec, or it may occur without them, 
through
the advent of more sophisticated noninvasive interfaces. One way or another,
though, I conjecture, it will fuse with the global Web.
The change which is about to come upon us as a consequence of the emergence 
of
the global Web mind will be a very large one -- comparable in scope, I 
believe,
to the emergence of tools, or language, or civilization. What will 
distinguish
this change from these past ones, however, is its speed. In this sense, 
those of
us who are alive and young today are very fortunate. We will be the first 
human
generation ever to witness a major change in collective humanity with a time
span the order of magnitude of a single human lifetime. We are not merely
creating artificial intelligence, we are creating an artificially 
intelligent
crystallization of the common, collective patterns of the global human mind.
This is a new adventure, something that does not merely bring us back to the
spiritual unity of past modes of culture, but brings on us to new frontiers 
that
we can barely even imagine. The universe is ever-expanding, not only 
physically
but informationally. What we are seeing is yet another manifestation of the
endless quest of the universe to bring new patterns out of itself. Though 
this
is an angle that will not be pursued here with any seriousness, I do believe
this is an aspect of emergent Internet intelligence which can validly be
explored from the perspective of transpersonal psychology.
3. Psychology and Ethics of the Global Brain
So, let us imagine the Internet, a few years from now, the home of an 
advanced,
self- organizing AI system, spanning tens of thousands of intranets 
worldwide.
Such a system would be an independent, intelligent entity on its own,
interacting with humans, and incorporating human workflow and 
question-asking
and answering behavior into its own intelligent dynamics. It would weave the
process of collective social inquiry and individual human mental inquiry 
into a
digital fabric of a texture guided by its own processes of intelligent
self-organization.
This view of the future of the Net is reminiscent of the archetypal idea of 
the
"global brain," which first started appearing in the 1970's. In Russell's 
book
on the Global Brain, for example, computer and communications technology are
assigned only a minor role, and it is argued that human society is reaching 
a
critical threshold of size and complexity, beyond which it will enter the 
realm
of true intelligence, and human consciousness will give rise to a higher 
level
of collective consciousness. Russell's hypothesized supra-human intelligence
might be called a "global societal mind," as distinct from the global Web 
mind
that is my central topic of interest here. Both the global societal mind and 
the
global Web mind, however, are specific manifestations of the general concept 
of
a "global brain" -- an emergent, distributed worldwide intelligence.
Russell ties the global brain in with new-age, consciousness-raising 
practices.
By meditating and otherwise increasing our level of individual 
consciousness, he
suggests, we bring the awakening of the global brain closer and closer. When
there is enough total awareness in the overall system of humanity, humanity
itself will lock into a new system of organization, and will become an
autonomous, self-steering, brainlike system.
Speaking generally, one can envision the global Web mind as leading to a 
global
societal mind a la Russell in two different ways. First, we might actually
become part of the Web in a physical sense. This could be accomplished 
either by
downloading ourselves into computers, by the fabled "cranial jack," or by 
some
kind of true VR interface. Or it could be done by incorporating our existing
bodies into the Web, via newfangled sensory and motor devices. Imagine 
brains as
Websites, and modem/cell-phones inserted directly into the sensory cortex! 
Or,
secondly, we might become part of the Web via our actions, without any extra
physical connections. This is already coming about, at least among certain
sectors of the population. As more and more of our leisure and work 
activities
are carried out via the Internet, more and more of our patterns of behavior
become part of the potential memory of the global Web mind.
Webmind or similar software, implemented widely and used intensively across
intranets, could lead to this effect quite easily.
The global Web mind and the global societal mind, then, are not really such
different things at all. If a global Web mind comes about, it will clearly 
link
humans together in a new way, leading to some kind of different and more
intelligently structured social order. This is one flavor of global brain. 
On
the other hand, if a global societal mind comes about, communications 
technology
such as the Internet will doubtless play a huge role in its formation. This 
is
another flavor of global brain. The question is, will there be an 
intelligent
Web interacting with humans in a subtle way, or will there be an intelligent
societal system incorporating the Web, human beings, and all their 
interactions.
What kind of global brain will we actually see?
In fact, Russell's book is only the best known of a host of independent
discoveries of the concept of a global superorganism over the past few 
decades.
Joel de Rosnay, for one, has published several books in French on the notion 
of
a "cybionte" or cybernetic superorganism. His earliest, Le Macroscope, was
published in 1975; L'Homme Symbionte, which appeared in 1996, updates the
concept with discussions of chaos theory, multimedia technology and other 
new
developments. And Valentin Turchin, in his 1977 book The Phenomenon of 
Science,
laid out an abstract, cybernetic theory of evolution, and used it to discuss 
the
concept of an emerging, meta-human "superbeing." His crucial concept is the
metasystem transition, a word for the phenomenon in which was previously was 
a
whole, suddenly becomes a part. For example, the cell, which has its own
systemic unity, its own wholeness, becomes a part when it becomes part of 
the
human organism. There is a metasystem transition between cells and 
organisms.
There is also a metasystem transition between computers and networks: one's 
PC
at home is a natural whole, but the networked PC of the year 2000 will be
something quite different, in that most of its software will require 
interaction
with outside computers, and most of its behaviors will be incomprehensible
without taking into account the network outside it.
Currently humans are whole systems, with their own autonomy and 
intelligence,
and human societies display a far lesser degree of organization and
self-steering behavior. But, according to Turchin, a transition is coming, 
and
in the future there will be more and more intelligent memory, perception and
action taking place on the level of society as a whole. Turchin's vision is 
one
of progressive evolution: as time goes on, one metasystem transition after
another occurs, passing control on to higher and higher levels. One of 
Turchin's
most active contemporary followers is Francis Heylighen, of the Free 
University
of Brussels. Heylighen believes that the Web will be the instrument that 
brings
about the meta-system transition, leading from humanity to the meta-human
superorganism. The Principia Cybernetica Website, which he administers, 
contains
an extensive network of pages devoted to super-organisms, meta-system
transitions, global brains, and related ideas. Together with his colleague 
John
Bollen, he has also experimented with ways of making the Web more 
intelligent,
by making its link structure adaptive, in the manner of a neural network.
Heylighen has done a comprehensive world-wide search for literature on the
global brain, and posted the results at Principia Cybernetica.
So far, rather than debating the merits of different approaches to making 
the Web
intelligent, the discussion group seems inevitably to veer toward the
philosophical -- toward the questions of what the global Web brain will be 
like,
and how it will relate to human beings, individually and collectively.
The most striking thing about the discussion on the Global Brain Study Group
list is not a presence but an absence -- the absence of serious disagreement 
on
most issues regarding emerging Web intelligence. Everyone who has thought
seriously about the global Web brain, it seems, has come to largely the same
conclusions. The Web will become more self-organizing, more complex, and
eventually the greater intelligence of its parts will lead to a greater
intelligence of the whole. Human beings will be drawn into the Web 
intelligence
one way or another, either by mind-downloading or virtual reality, or simply 
by
continual interaction with Web-brain-savvy technology. In this way, human
society will begin to act in a more structured way -- in a way directed by 
the
global Web mind, which recognizes subtle emergent patterns in human 
knowledge,
and creates new information on the basis of these patterns.
A brief stir on the Global Brain Discussion Gruop was made by the appearance 
of
a brief satirical piece in Wired in mid-1996, called The Human 
Macro-Organism as
Fungus (online at 
http://www.hotwired.com/wired/4.04/features/viermenhouk.html).
This article is an interview with a fictitious scientist named Dr. 
Viermenhouk,
who parodies Heyligen by taking the absurdist line that the global 
superorganism
is already here. Here is the final part of the interview:

The Internet provides a big leap forward. As an organism grows more complex, 
it
requires a sophisticated means of transferring data between its constituent
entities. The Internet is little more than the nervous system of our human
macro- organism.
Isn't your work derivative of other cybertheoreticians? Francis Heylighen, 
for
example, has postulated the technology- driven transformation of humanity 
into a
"super- being" or a "metabeing." Heylighen ...
... walks around all day with a printer port up his ass. I've seen the 
pictures.
He's obsessed with a direct neural interface. His concept of a metabeing, a
single unitary organism, hinges on us physically plugging into a "super- 
brain."
He's missing the point. We already have.
Cells don't communicate through direct physical connections; they use 
electrical
interfaces. The neural cells in our skulls communicate through an intricate
chemical dance. To expect a macro- organism to develop differently from a
multicellular organism is foolish.
Now that we monkeys are part of a greater being, the connection we share is
through symbol. Human language, with all of its limitations, is sufficiently
complex to support the information- transfer needs of an organ- ism never 
seen
before on Earth. You don't need wires up your butt. Just look at the symbols 
on
your screen. Click on that hypertext link. Send that email. Be a good little
cell.
And Heylighen's bizarre notion that this metabeing is an improvement; 
delusion!
Individual humans are intriguing, sensual, spiritual creatures. The human 
macro-
organism is more of a fungus. A big, appallingly stupid fungus. It only 
knows
how to eat and grow, and when all of the food is gone, it will die. It has 
all
the charm and wit of something growing in a dark corner of your basement. 
Adds a
whole new dimension to the concept of human culture.
But what of individuality?
Humans are already too specialized to survive outside of their host 
organism.
Pull a nerve cell from the brain and put it on the ground; within minutes 
it's a
tiny gray blob of snot. Pull Bill Gates out of his office and put him in the
veldt; in four days he's a bloated corpse in the sun. With the exception of 
a
few madmen and anarchists, most of us can't even feed ourselves anymore; or
communicate outside of our specialized fields. Put an anthropologist, a 
supply-
side economist, and a mechanic in the same room. What the hell will they 
talk
about? O. J. Simpson?
David Williams' notion of the superorganism as a fungus is humorous, but it 
also
conceals a serious point. Yes, the fictitious Dr. Viermenhouk is wrong; the
superorganism is not here yet, at least not in full force. But when it is 
here,
will it necessarily be a boon to humanity? Or will it, indeed, be a fungus, 
a
parasite on man, sucking the life-blood from human-created technology for 
its
own purposes?
Heylighen himself appears to have taken the parody in good cheer But not all
global brain advocates have been so charitable. Valentin Turchin, for one, 
was
deeply annoyed. In a message posted to the Global Brain Study Group, he made 
the
following remarks:
Wired's interview with "Dr.Viermenhouk" which Francis calls a parody and 
fun, is
rather a lampoon, in my view.
...
The only "fun" in the interview is the vulgar language, which allows the 
author
to convey his negative emotion. I think that he is genuinely outraged by the
very idea of direct (not through language) brain links. And he is speaking, 
I
believe, for the majority.
The fact that we take this idea seriously, and explore its significance and
various aspects, will upset most people. We must be prepared for this. I 
have
already had some experiences of this kind....
Turchin believes that the global brain will have deep, positive, profound 
human
meaning. That it will provide a way of bridging the gaps between human 
beings
and fusing us into a collective awareness -- something that spiritual 
traditions
have been working on for a long time. From this point of view, direct
brain-computer links should not be viewed as tools for escape from human
reality, but rather as gateways to deeper connection with other human 
beings.
And, from this point of view, Williams' remarks are destructive, pointing 
the
readers of Wired away from something genuinely valuable -- they are about as
funny as someone going into schools and teaching children that vegetables 
are
bad for your teeth.
It is not only the fictitious Dr. Viermenhouk, however, who has a negative
attitude toward the global brain. Related fears have been voiced by Peter
Russell himself, who started a thread in the Global Brain Study Group on the
striking topic: Superorganism: Sane or Insane. Russell says,
I first explored the notion of superorganisms in my book "The Global Brain" 
- -
written back in the late seventies before the Internet really existed. There 
I
showed that, from the perspective of general living systems theory, human
society already displays 18 of the 19 characteristics of living organsims 
(the
missing one is reproduction - we haven't yet colonised another planet, 
although
we have the potential to).
...
The interesting question for me is not whether a global brain is developing. 
It
clearly is. But will this growing global brain turn out to be sane or 
insane? If
civilization continues with its current self- centred, materialistic 
worldview
it will almost certainly bring its own destruction.
I have long been fascinated by the striking parallels between human society 
and
cancer. Cancers have lost their relationship to the whole, and function at 
the
expense of the organism - which is insane, since a successful cancer 
destroys
its own host. This is what we appear to be doing, and very rapidly. Our
embryonic global brain would seem to have turned malignant before it is even
fully born.
I believe the reason for our collective malignancy comes back to individual
consciousness. We are stuck in an out- dated mode of consciousness, one more
appropriate to the survival needs of pre- industrial society. Thus the real
challenge is for human consciousness to catch up with our technology. We 
need to
evolve inwardly before any of our dreams of healthily-functioning global 
brains
can manifest.
This is more intelligently and respectfully stated than Williams' parody, 
but in
the end it is somewhat similar. Instead of fungus, we have cancer -- a far
better metaphor, because cancer cells come from within, whereas fungus comes
from outside. Russell believes that we are on a path toward the emergence of 
the
global brain, and that the Web is just one particular manifestation of this
path. But, observing that we humans ourselves are riddled with neurosis and
interpersonal conflict, he wonders whether the collective intelligence that 
we
give rise to is going to be any better off.
On the one hand, Russell believes that the global brain will go beyond
individual human consciousness, with all its limitations. In response to a 
post
of mine, questioning whether the Internet might eventually develop a sense 
of
"self" similar to that of human beings, he responded as follows:
The question is whether this superorganism will develop its own 
consciousness -
and sense of self - as human beings have done.
Back then [in The Global Brain] I argued that there were close parallels 
between
the structure and development of the human brain, and the structure and
development of the global telecommunication/information network, which 
suggested
that when the global nervous system reached the same degree of complexity as 
the
human nervous system, a new level of evolution might emerge. But it would be
wrong to characterize this new level as consciousness. It would be as far 
beyond
consciousness, as we know it, as our consciousness is beyond life, as a 
simple
organism knows it. So I don't think discussions as to whether the global 
social
superorganism will develop a self akin to ours are that relevant.
Despite this conviction that the global brain will be far above and beyond 
human
consciousness and human mental dynamics, however, he is worried that the 
flaws
of individual human consciousness may somehow "poison" this greater emergent
entity, and make it fatally flawed itself.
Responses to Russell's pessimistic post were mixed. Gregory Stock, for 
instance,
took issue with Russell's generally negative judgement ofthe psychology of 
the
average modern human. A biologist, Stock views human selfishness and
shortsightedness as biologically natural, and believes that modern society 
and
psychology, for all their problems, are ultimately wonderful things. His 
book
MetaMan treats contemporary technological society as a kind of 
superorganism,
and views this superorganism in a very positive light.
Turchin, on the other hand, agrees substantially with Russell's pessimistic 
view
of human nature and its implications for the mental health of the 
superorganism.
He believes, however, that it may be possible to cure human nature, at the 
same
time as developing new technologies that extend human nature, overcoming its
limitations:
>We need to evolve inwardly before any of our dreams of healthily 
>functioning
global
>brains can manifest
Yes. This is why the Principia Cybernetica project came into being. Our goal 
is
to develop - - on the basis of the current state of affairs in science and
technology - - a complete philosophy to serve as the verbal, conceptual part 
of
a new consciousness.
My optimistic scenario is that a major calamity will happen to humanity as a
result of the militant individualism; terrible enough to make drastic 
changes
necessary, but, hopefully, still mild enough not to result in a total
destruction. Then what we are trying to do will have a chance to become
prevalent. But possible solutions must be carefully prepared.
More positive than Turchin or Russell, though less so than Stock, the 
physicist
Gottfried Mayer-Kress expressed the view that, perhaps, the global brain 
itself
might represent the solution to the problems of individual human 
consciousness,
rather than merely transplanting these problems onto a different level:
Peter Russell writes:
>The interesting question for me is not whether a global brain is 
>developing. It

>clearly is. But will this growing global brain turn out to be sane or 
>insane?
If >civilization continues with its current self- centred, materialistic
worldview it will >almost certainly bring its own destruction.
...
I thought a coherent world civilization was what we expect to emerge from a
GlobalBrain.
In the context of complex adaptive systems we always would expect a "self-
centred, materialistic worldview" of all subsystems (e.g. cells in the body,
nation based civilizations etc.). Through the emergence of order parameters 
the
subsystems begin to act in a coherent fashion and thereby increase the 
payoff
for the subsystem.
A self-organized structure (order- parameter etc) will be stabilized if in 
the
competition between the interests of the sub- system and those of the super-
system if the subsystem recognizes that it is better off if it supports the
larger structure (e.g. pay taxes etc). A neccessary condition for that to 
happen
is that the coupling (communication) between the subsystems is strong enough
that local modes cannot grow at the expense of the order parameter. In a 
social
context that could mean that violations of the global rules/laws could be
detected and suppressed.
For example, on a global scale it is still cheaper for most nations to 
choose to
pollute the environment and waste energy. In a GlobalBrain world China would
recognize that it is better not to introduce large scale individual
transportation (cars) and Brazil would find it better for its own economy 
not to
destroy the rainforest.
Regarding the "cancer" metaphor, Meyer-Kress observes that
even an "embryonic global brain" would be a coherent global structure and
thereby directly contradict the basic definition of cancer. I would see the
cancer analogy more the global spread of a drug culture.
Essentially, Mayer-Kress's point is as follows: saying that humans are
"individualistic" is the same as saying that humans represent the "top 
level" of
a hierarchy of systems. An individualistic system is just one that has more
freedom than the systems within it, or the systems that it is contained in.
Cells within individual organisms are individualistic only to a limited 
extent;
they are behaving within the constraints of the organism. Cells that make up
single-celled organisms, on the other hand, are far more individualistic: 
they
have more freedom than the systems of which they are parts.
The global brain, according to Mayer-Kress, is almost synonymous with the
decrease of human individualism. We will still have individual freedom, but 
more
and more it will be in the context of the constraints imposed by a greater
organism. And so, in this view, Russell's idea that the global brain might
inherit the problems caused by human "self-centredness" is 
self-contradictory.
The global brain, once it emerges, will be the top-level system, and will be
individualistic -- but, as Russell himself notes, the nature of its
individualism will be quite "inhuman" in nature.
Mayer-Kress, in this post, did not address the question of whether the 
global
brain would be sane or insane in itself; rather, he defused the question by
breaking the chain of reasoning leading from human neurosis to global brain
neurosis. In my own reply to Russell's message, on the other hand, an 
attempt
was made to take the bull by the horns and answer the question: What would 
it
even mean for a global Web brain to be insane?
About sanity or insanity. Surely, these are sociocultural rather than
psychological concepts. However, they can be projected into the individual 
mind
due to the multiplicity of the self. An insane person in a society is 
someone
who does not "fit in" to the societal mindset, because their self- model and
their reality- model differ too far from the consensus. In the same vein, if 
one
accepts the multiplicity of the individual self (as in Rowan's book
SUBPERSONALITIES), one finds that in many "insane" people, the different 
parts
of the personality do not "fit in" right with each other. So the jarring of
world- models that characterizes the insane person in a culture is also 
present
within the mind of the insane person. Because, of course, the self and mind 
are
formed by mirroring the outside!
What does this mean for the global brain? View the global brain as a 
distributed
system with many "subpersonalities." Then the question is not whether it is 
sane
with respect to some outside culture, but whether it is sane WITH RESPECT TO
ITSELF (a trickier thing to judge, no doubt). Do the different components of 
the
global brain network all deal with each other in a mutually understanding 
way,
or are they "talking past" each other...
A key point to remember here is that the global brain can be, to a large 
extent,
real- time engineered by humans and AI agents. So that, if any kind of
"insanity" is detected, attempts can be made to repair it on the fly. We are 
not
yet able to do this sort of thing with human brains, except in the very 
crudest
way (drugs, removing tumors, etc.).
The belief which I expressed in this post is that the sanity of the global 
Web
brain is an engineering problem. By designing Web software intelligently, we 
can
encourage the various parts of the global Web brain to interact with each 
other
in an harmonious way -- the hallmark of true sanity. The various neuroses of
human mind and culture will be in there -- but they will be subordinate to 
an
higher level of sanely and smoothly self-organizing structure.
The biggest potential hang-up, in this view, is the possibility that forces 
in
human society may intervene to prevent the software engineering of the Web 
mind
from be done in an intelligent way. Perhaps it may come about that a 
maximally
profitable Web mind and a maximally sane Web mind are two different things. 
In
this case we will be caught in a complex feedback system. The saner the Web
mind, the saner the global brain of humanity, and the less likely the forces 
of
greed will be to take over the Web mind itself.
One thing is noteworthy about this particular thread on the Global Brain 
Study
Group: in spite of our disagreements on details, everyone in the Study Group
seems to concur that a healthy, sane global brain would be a good thing. An
alternative view was given by Paulo Garrido, in a message on the Principia
Cybernetica mailing list, forwarded by Heylighen to the Global Brain Study
Group. Garrido made the following remarks:
IF human society is an organism (in the autopoietic sense) and has a (the 
super)
brain THEN most probably we should KILL such a being.
Because, societies, or better, the social interaction should be a TOOL to
enlarge individual power and freedom or, if one prefers, individual survival 
and
development. There is no point in maintaining a society if it is not that. 
If a
society becomes an organism, chances are that individual power and freedom 
are
diminished: to exist as such an organism must limit the degrees of freedom 
of
its components. And in the case of human societies - the components are us! 
Only
one type of autopoietic system should be allowed to emerge as a result of 
social
interactions: the one that enlarges individual power and freedom - for all 
the
individuals. Maybe such a system is possible if it is built in the emotional
domain of love, corresponding to the goal of development. If it is not the 
case,
it should be destroyed. Otherwise, we may see ourselves with no survival or
comfort problems ... and with no reason to live.
Garrido's remarks, though somewhat paranoid in tone, are well-thought-out 
and
represent a natural human fear. Are we all going to be absorbed into some 
cosmic
organism, to lose our human individuality, our freedom, our sense of 
individual
feeling and accomplishment? After all, does computer technology not 
represent
the ultimate in de-humanizing technology?
The difficulty, of course, is that freedom is difficult to assess. Every 
major
change in the social order brings new freedoms and eliminates old ones. And 
the
relative "goodness" of one thing or another is not an objective judgement 
anyway
-- standards of morality vary drastically from culture to culture.
4. The Internet as Collective Unconscious
A different but related perspective may be found by introducing the ideas of 
the
psychologist Carl Jung. The Web provides a whole new way of thinking about
Jung's concept of the collective unconscious -- a realm of abstract mental
forms, living outside space and time, accessible to all human beings, and
guiding our thoughts and feelings. And it gives new, concrete meaning to his
concept of archetypes -- particular mental forms living in the collective
unconscious, with particular power to guide the formation of individual 
minds
(Jung, 1955).
The concept of the collective unconscious has never achieved a highly 
respected
status within scientific psychology. Science, perhaps rightly, perhaps (as 
Card,
1996, argues) wrongly, has no place for an incorporeal realm of abstract 
forms,
interacting with individual minds but standing beyond them. The global Web 
mind,
however, will actually be an incorporeal -- or at least digital -- realm of
abstract forms, interacting with individual minds but standing beyond them!
Some of the "archetypal forms" that Jung believed we absorb from the 
collective
unconscious are basic psychological structures: the Self, the Anima/Animus 
(the
female and male within), the Shadow. Others are more culture in nature, e.g. 
the
First Man. Some are visual: e.g. the right-going spiral, signifying being
"sucked in"; the left-going spiral, signifying being "spewed out." But the 
most
basic archetypes of all, in Jung's view, are the numbers. Small integers 
like 1,
2, and 3, Jung interpreted as the psychological manifestation of order. In 
fact,
Jung suggested that all other archetypes could be built up out of the 
particular
archetypal forms corresponding to small integers. This is a strikingly
computer-esque idea: it is a "digital" view of the world, in the strictest
sense. And so we see that Jung's thought, for all its obscurity and
spirituality, was at bottom very mathematical: he viewed the abstract 
structures
of the mind as emanating from various combinations of numbers. He viewed the
collective unconscious as a digital system.
The global Web mind will, I suggest, fulfill Jung's philosophy in a striking 
and
unexpected way: it will be a digital collective unconscious for the human 
race.
For after all, the memory of the global Web mind is the vast body of
humanly-created Web pages, which is a fair representation of the overall 
field
of human thought, knowledge and feeling. So, as the global Web mind surveys 
this
information and recognizes subtle patterns in it, it will be determining the
abstract structure of human knowledge -- i.e., determining the structure of 
the
human cultural/psychological universe. This is true even for the global Web 
mind
as an independent entity; and it will be far more true far more true if, as
human beings integrate more and more with the Web, the global Web brain
synergizes with humanity to form a global digital/societal mind.
Specifically, the most abstract levels of the global Web mind will bear the
closest resemblance to the collective unconscious as Jung conceived it. 
These
levels will be a pool of subtle, broad-based patterns, abstracte from a huge
variety of different human ideas, feelings, and experiences, as presented on 
the
Web. And this body of abstract information will be active. Initially, it 
will be
involved in creating new links on the Web, in creating new Web content, in
regulating various real-world and virtual activities based on this content. 
And
more and more, as it grows more pervasive, it will become involved in an
interactive way with human thoughts and feelings themselves. In other words,
precisely as Jung envisioned, the digital collective unconscious will be
involved in forming the thoughts, feelings and activities of human beings'
individual consciousnesses.

                              Chapter 7

                       The Webmind Architecture


1. Webmind
AI is defined in the Artificial Intelligence Dictionary as "a 
multi-disciplinary
field encompassing computer science, neuroscience, philosophy, psychology,
robotics, and linguistics; and devoted to the reproduction of the methods or
results of human reasoning and brain activity." In spite of these lofty
ambitions, however, the reality of AI has up to this point been severely
disappointing. Outside the academic world, AI has adopted a role as one
technology among many, suitable to be applied to tasks that are too 
non-routine
for ordinary software, and are too repetitive or mathematically involved for
humans to do efficiently, but do not require a general, intuitive 
understanding
of the world. As such, various AI technologies have proved particularly 
useful
in the finance industry, and in other specialized areas such as process 
control.

Webmind, the Java-based AI system under development
overcomes many of these historical limitations of AI technology. Whereas
previous AI technologies focus on specific mental qualities such as 
knowledge
representation, perception, reasoning, or learning, Webmind is based on a
comprehensive mathematical model of the mind as a whole. It perceives
information, analyzes information, remembers information and acts on 
information
in an integrated way. It adapts to its own environment much as human beings
adapt to the physical world. It evolves its own spontaneous structures, 
rather
than relying too heavily on the intuitions of human programmers.
Webmind does not have a general intuitive understanding of our world, but it
does have a general intuitive understanding of its own world, the data 
stored on
the Internet, and the interests and responses of human Internet users. This 
is
the qualitative difference between Webmind and previous AI systems; and this 
is
why Webmind has the potential to bring AI beyond the role of a specialized
adjunct to unintelligent computer systems, and to make AI the heart of 
computing
itself as we move on into the new era of super-powerful digital networks.
At the present time, the domain in which Webmind has been most 
systematically
and scientifically tested is financial analytics. The goal in these tests 
was
simple, to be able to make better financial predictions and trading 
decisions
than one could without the software. The finance industry has already 
developed
sophisticated tools for projecting trends in markets, using quantitative 
data
about what the markets have done in the past. The problem with these 
techniques
is that they do not encompass the full realm of information available to the
markets.
The key advantage of Webmind for this purpose is that it can incorporate a
logical analysis of textual material, material from financial news sources.
This, of course, is very similar to what market analysts do. They study the
statistical data, then interpret it in view of information which they obtain
from their reading and conversations with industry leaders. However, there 
is a
tremendous mass of written material available, and not enough time for any
analyst to digest it all. Traders for large investment firms have difficulty
even finding the time to read the reports generated by their own staff 
analysts.
Webmind computerizes this process.
Webmind deals with textual and numerical information on an equal footing, 
freely
making generalizations spanning different types of data. Unlike any other AI
system, it forms its own intuitions regarding concepts by dynamically
"grounding" linguistic concepts in terms of its own nonlinguistic 
experience. It
is the first AI system able to create its own concepts that apply across
different contexts. Webmind runs efficiently on powerful stand-alone 
computers,
but is most powerful when run over a network of computers, in which case its
sophisticated server-server communication methods allow its internal network
structure to harmonize with the connectivity structure of the computer 
network.
Webmind does not aim to emulate human intelligence in an artificial way, but
rather to be a natural intelligence for the digital world -- solving human
problems with digital intuition. Initial tests of Webmind's ability in the 
area
of financial decision support have proved dramatically successful. Further
application areas under active development include information retrieval, 
data
mining, and enterprise analysis.
Though it is a new kind of software system, Webmind incorporates elements or
functions of various standard AI technologies. It can serve as a search 
engine,
as a Web visualization tool, or as a data mining engine -- and it can also 
serve
functions not implied by any of these pre-existing categories. Most
fundamentally, it is a system for the structuring and dynamic evolution of 
large
bodies of data. It is an intelligent system for knowledge management and
knowledge creation, with an almost boundless array of potential applications 
in
business, science and personal computing.
Webmind collects and analyzes information in much the same way as the human 
mind
does, but with different strengths and weaknesses, befitting its physical 
basis
in silicon rather than carbon, and its perceptual embedding in the world of 
data
and text feeds rather than the world of sights, smells, sounds, tastes and
touches. In the human mind, everything above the lowest levels of perception 
and
action is dynamic and constantly evolving. The memory constantly studies 
itself,
reorganizing to enhance and inhibit recall and performance. Current 
information
discovery tools impose an artificial distinction between static data
repositories and dynamic search and analysis tools. But what is needed, in 
order
to really make use of the vast amount of information we are amassing, is a
dynamic information system that mimics, on an abstract level, the structure 
of
the human mind, with all its reflexive, self-analytical capabilities. What 
is
needed is a way to make the data come alive and study itself, rather than 
lying
there inert and being probed by various computational instruments, of 
various
degrees of intelligence. This is what Webmind accomplishes. It allows
collections of data to structure themselves and recognize patterns in
themselves. Webmind transforms each informational element within a computer
network into an independent agent which cooperates with other agents in
organizing and reorganizing data for search and retrieval. It interacts with 
and
builds on the intelligence of the user. Users' queries are treated as
informational agents interacting via the collective Webmind agents system to
locate information. Webmind agents take advantage of the sum collection of 
their
organization's intelligence to discover context and concepts on behalf of 
users
and their applications. In short, Webmind is the first unified computational
mind, complete with integrated systems for perception, action, cognition,
language, short and long term memory, etc., all embedded in a data-rich
environment.

2. From Theory to Implementation via Five Questions

According to the psynet model, as reviewed above, the achievement of
intelligence does not depend on the implementation of particular mechanisms, 
but
rather on the implementation of mechanisms that give rise to appropriate
emergent structures. The unique achievement of the Webmind design is to 
cause
these emergent structures to emerge from current computers and networks in a
computationally efficient and humanly useful way. Underlying the psynet 
model is
a vision of intelligence as "the capability to achieve complex goals in 
complex
environments." Webmind's complex goals are firstly posed to it by human 
users
and secondarily evolved by itself; its complex environment consists of the 
human
beings, machines and networks with which it interacts.
The Webmind system provides a general "agents operating system" for managing
systems of software agents that share meaning amongst each other, transform 
each
other, and interact in various ways; the agents may live in the RAM of a 
single
machine, may be run on multiple processors, and may live across many 
machines
connected by high-bandwidth cable. It provides mechanisms for agents to
represent patterns they have recognized in other agents, and patterns they 
have
recognized in the overall agents system of Webmind. And it also provides an
appropriate assemblage of agents, specialized for such things as
    text processing
    numerical data processing
    text-numerical interrelation
    supervised and unsupervised concept formation
    reasoning
    social interaction
    language processing
These agents have their own localized behaviors but achieve their true
intelligence only in the context of the whole Webmind system.
Several important questions arise in designing an AI implementation of the
psynet model:
    How to represent the agents
    How to manage the dynamics of a collection of agents, on the hardware
    provided
    What to build in, and what to require the system to learn
    How to embed the system in a rich environment, including other systems 
of
    like kind
    What goals to give the system
    What programming language to use?
How one answers these questions determines what kind of system one comes up
with. One way to get across the essence of Webmind is to describe how these
questions were answered in the design of Webmind, on a general level.



Question 1: How to represent the agents?
Answer: As nodes in an abstract network, and relational and mobile agents
associated with nodes
In Webmind agents are realized as data structures called "nodes," with
particular structural and dynamical properties. A psynet node is defined as 
a
container for one or more of the following:
    possibly some "raw" perceptual data
    a record of the meaning of the current node, in the context of the mind 
as a
    whole
    a transformation which acts on a node, and produces a node
    a transformation which acts on a node, and produces a number
    tables of links to nodes with which it shares meaning in certain ways
    mobile agents which seek other nodes to add to the link tables
    information on how much attention is paid to the node currently and over 
the
    recent past, in total and from the perspective of various other nodes
    information about how important the node is, and how fast its importance 
is
    expected to decay over time
It should be clear that the node data structure, defined in this way, is
adequate for expressing the "agent" concept of the psynet model of mind.
Transformations taking agents into agents and agents into numbers are here, 
as
required; and the dynamical law of mind, which says that attention follows
meaning, is explicitly supported via each node carrying lists of other nodes
that share its meaning, and agents searching for nodes that share its 
meaning.
The view of the mind as a network of psynet nodes is a kind of "middle way" 
for
AI -- or in other words, an "intermediate level" model of intelligence. It
stands between, on the one hand, neural network and other physiologically
inspired models of intelligence; and on the other hand, abstract logical and
psychological models of intelligence that speak of mind in the abstract,
completely apart from its physical implementation. Neural network and 
logical
and psychological models all have their particular validity, but considered 
as a
collective, they leave a crucial gap: how do the complex, self-organizing
physical dynamics of neural networks and the like give rise to the 
abstractions
of the mind, such as syntax, logic, inference and self? "Psynet" means "mind
network," and so it is only appropriate that the network interpretation of 
the
psynet model fills in this gap by positing a network that is not a network 
of
neurons but a network of intermediate-scale, mental entities ... "agent" 
nodes.
In a neurophysiological context, it may be useful to think of a psynet node 
as a
neuronal group in the sense of Edelman's Neural Darwinism brain theory: each 
of
Edelman's neuronal groups carries out a specific useful function, quite
different from a single neuron, which in most cases has no particular 
function
in itself. In some computer implementations of the psynet model, it may be
useful to take this mapping fairly literally, and implement nodes in terms 
of
networks of formal neurons -- as we shall see, something like this is done 
in
Webmind, in certain cases, though not in general. But this is not a 
necessary
approach: psynet nodes could be implemented as LISP program trees, as
high-degree polynomials, as physical entities outside the realm of digital
computing (Bose-Einstein condensates?), etc. The psynet model is abstract: 
it
has to do with the structure and dynamics of the network of agents, not with 
the
internal makeup of the agents themselves. Mind design, in this perspective, 
is
largely about creating the correct agents out of the tools at hand, to 
ensure
the emergence of the desired structures and dynamics from the system as a 
whole.
In "agent system" terms, we may say that agents in Webmind fall into several
classes. First there are nodes, static agents, which have a relatively
persistent state. Nodes may contain information relating to data such as 
texts
or database files or internal information such as categories or other 
patterns
recognized among other nodes. Nodes are not truly static, they update 
themselves
continually, but they represent the backbone of 's intelligent internal 
network.
Most of the information contained in a node actually consists of a record of 
the
node's relationship to other agents in the Psynet. There are many different
types of node, pertaining to different types of information.
In addition to nodes, there are relational agents or links, which establish 
and
retain relationships among nodes. These relationships may fall into many
different categories, and may be parameterized by qualities such as 
intensity,
uncertainty, and so forth. Finally, relational agents are formed by the 
third
kind of agent, mobile agents which move around within the internal 
relational
geometry of an individual Psynet, potentially from one server to another,
actively seeking new relationships to embody in relational agents. Mobile 
agents
associated with nodes can also move from one to another, across the Net, but
this requires a different brokering mechanism than the free flow of agents
within the same server, or between different servers on the same Intranet.

Question 2: How to manage the dynamics of a collection of agents, on the
hardware provided
Answer: On contemporary computers and networks, collections of nodes must be
stored in large amounts of RAM; they must be dynamically groupable into sets 
to
allow multiprocessing; and they must be able to migrate and send messages 
across
network cable in order to allow a single mind to span multiple machines
The dynamics of the network of nodes in Webmind's psynet is as follows. When 
a
node receives attention, it spreads some of its attention to the 
meaning-sharing
nodes that it knows about -- i.e. attention follows the links in its link 
table.
It may also update its record of its own meaning. Finally, it may carry out 
its
transformations to produce new nodes, or it may modify its transformations, 
with
the goal of making them more prominent patterns in the collection of other 
nodes
with which it shares meaning.
Ideally, each node would run independently on its own processor. However, 
that
is not feasible at the present time: massively parallel processors do exist, 
but
are orders of magnitude less efficient than networked top-of-the-line serial
computers. Fortunately we are presently blessed with computers that, though
serial in nature, possess multiple rapid processors, large amounts of RAM 
(in
the gigabyte and up range), and fast cables (e.g. Ethernet) networking them
together. This combination allows us to run large magician systems on serial
machines.
The system architecture of Webmind is roughly as follows. A single mind 
network,
may run across a number of machines, interconnected by high-bandwidth
connections. Nodes may send messages across machines, and may share meaning 
with
nodes on other machines; they may also move from one machine to another if 
this
is judged dynamically efficient. Each machine contains a managerial process
which determines when dispatching of nodes and messages between itself and 
other
machines is appropriate. This managerial process must pay attention to the
psynet dynamics on its own machine, for various reasons.
Within a single machine, nodes are divided into groups called pods. Grouping
into pods is useful for human saving/loading of groups of nodes -- a pod of
nodes may serve a particular human purpose, e.g. it may contain nodes 
concerning
personnel records, nodes concerning financial data, etc. It is also useful 
for
multiprocessing: a pod must live on a single processor, but multiple pods 
living
on multiple processors may be very elegantly and efficiently handled. When a
node wants to transform or spread attention to another node, living in a
different pod, it sends a request to its pod's managerial process, which 
then
sends the request on to the managerial process of the pod at the receiving 
end.
Question 3: What to build in, and what to require the system to learn?
Answer: Build in "preprocessing" routines for transforming particular kinds 
of
input data into patterns, and for producing particular kinds of output data 
from
node states. Build in a vocabulary of useful magician types. Beyond this, 
build
in other things as short-cuts, but nothing which the system is not capable 
of
learning on its own in a non-absurd period of time
During the early development of Webmind, this was the toughest question to
answer. One could seek to implement the psynet model of mind by setting up a
kind of "primordial soup" of transformations, mapping into numbers and 
mapping
into other transformations. Assuming this soup was hooked up to rich data 
feeds,
given extra resources in return for success at practical tasks, and given a
reasonable capacity to mutate itself, it would eventually transform itself 
into
something intelligent, arriving at the structures described above by
self-organizing evolution. This methodology, however, is extremely 
inefficient,
and is extremely unlikely to work given the limitations of current 
engineering
technology. I tried this approach in 1994 and failed miserably; my comoc.com
collaborator tried it in a different way in 1997, and also failed. This 
approach
is conceptuall flawed in that mixes up what in biology are three different
levels of intelligence: evolution, epigenesis, and learning. In order to 
arrive
at a practical implementation of the psynet model, one must build in more of 
the
desired structure -- just as a human being is born with some of the desired
structure of mind already present in the structure of the brain.
The question then becomes how much one can build in without sacrificing the
dynamicity, creativity, and self-modification that are at the essence of
intelligence. The correct answer, we have found, is to engineer:
    the processing of low-level perception and action
    the types of agents potentially in the system (which determine not which
    mind you have, but what kind of mind you have)
    the specific configuration of agents, but only as a guide for future
    learning and not as a fixed, inorganic base of knowledge
The built-in preprocessing of low-level perception and action is necessary
because of resource constraints. All that the human brain does by hard-wired
visual circuitry -- line extraction, etc. -- could be done by more 
intelligent
parts of the brain, but this would be a waste. The amount of pattern 
recognized
in the world is maximized by allowing pre-processing to take care of the 
lowest
level of pattern-recognition, and focusing the attention of intelligent on 
the
output of the preprocessors. In the current Webmind system this lesson is
reflected as follows: some basic text processing routines are built in, as 
are
some numerical data processing routines that recognize dates in numerical 
data
files, etc.
In the current system, a carefully chosen variety of node types is provided
also: nodes for text, nodes for numerical data, nodes for words/phrases, 
nodes
for senses of words, nodes for categories, etc. Each node type has specific
means for determining shared meanings and for transforming itself and other
nodes. The strategy is to build in the node types, and then allow the nodes, 
of
selected types, to self-organize into their own structure based on their 
inputs.
The key principle to remember here is the gradation of specialization. A 
system
consisting of only highly specialized node types will never have the 
flexibility
to be intelligent in the real world; but a system consisting of only highly
general node types (not evolved or engineered for any particular world) will
have a hard time coming to grips with its environment. Specialized node 
types
are a step up from perceptual/active preprocessing, but serve a similar 
role: to
enable the system to focus most of its attention on problems which require
general intelligence.
In the current Webmind system, we have highly specialized node types for two
things: language processing, and analysis of numerical data, particularly
financial data (e.g. a node type specifically representing a trading 
system).
Preprocessed data from these sensory modalities feeds into specialized nodes 
at
the same time as it feeds into general-purpose nodes, and the interaction
between the specialized and general-purpose nodes at this stage of 
processing is
intense.
The interplay of different node types is is in some sense the hardest part 
of
computational mind engineering. The essence of mind, as depicted by the 
psynet
model, is relatively simple: what distinguishes one type of mind from 
another is
mainly the types of agents that live inside it. The interaction between
specialized agents and general-purpose agents is something that, in humans,
evolved over millions of years; it can be engineered, but only with a firm
understanding of the general underlying dynamics of mind, and a healthy 
respect
for the difficulty of the task.
Question 4: How to embed the system in a rich environment, including other
systems of like kind?
Answer: The Internet!
Humans receive a vast amount of perceptual data via their sense organs; and 
we
interact with the same world producing this data via speaking, moving, 
stinking,
etc. In short, we are embedded in our bodies. The body of Webmind consists 
of
the computers on which it is implemented, which have their own sensory and
active organs, for reading information from other computers and from human
users. Pre-Internet, the data feeds available to computers were fairly 
sparse,
and the ability for a computer program to interact with humans and other
programs was strictly limited. At the present time, however, it is possible 
to
hook up Webmind to a vast number of textual and numerical datafeeds, to 
supply
with an unlimited amount of user interaction via corporate intranets and the
Web, and to allow it to act by sending messages to individuals and other
computers. For the first time, there is now a rich perceptual environment
available to a computational intelligence. Furthermore, there is the 
opportunity
to place a community of AI systems within this common environment, giving 
them
access to common perceptual data and the ability to interact with one 
another in
a variety of ways.
This is a somewhat obvious point, but it is one that has not been exploited 
by
the mainstream of AI researchers, who do not accept the essentiality of a 
rich
environment for the emergence of intelligence.
The social interaction of Webminds is explicitly mediated by data structures
corresponding to the Self (the Webmind itself), other Webminds (needed for
intelligent inter-Webmind querying), and human users (needed for intelligent
customized response).
Question 5: What goals to give the system?
Answer: The goal of giving meaningful answers for questions asked by 
important
entities (human and computer), at present and into the future
In order for a mind network to evolve effectively, it must be given goals to
carry out, and must be given feedback on the quality with which it has 
carried
out these goals. Just as humans are built to want to eat, have sex, drink,
exercise, etc., so Webmind is built to want to answer queries effectively. 
In
order to assure this, each time it sends a query, it should receive an
assessment of the perceived quality of its response. This may be explicit 
(e.g.,
the user may be asked), or it may be implicit (e.g., the client may 
calculate
the number of times the human user needed to rephrase his question to get 
the
answer desired). We all know how hard it is to guess what humans want -- 
what
this means is that answering human questions is certainly a "complex goal" 
as
required in our working definition of intelligence, which is "the ability to
achieve complex goals in complex environments."
Question 6: What programming language to use?
Answer: Java
The Java language was one of the inspirations for Webmind, so it might seem 
a
100% obvious choice. Even so, though, serious consideration was given to the
possibility of programming Webmind in C++, a language which has the 
advantage,
at this point, that C++ programs run faster than Java programs. But, in 
spite of
its current speed advantage, C++ lost the context, for a number of technical
reasons:
    Java shares with LISP and other artificial intelligence languages a 
feature
    called automatic garbage collection. This means that, when a certain 
piece
    of the computer's RAM is no longer being used by a Java program, the 
Java
    language will take the initiative to free up this piece of memory so it 
can
    be reused by another part of the program. In C/C++, this free-ing up of
    memory has to be done explicitly by the programmer.
    Java is Internet-friendly -- its sophisticated networking libraries 
allow
    flexible inter-machine communications to be programmed relatively easily
    Platform independence, for the client and the server! Programmed in 
Webmind
    can run on PC's, Macs or Unix boxes
    Multi-threading is a programming technique used to make programs that 
act
    like they're doing a lot of things at once, even though they're really
    running on a one-thing-at-a-time serial processor. Of all popular
    programming languages, Java provides the most effective and easy to use
    multi-threading. In particular, this means that, in Java, it's really 
easy
    to make a program that keeps doing its own internal processing in one
    "thread" of control while communicating with users or with other 
computers
    in other "threads."
With the advent of ultra-efficient compilers such as IBM's Formula, the 
speed
gap between C++ and Java is narrowing; our own tests show a ratio of 2.5:1. 
The
Webmind server will also be able to exploit the added power of the Java 
chip, to
be introduced in 1999. Finally, the much-publicized "Java browser wars" have 
no
significant effect on Webmind, as the Webmind server is not browser-based, 
and
the Webmind user interface doesn't need to run in the browser, or even need 
to
be written in Java - in fact, our first user interface is actually written 
in
C++, and talks to the Java portion of Webmind using (what else?) standard
Internet communication protocols.

3. How Webmind Works

The nitty-gritty details of the Webmind software architecture are 
proprietary
and cannot at this stage be discussed in detail in a public document. 
However,
some important aspects of the design may be discussed at a high level of
granularity, and may help the reader to grasp the system's essential nature.
A single Webmind system consists of a collection of nodes called a Psynet, 
which
spans a collection of one or more Webmind servers, each one running on its 
own
(possibly multiprocessor) machine. The intelligent core of a Webmind server 
is
its collection of nodes, which is called a PsynetLobe. Each PsynetLobe, in 
turn,
consists of multiple pods, each pod containing a number of nodes.
At the present time, the boundaries of a single Psynet are the boundaries of 
the
intranet or extranet: Ethernet connections are fast enough that Webmind 
servers
thus connected can act as an emergent whole; whereas Internet connections 
are
too slow. Thus, one has a model of the intranet as the individual; and the
Internet as a population of Webminds interacting in a manner best described 
as
"social", exchanging information and questions and answers but not 
participating
in collective trains of thought on a real-time basis. This of course may 
change
as bandwidth increases over the next decade; at some point the Internet will
become as fast as today's Ethernet, and a single global Psynet will become
achievable.
Every Webmind server comes with a standard collection of pods: pods for 
standard
data types such as text and numerical data, several pods for natural 
language
understanding, a pod for query processing, and other pods devoted to such 
tasks
as creative brainstorming, self-reflection, socialization, etc. Other pods 
will
be application-specific: e.g., a Webmind server aimed at financial 
applications
would contain a pod of historical financial data; an enterprise application
would contain a pod of transient e-mail data; etc. Additional pods may be 
loaded
and saved by the user.
The dynamics of the Psynet is fairly simple, though the emergent structures 
that
these dynamics give rise to may be quite complex. Each pod carries its own
threads, and these threads cycle through all the agents in the pod, allowing
them to carry out "elemental actions." Each node carries a "heat" indicating 
its
importance at a particular time, and hot nodes are given a greater priority.
The starting-point for Webmind's intelligence is a collection of basic node
types used to represent data found on current computer networks. There are
TextNodes for representing text documents, DataNodes for representing 
numerical
data, and several types of CategoryNodes for representing concepts inferred 
from
text and data. There are also TimeSets, representing collections of data 
that
occur over time; and TimeNodes, representing collections of information
considered roughly simultaneous. These node types are the basis for 
Webmind's
intelligence; other processes work with these nodes to generate meaningful
information of various forms. Information mined from data warehouses is read
into the system and translated into a collection of TextNodes, DataNodes,
CategoryNodes, TimeSets and TimeNodes; a flexible format which is amenable 
to
sophisticated intelligent analysis by Webmind's various internal agents.
What is universal to all nodes in Webmind is:
    attention (activation), current and cumulative
    links representing direct, local relationships to other nodes
    links containing information on other "distant" nodes that are bound up 
with
    the node in emergent dynamical reactions (this is the "halo" of the 
node)
    the process of spreading attention along links
    the process of building links to other nodes, by sending out agents 
called
    Wanderers
That is, all Webmind nodes manifest Peirce's Law of Mind, which states that
attention follows meaning, that each aspect of mind spreads some of its 
energy
to each other aspect of mind which stands to it in a "peculiar relation of
affectability."
Wanderer agents, resident at each node, travel through the network of nodes 
in
search of other nodes that stand toward their home node in a peculiar 
relation
of affectibility. Specifically, they seek out symmetric (similarity), 
asymmetric
(inheritance) and emergent relationships. In the case of emergent 
relationships,
they may create new nodes embodying emergent patterns between their home 
node
and the node they are currently visiting. Wanderers move through the net in 
a
stochastic way, following links from one node to another and honing in on 
those
regions of the net containing the most nodes relevant to their home node.
Beyond this, specific capacities are carried by specific node types. The
assemblage of built-in node types is important and is considered analogous 
to
the genetic heritage of specialized human brain modules. Humans are born 
with
some pre-existing ability to process language, to socialize, to perceive and 
to
integrate perceptions cross-modally, to form abstract concepts according to
certain archetypal templates, etc. Webmind similarly enters the world with 
these
pre-existing abilities, albeit in somewhat different form.
First of all, Webmind starts with two low-level "perceptual" modules: one
containing nodes for text processing, another containing nodes for numerical
data processing. Text and numbers are Webmind's perceptual world, and it 
deals
with both in great profusion. Specialized preprocessing is necessary and
appropriate here, just as our eyes and ears preprocess sensory data for our
brains.
In order to intelligently deal with text information, Webmind starts with a
natural language module. The nodes contained herein carry out both semantic 
and
syntactic analysis. The NL system takes in text and parses it, and outputs 
the
parsed text into nodes and links, namely "inheritance links" representing
relations of categorial inclusion (e.g. from "big cat" to "cat") and ternary
links representing general triadic relationships (e.g. joining "big", "cat" 
and
"big cat"; or "to" ,"run" and "to run"). The parser, whose unique and 
innovative
design is described in a separate document, is specifically designed to 
parse
sentences into collections of trilinks.
Note that language processing is not rigidly separated from thought 
processing
here. Webmind does not have a separate "natural language system" as such.
Language processing is carried out by nodes representing words and phrases,
nodes carrying short term memories of recently read material, nodes 
concerned
with categorization, and mobile agents concerned with recognizing similarity 
and
other relationships. To disentangle language from thought is to remove most 
of
what is interesting and important about language.
The first part of Webmind's language understanding is linguistic 
categorization,
which is carried out by category nodes according to the normal dynamics of 
the
Psynet. Webmind's self-organizing categorization is capable of determining 
parts
of speech for words, and of determining semantic categories such as "house",
"action", etc. as well.
Once straightforward categorization has been used to create semantic and
syntactic categories, attractor neural network methods are used to refine
semantic links between words, based on contextual information. Words are 
then
joined by relational agents if they are related closely in the bodies of
information that has been exposed to.
Parsing is carried out by a proprietary, original self-organizing parsing 
method
based on L-system grammar. This system takes a sentence, uses the Psynet to
determine syntactic and semantic categories for the words in the sentence, 
and
then maps the sentence into a collection of nodes and relational agents, 
which
are placed into the PsyContext and then into the Psynet. If the sentence
represented a query, then the collection of nodes and agents corresponding 
to
the sentence is activated rather than merely stored, and a query agent is
created which collects the results of this activation after a period of 
time.
The ingestion of linguistic as well as numerical data is mediated via the 
short
term memory module. The most recent items read in through a given input 
stream
are stored in short-term memory and the various intercombinations of these 
items
are explicitly represented in a temporary way. This system is crucial, among
other things, for disambiguation of linguistic terms.
As important as language understanding is, however, it is not all-powerful.
Relations learned through the natural language system are not intrinsically
"understood" by the system -- they represent purely formal knowledge. 
Webmind's
grounding module, on the other hand, contains nodes that allow the system to
derive knowledge directly from its environment. This module contains
GroundedRelationNodes, which evolve automata that recognize relations among 
data
that Webmind has directly at its disposal (e.g., recognize which DataNodes
correspond to volatile markets, which TextNodes correspond to novels, which 
ones
to long novels, etc.).
Mobile agents of various types create associative and hierarchical
relationships, thus instantiating the elements of a cognitively powerful 
"dual
network" structure. But what guarantees that these two structures, 
heterarchy
and hierarchy, will work effectively together? There are several answers to 
this
question within the Webmind architecture, but one of the keys is the 
explicit
search for meaning. Each portion of the net, searching for its own meaning,
re-expresses itself in terms of the hierarchical and heterarchical 
connections
surrounding it, thus causing the hierarchical and heterarchical connections 
in
its area to function effectively together.
In order for an intelligence to understand the meaning of a word, concept or
other entity, this entity must be "grounded" in the system's own experience.
Consider, for example, the position of Webmind on encountering the word 
"two."
In order to deal with "two" in an intelligent way, the system must somehow
understand that "two" is not only a pattern in texts, it also refers to the 
two
computer users who are currently logged onto it, the two kinds of data file 
that
it reads in, etc. These are all patterns associated with "two" in various
situations. Associating all these patterns with the "two" category node is 
what
fully integrates the "two" node with the rest of the network, guaranteeing 
that
the behavior of the "two" node is in harmony with its surroundings.
Through processing of text alone, part of the fuzzy set that is the meaning 
of
"two" can be acquired: that part which is purely linguistic, purely 
concerned
with the appearance of "two" in linguistic combinations with other words. 
But
the other part of the meaning of "two" can only be obtained by seeing "two"
used, and using "two," in various situations, and observing the patterns in
these situations that are correlated with correct use of "two." In general, 
the
fuzzy set of patterns that is the meaning of a symbol X involves not only
patterns emergent among X and other symbols, but also patterns emergent 
among X
and entities involved in various situations. Recognition of patterns of this
type is called symbol grounding.
Symbol grounding is connected in a fascinating way with introspection. In
Webmind, each agent has the ability to look inside itself and report various
aspects of its current state. Symbol grounding is achieved by special mobile
agents attached to individual nodes representing concepts (e.g. the node for
two"). These agents are evolved by the genetic algorithm; they survey the
introspections of other nodes and combine the results of these 
introspections to
form "groundings" of the node to which they are attached.
The actual mechanics of symbol grounding is carried out in within Webmind in 
a
unique way: groundings for symbols are evolved by special nodes that carry 
out
processes similar to the genetic algorithm. As noted earlier, the relation
between the genetic algorithm and the brain is somewhat uncertain. Gerald
Edelman's "Neural Darwinism" theory argues that the brain can be viewed as a
genetic algorithm with mutation only; however, as I argued in my book The
Evolving Mind, a case can also be made for crossover between "neuronal 
maps." At
very least, we may view the use of the GA in Webmind as a more efficient way 
of
accomplishing, using crossover and mutation, what the brain may accomplish 
with
mutation only.
Both language based understanding and grounded understanding are essential 
to
mind. But with just grounding nodes and natural language nodes, the system 
would
have a divided mind: formal (social) knowledge on the one hand and 
experiential
(individual) knowledge on the other. What bridges the two is reason. The 
purpose
of the Reason module is quite simple: to derive new links from existing 
links.
This is useful not only because some links are derived from the outside, 
e.g.
via the NL system; but also because some purely internal links in the 
system, at
any given time, were derived from information that is currently forgotten.
Webmind contains node-resident and mobile agents whose specific purpose is 
to
carry out this dynamic of building new links from old.
The specific behavior of these agents is in part borrowed from the NARS
(Non-Axiomatic Reasoning System) AI system developed by Dr. W. In his 
previous research work,
has shown mathematically that reasoning agents such as those used in Webmind 
are
capable of carrying out deductive, inductive and abductive (hypothetical
reasoning) in accordance with the requirements of mathematical logic. Unlike 
in
typical AI reasoning systems, however, here reasoning is carried out via
self-organizing processes that are integrated with the remainder of 
Webmind's
intelligence.
The importance of reasoning can be well understood in the context of symbol
grounding: reasoning allows Webmind to extend conclusions from grounded 
domains
to ungrounded domains. For example, if Webmind has learned that "big" means 
a
certain thing in the context of financial data (where it can ground the 
symbol
"big" in reference to its direct experience), then it can use reasoning to
transfer the relations it has derived about "bigness" to domains where it 
has no
direct experience, e.g. to understand a phrase like "a big upheaval in the
Japanese government."
Next, the categorization module allows the system to create new concepts 
from
existing ones via statistical grouping. This leads to concepts of broader 
scope
than are obtained through the grounding system. It does not act on 
perceptual
data directly but rather on higher-level nodes in the system. For instance,
applied to TextNodes representing Web pages, it might arrive at a Yahoo-like
categorization of texts.
There is a text-numerical interrelation module -- Webmind's analogue of the
parts of the brain performing crossmodal correlation between the different
senses, such as sight and hearing. For example, it evolves special nodes 
which
are linked to both DataNodes and collections of TextNodes, and which are 
highly
useful for such applications as using information in newspapers to predict
financial markets. These cross-sensory links allow activation spreading to 
move
between text and data in an intelligent way.
There are also QueryNodes, for mediating interactions with human queriers.
Webmind's query processing is integrated with the rest of its mental 
activity,
just as for a human being, question-answering is not so different from 
purely
internal thought processing. When a query (be it a series of key words, a
paragraph of natural language, or a series of commands requesting particular
data operations) is entered into the system, a node is created for it, the 
query
node sends out mobile agents, and these agents create new relational agents
joining it and other nodes. Activity related to the query node spreads 
through
the Psynet, and after a certain period of time, the nodes with the highest
activity relevant to this particular activation process are collected, and
returned as the answers to the query. The distinction between activity due 
to a
particular query and activity due to general Psynet thought processes or 
other
queries is carried out via an innovative, proprietary technique of "parallel
thought processes," which allows to do one thing the human mind cannot: 
carry
out hundreds or thousands of simultaneous trains of thought, and keep them 
all
straight!
Every Webmind has a "self pod", containing a SelfNode, and various nodes 
linked
to by the SelfNode. No Webmind should have multiple self pods, as this would
lead to dysfunctions similar to multiple personality disorder (MPD). A 
SelfNode
can have subcategories corresponding to "subselves," a feature which also 
has
the potential to lead to MPD, but only if managed improperly. The purpose of 
the
SelfNode is multiple, including at least the following:
    self-control: to enable the system to determine its own resource 
allocation,
    apart from the resource allocation directed by queries and by basic
    activation and wanderer spreading
    episodic memory: to enable the system to compare present situations to 
past
    situations, and draw inferences therefrom, and summon appropriate past 
nodes
    from disk as necessary
    self-description: to enable the system to describe itself to other 
Webminds
    or humans
    social understanding: to enable the system to conceptualize its
    relationships to other Webminds or humans and draw conclusions therefrom
Webmind also contains a social module, with specialized nodes mediating its
interaction with humans and with other Webminds. There is a "friend pod",
containing FriendNodes corresponding to other Webminds. And of course a 
"user
pod" containing UserNodes corresponding to users; and a pod of nodes
corresponding to other agents acting over time, with whom the system does 
not
have direct experience, but whom the system reads about. UserNodes 
representing
human users are an essential part of the user-interaction system, allowing
Webmind to disambiguate query terms based on its knowledge of the particular
user who asked the query. FriendNodes representing other Webminds are 
essential
in the task of figuring out, when the answer to a query is not known, which
other Webmind to direct the query to.
Finally, a given Webmind potentially contains, at any given time, a number 
of
short-term memory nodes, containing information regarding what has been 
recently
experienced by a certain node, and other things judged to be associated to 
these
recent experiences. The short-term memory node's dynamic is called "lineal
awareness," and consists of a loop that takes in new data, processes it in
conjunction with the nodes in the short-term memory, and then places it in 
the
short-term memory, removing some obsolete information. Specialized pods for
dealing with self, social interaction and creativity are dealt with in a 
fairly
similar way to the PsyContext; they contain agents which, rather than being
relevant to recent experience, are relevant to particular areas of
moderately-recent experience, e.g. experience with oneself, experience with
other intelligent entities, and experience with difficult problems.
In short, we have a system whose immediate perceptual reality is numerical 
data
and linguistic communication, which perceives this reality using a freely
combinatorial short-term memory, derives its own concepts based on numerical
data using evolving automata, which learns formal ideas from parsing 
language,
and then extends its ideas from its experiential domain to other domains 
using
reasoning, where the focus of its reasoning at each time is guided by the 
spread
of attention in the direction of meaning, and the meaning of each part of 
the
system is determined by both global dynamic relations and local static
relations.
4. Webmind as Distributed Intelligence
An individual Webmind server running in a single Java process is, 
potentially,
an autonomous Webmind. On a multiprocessor machine with a large amount of 
RAM
(say, 4 Gig), one can achieve quite a powerful Webmind in this way. Given 
the
current state of computer hardware, however, the most cost-effective way to
achieve a very smart Webmind will be to network a number of Webmind servers
together into a single multi-machine Psynet.
To understand the implications of these hardware restrictions, which push 
toward
a network implementation, one must consider them in a broader context. They
overlap with fundamental system-theoretic requirements in an interesting 
way.
Hardware considerations lead us to distinguish three levels of Webmind 
system:
    Single machine
    Network of machines connected by fast cable (intranet)
    Network of machines connected by slow connections (internet)
These levels may shift over time, as the Internet becomes faster and faster.
On the other hand, analogy with biological intelligence leads us to 
distinguish
two levels of systematic intelligence:
    individual intelligence
    society of individual intelligences
Intra-mental interactions may encompass social interactions, but not 
vice-versa:
parts of the same mind have access to each other in ways that separate minds 
do
not.
We use the term an "individual " or simply "a " to refer to an individual
intelligence; and the term "global " or " society" to refer to a collection 
of
intelligences. At the present time, the most natural mapping from
social/individual concepts to computer hardware is
    machine = part of individual
    intranet = individual
    internet = society
For now, therefore, when talking about a single "Webmind" we generally mean 
an
intranet, a small network of machines running with a single self. However, 
this
mapping may not remain ideal forever: eventually it will be possible to have
coherent Webmind individuals running across thousands or millions of 
machines,
around the world.
Social interactions between Webminds are as follows:
    One Webmind can query another Webmind, just as if it were a client.
    One Webmind can send a agent to visit another Webmind, to gather 
information
    (create relational agents involving agents on the other Webmind).
Determining which Webmind is best suited to receive a query or a visiting 
node
to is done using a fairly sophisticated "socialization" system, involving
FriendNodes and the like.
Intra-mental interactions encompass social interactions but add an 
additional
components. For example, nodes within a single Webmind (currently, a single
intranet) will reach out to one another when activated and reinforce the
relational agents intertwining them; nodes separated by a slower Internet
connection will not.
To ensure clarity for the reader with a computer science rather than
philosophical perspective, it is perhaps worthwhile to cast these ideas in 
the
language of distributed computing. Gul Agha has isolated the following five
design decisions as being essential to the characterization of any system
carrying out concurrent computation:
    the nature of the computing elements
    global synchrony versus asynchronous elements
    the mode of interaction between computing elements
    the degree of fairness
    reconfigurability and extensibility
The "nature of computing elements" is that they are nodes, which are a 
special
kind of actor, in Agha's terminology. Processing is asynchronous, not 
globally
synchronous, as is necessary given the desired Internet/intranet 
functionality
of the system. The mode of interaction between the nodes in Webmind is
determined on the mechanical level by the Pods system, that handles node
addressing in a URL-like manner and dispatches communication from nodes to 
their
containing Pods, on to other Pods and the nodes contained therein. On a
functional level the interactions between nodes are diverse and
node-type-dependent, but centrally include neural-net-like activation 
spreading,
genetic-algorithm-like node creation and deletion, and internal 
reorganization
of a node's perceived relationship to other nodes, enabling the node to 
better
represent a category or other relation amongst the nodes around it.
By "fairness" Agha means such things as guaranteed delivery of messages: in 
this
sense Webmind does not place a premium on fairness, except in a 
probabilistic
sense. At any one moment one node may be treated better than another by some
fluke of the system; but, overall, the system acts homeostatically to ensure
approximate fairness, in the sense that nodes that the system assigns a 
greater
importance will get more CPU time. The system is comfortable with lossy
communications in some contexts (e.g. spreading of activations from machine 
to
machine) and not in others (e.g. the return of query answers to human users; 
the
transport of nodes among machines).
Finally, reconfigurability and extensibility are assured by the modular 
system
of node definition -- different functionalities are induced in the overall
system by different node types. This is not a system with a population of 
agents
following a fixed structure and dynamics, designed to be ideal and perfect 
once
and for all. Rather, there are core node types which are seen as essential 
to
realizing the structures and dynamics of the psynet model of mind, but these 
may
be altered based on experience, and may be augmented by a limitless variety 
of
specialized agents to deal with particular situations. In this sense Webmind 
is
almost an "agents operating system" -- any type of agent, within quite broad
constraints, could be integrated into the Psynet. Half of the task of
engineering Webmind is in setting up the "operating system" mediating node
interaction, and the other half is in determining the correct node types to 
give
rise to the emergent structures of mind.


6. The 37 Observations Revisited
In this section, I will run through the 37-Observation version of the psynet
model introduced earlier, and briefly indicate how each of the points is
addressed within the Webmind architecture and its current implementation.

Observation 1. Mind is a set of patterns, where a pattern is defined as "a
representation as something simpler".
Each Webmind node contains tables of links, which represent patterns 
recognized
between that node and other nodes. A link representing a similarity is
representing an approximate repeated pattern; a link representing an 
asymmetric
(inheritance) relation is stating that one thing can be used in the place of
another, i.e. as a pattern in another. Etc.
Observation 2: To be able to have patterns, one must have entities that can
transform entities into numbers (or some other ordered domain, to give
simplicity judgements), and one must have entities that transform entities 
into
other entities (so as to enable representation)
Wanderers, agents which create links, are mapping entities into numbers; 
they
are the valuers in Webmind. Transformation is carried out on several levels, 
but
most centrally by Wanderers which emanate from one node but build links at 
other
nodes.
Observation 3: Mind is intrinsically dynamical. The transformation C(B)=A, 
in
the definition of pattern, implies a "change": C changes B into A
Webmind is obviously a dynamical system; its valuers, Wanderers, are 
dynamical
agents.
Observation 4: Magicians can be combined in a way other than transformation;
they may be combined in space. The result of joining A and B in space may be
denoted A#B.
Special links exist for representing combination; for example the ternary 
link
that joins "big", "cat" and "big cat" to represent the modifier 
relationship; or
the binary "part-of" link that joins the node representing a phrase 
extracted
from a text to the node representing the text itself.
Observation 5: Spatial combination gives rise to the possibility of emergent
pattern: patterns that are there in A#B but not in A or B individually.
Wanderers may explicitly search for emergent pattern, and create emergence
nodes.
Observation 6: The meaning of an entity may be defined as the set of all
patterns associated with that entity -- where a pattern P may be associated 
with
an entity A in several ways: P may be a pattern in A, P may be an emergent
pattern in the combination A # B. or P may close to A in spacetime (P and A 
may
have occurred in about the same place at about the same time)
The meaning of a node is effectively taken as its link tables. Link types 
exist
to denote spatial and temporal relationships.
Observation 7: Pattern and meaning are subjective, in that they depend on 
who is
measuring simplicity, and who is defining the set of permissible 
transformations
Wanderers take the subjective view of the node who sends them out
Observation 8: Meaning is dynamic as well as static. The patterns in an 
entity
include patterns in how that entity acts, and interacts, over time.
The halo of a node encapsulates the dynamic component of meaning
Observation 9: In any given mind, at any given time some magicians are given
more attention than others. Attention means that a magician is allowed to 
carry
out transformations on other magicians.
In Webmind, the nodes with the highest heat, the greatest importance to the
system, are given more CPU time
Observation 10: A mind always possesses some degree of randomness (i.e.,
dynamics which has no patterns that it can detect). Randomness occurs 
wherever
attention occurs.
Wanderers use randomness in their motions. The dynamics of activation 
spreading
will often be chaotic and will hence propagate this randomness widely.
Observation 11: The basic law of mind dynamics is: A magician passes some of 
its
attention to other magicians with whom it shares meaning
This is activation spreading. The thresholding adopted from neural net 
models is
simply a way of making attention spreading efficient.
Observation 12: Sharing of meaning may take many forms. Primally, meaning
sharing may be of three different kinds:
symmetric
asymmetric
emergent.
Symmetric meaning sharing is encapsulated in "similarity links", asymmetric
meaning sharing in "inheritance links" (member-category), emergent meaning
sharing in multiple-target inheritance links.
Observation 13: Intelligence may be defined as the ability to achieve 
complex
goals in complex environments
Webmind's goal of answering human questions meaningfully is complex, and the 
Net
is plainly a complex environment
Observation 14: The complexity of an entity may be defined as the total 
amount
of pattern in that entity, or equivalently, the amount of meaning in the 
entity.
Thus, intelligence is the ability to achieve meaningful goals in meaningful
environments.
Webmind has the ability to determine its own meanings, to understand human
meanings, and to correlate the two.
Observation 15: In order to achieve complex goals in complex environments --
i.e., to be intelligent -- a complex mind is required
Webmind's mind is certainly complex!
Observation 16: A complex mind, implemented in a physical medium, will 
require
continual modification of its internal parameters to assure steady 
intelligent
functioning. This modification must be done intelligently in some cases, and 
so
there must exist certain magicians with a special feedback relation to the
physical medium determining the parameters of mental action.
Webmind has a homeostatic system built in, to regulate its parameters for
optimum performance.
Observation 17: Pattern recognition in a complex environment is best done by 
a
combination of perception, cognition (internal transformation of 
perceptions),
and action
Webmind's perception is reading text and numbers, its cognition is most of 
its
activity, and its action is providing answers, asking questions, and 
possibly
carrying out electronic commercial transactions.
Observation 18: A substantial amount of a mind's attention must often be
allocated to recognizing pattern in its environment, i.e. to this threefold
"perceptual/cognitive/active loop."
This means that pods containing lower-level node types must be allocated
significant CPU
Observation 19: A "magician system" is a collection of magicians that is
self-producing, in the sense that any magician in the system can be produced 
by
the combination of some other magicians in the system. Minds are magician
systems, at least to within a high degree of approximation.
This is an experimental hypothesis that can be carried out on Webmind. If we
remove a node corresponding to an abstract category -- a CategoryNode, a
GroundedRelationNode, etc. -- will the system regenerate something similar 
on
its own? The working hypothesis is YES.
Observation 20: Highly intelligent minds are characterized by hierarchical
structures. The definition of hierarchy in this context is: A relates to 
{B1,
B2,...,Bk} hierarchically if each of the B[i] asymmetrically shares much 
meaning
with A. The process of creating hierarchical structure is called 
"clustering" or
"categorization."
CategoryNodes provide hierarchical structure, as do inheritance links 
inferred
from natural language.
Observation 21: Highly intelligent minds are characterized by heterarchical
structures, large connected networks of symmetrically meaning sharing 
entities
Similarity links provide heterarchy.
Observation 22: In a highly intelligent system, the hierarchical and
heterarchical structures of mind are aligned, so that in many cases, when A
relates to {B[1],...,B[k]} hierarchically, each B[i] relates to a number of
other B[i] symmetrically
This alignment is automatic in Webmind because categorial links are formed 
by
inference from similarity links; and similarity links are then formed 
amongst
category nodes.
Observation 23: Minds are finite, so that if they live long enough, they 
must
forget. They will run into situations where they lose the B involved in a
representation C(A)=B, but retain the pattern A that was recognized.
Webmind contains explicit node and link forgetting routines.
Observation 24: A mind's intelligence will be enhanced if it forgets
strategically, i.e., forgets those items which are the least intense 
patterns
Each node and link has an "importance" value determining its forgettability.
Observation 25: A system which is creating new magicians, and then 
forgetting
magicians based on relative uselessness, is evolving by natural selection. 
This
evolution is the creative force opposing the conservative force of
self-production.
This occurs implicitly in Webmind, and explicitly as well via node types 
that
evolve other node types using the genetic algorithm (this occurs in the
text-numerical cross-correlation module, for the evolution of relations 
between
text and numerical data; and in the grounding module, for the evolution of
automata representing patterns in experiential data).
Observation 26: A pattern A is "grounded" to the extent that the mind 
contains
entities in which A is in fact a pattern
GroundedRelationNodes represent grounded patterns; many SenseNodes and
CategoryNodes derived therefrom represent ungrounded patterns, to do with 
things
like cats and showers that the system has no experience with.
Observation 27: "Reason" is a system of transformations specialized for
producing incompletely grounded patterns from incompletely grounded 
patterns.
ReasonableNode (which CategoryNode extends) contains specific behaviors 
which
look at the links coming out of that node and try to infer new links 
therefrom.
This is Webmind's "reasoning system."
Observation 28: The specialization of a transformation may be defined as the
variety of possible entities that it can act on. The magicians in a mind 
will
have a spectrum of degrees of specialization, frequently with more 
specialized
magicians residing lower in the hierarchy.
This is true within Webmind, there are highly specialized and highly general
node types.
Observation 29: A mind must contain magicians specialized for mining 
emergent
information recently obtained via perception. This is "short term memory." 
It
must be strictly bounded in size to avoid combinatorial explosion; the 
number of
combinations (possible grounds for emergence) of N items being exponential 
in N.

ShortTermMemoryNode accomplishes this, which is necessary among other 
reasons
for semantic disambiguation of linguistic input.
Observation 30: The short-term memory may be used for tasks other than
perceptual processing, wherever concentrated attention on all possible views 
of
a small number of things is required
ReasonableNode uses short-term-memory to determine where to direct the 
course of
reasoning next. A ReasonableNode maintains a "trail" of the things it has 
most
recently reasoned about, which maps naturally into the contents of 
short-term
memory. We call this a "train of thought."
Observation 31: Syntax is a collection of categories, and "syntactic
transformations" mapping sequences of categories into categories. Parsing is 
the
repeated application of syntactic transformations; language production is 
the
reverse process, in which categories are progressively expanded into 
sequences
of categories.
The Webmind NL system represents this specifically in terms of 
transformations
taking pairs of category labels into individual category labels (and vice
versa).
Observation 32: Semantics is a collection of categories, and "semantic
transformations" mapping: categories into categories, category elements into
category elements, transformations into categories, and semantic 
transformations
into semantic transformations.
This is provided automatically by entry of words into Webmind as SenseNodes, 
and
action of the CategoryNode categorization system on SenseNodes.
Observation 33: A key psychological role of syntax is to transfer semantic
knowledge from strongly grounded patterns to weakly grounded or entirely
ungrounded patterns.
The reasoning system automatically uses links from words corresponding to
grounded concepts up to syntactic categories, and from syntactic categories 
down
to words corresponding to ungrounded concepts, in order to build links 
between
grounded and ungrounded concepts.
Observation 34: Language is useful for producing magicians specialized for
social interaction. Syntax in particular is crucial for social interaction,
because another intelligence's observations are in general ungrounded in 
one's
own experience.
Only through the NL system can Webmind get a grip on human concepts that are
outside its experiential domain.
Observation 35: The most intelligent minds have selves, where a "self" S is 
a
pattern which a mind recognizes in the world, with the property that, 
according
to the mind's reasoning, the substitution of S for the mind itself would 
produce
few changes. I.e., the self asymmetrically shares meaning with the entire 
mind.
Webmind contains a Self node, which contains its image of itself in its own 
and
others' eyes.
Observation 36: The "self" of a mind is a poorly grounded pattern in the 
mind's
own past. In order to have a nontrivial self, a mind must possess, not only 
the
capacity for reasoning, but a sophisticated reasoning-based tool (such as
syntax) for transferring knowledge from strongly grounded to poorly grounded
domains.
Reasoning automatically operates on the Self node as it does on all other 
nodes.

Observation 37: The existence of a society of similar minds makes the 
learning
of self vastly easier
Webminds can communicate with one another, using the query language and more
intimately using direct exchanges of nodes.
7. Webmind for Business Intelligence
Having reviewed the general nature of the Webmind architecture, I will now
outline its application to various particular problem areas, at a very 
coarse
level of detail. In this section we will consider several broad application
domains: text information retrieval, data mining and enterprise analysis, 
and
financial decision support.

Textual Information Retrieval
An application area for whichWebmind has been domain-engineered to some 
extent
is information retrieval from document databases. It is important in this
context to distinguish intelligent information retrieval from "search 
engine"
software. The notion of searching for documents based on topic descriptions 
is
highly limiting. What is really interesting is having human minds relate 
with
textual and other digital information in creative and productive ways. A 
final
implementation of Webmind for information retrieval will incorporate search
engine functionality, but will also involve visual exploration of 
information,
and interactive processing whereby the software asks the user questions to
clarify their intentions.
Here we will not go into these user interface matters in detail, but will 
rather
give a cursory description of the internals of Webmind as applied to text
information retrieval. This will serve to indicate how drastically different 
it
is from its current competitors, and to illustrate the algorithmic roots of 
the
different behavior that it displays. Each document being indexed is 
described by
a TextNode, which contains links to various nodes representing the 
meaningful
information in the text. There are also nodes corresponding to categories 
among
documents. At the most abstract level, these categories might correspond to
general topics such as "Business" or "Computing." At a level closer to the
bottom, these might correspond to very particular groupings of pages, 
reflecting
topics such as "artificial intelligence and finance." The process of 
category
formation follows immediately from Webmind's self-organizing dynamics.
Each node on every level, when it is sufficiently stimulated, sends out a 
flock
of intelligent agents, which wander the network of nodes in search of other
nodes displaying similar or otherwise appropriately related patterns. When a
query (be it a series of key words, a simple question, or a paragraph of 
natural
language) is entered into the system, a node called a QueryNode is created 
for
it. Then, the QueryNode is entered into the network much as if it were a new
document to be read, classified and understood. First, using higher-level 
nodes
to guide it, it finds a natural position in the network; then it sends out
intelligent agents and these agents create new links between it and other 
nodes,
other documents. Then activation spreads through the network, somewhat in 
the
manner of a neural net, and after a certain period of time, the nodes with 
the
highest activation are detected, and returned as the answers to the query.
What is fundamentally different about Webmind, as opposed to search engines, 
is
that its document database is not a static entity, which sits there 
passively
waiting for queries to set it into motion. Instead, it is constantly 
refining
its structure -- and its structure is, like the structure of the human
mind/brain, inseparable from its complex dynamics. The categories that it 
forms
are not defined by mere statistical clustering, but by the outcome of
self-organizing, reflexive dynamics. When a query is entered into Webmind, 
it
sets off a chain of reactions that ultimately lead to it finding its natural
places in the network, i.e., finding those groups of documents that pertain 
to
the same topics as it does, as judged by the network's own complex dynamics.
In an ordinary search engine, a collection of pages is judged to have the 
"same
topics" as each other if they have the same words in them. In a statistical
search engine like Excite or SemioText, a collection of pages is judged to 
have
a common topic if they have a mutually similar distribution of words and
phrases. In Webmind, on the other hand, common topicality is not defined
entirely in terms of structure but also in terms of process. A group of 
pages
has a common topic if they have been judged as related by the 
self-organizing
dynamics of the network. The distribution of words and phrases does play a 
role
in this dynamics, as does the distribution of human-created links between
phrases. But these factors only give the initial conditions for the 
dynamics. In
the end, the distribution of links in Webmind is determined as much by the
distribution of links in Webmind as by any specific identifiable factor. A 
query
entered into Webmind is not only a request for information, but a 
perturbation
of the complex dynamics of Webmind's constant study of itself.
It was noted above that Webmind can be used to form a more complete and 
complex
version of the Yahoo category system. However, there are significant
differences, which are conceptually relevant. The Yahoo category structure 
is
very deep, for one thing; Webmind's structures tend to be somewhat flatter.
Also, the Yahoo structure is strictly hierarchical, whereas, Webmind , like 
the
human mind, is capable of operating with a patchwork of overlapping,
complementary categorization schemes. In other words, the hierarchy of nodes 
can
be dissociated, branching off in a number of different directions as it 
rises.
Without Webmind's integrative dynamics, this could lead to confusion, but 
the
dynamics are capable of integrating diverse points of view into a single
coherent system -- in fact, this is what they do best. The different views
within Webmind will not contradict but rather enhance each other. 
Ultimately, in
a situation of extremely favorable computational resources, one can envision
each user creating their own view of the Web, and allowing Webmind to
crystallize a universal categorization system reflecting their individual
perspective. The synthesis of all these individual views, according to 
Webmind
dynamics, would be a kind of collective mentality -- not an average, but a
complex, polyphonic blending of the different subjective world-views of the
different members of the human race.
Webmind can use the same text-based interface as existing search engines.
However, it also lends itself extremely naturally to visual,
dynamic-image-directed search. Instead of returning lists of pages, a 
Webmind
search engine can just as naturally return visual images of those portions 
of
the Webmind network containing relevant pages. Different levels of 
magnification
reveal lower and lower levels of the network, different colors represent
complementary categorization schemes, etc. There are endless possibilities.
Next, regarding Webmind and the network on which it runs, it should be noted
that, just as the general Webmind architecture is perfectly designed for 
network
implementation, so is the Webmind search engine in particular. Today, it is
assumed that a search engine database is a single, monolithic mirror of the 
Web.
Already, however, some firms are breaking with this model; for example,
FunnelWeb has constructed a network of search engines, one for each 
continent.
When the Australian FunnelWeb engine wants information about the European
FunnelWeb engine, it calls it up and asks it. This kind of networked search, 
on
the level of passing queries from one engine to another, is relatively 
primitive
and unexciting. However, Webmind opens up a whole new spectrum of 
possibilities.

Imagine, as the simplest case, that one set up two separate Webmind 
databases
corresponding to .com and .edu domain names. Each one would have its own
integrity, its own hierarchical/associative structure. But there could be 
links
from one database to the other. And if agents from one database were allowed 
to
wander through the other, then the two databases would be linked together on 
the
level of dynamic structure: they would be two entrained, complex dynamical
systems. The closeness of the link between the two systems would be 
determined
by the propensity of the agents to leave their home database.
Imagine, next, that instead of two separate Webmind databases, one had, say, 
ten
million. Imagine, in other words, that every website was its own Webmind
database, and that, in order to do a search, one simply sent out intelligent
agents through the Web. This is the opposite extreme from the
one-search-engine/one-database model, and it is not yet feasible given 
current
bandwidth limitations. But a search engine distributed across hundreds of 
sites
is feasible today, and given the projected rapid growth of network 
bandwidth,
the millions-of-sites scenario may well be feasible in ten years time. Once 
it
is feasible, one will have something remarkable indeed: one will have the
possibility for a World Wide Web which is, structurally and dynamically, a
mirror of the human mind. Instead of a World Wide Web, we will genuinely 
have a
World Wide Brain, a global network intelligence.

Data Mining
The term "data mining" refers to the process of searching through a large 
volume
of data to discover interesting and useful information. Text information
retrieval, as discussed above, may be considered a special case of data 
mining,
in which the data involved is purely textual. Much of human intelligence is
"data mining" -- we are concerned with searching through a vast amount of
perceptual information, regarding the outside world and our own brains as 
well,
and finding the interesting and useful information obscured wherein. In
practice, though, the term "data mining" is generally used to refer to the
elicitation of patterns from numerical data such as financial time series, 
or
data stored in databases such as customer information; and it is these
applications that will be the focus in the following comments. Most 
applications
labeled "data mining" involve prediction of future trends, or categorization 
of
data collected over an interval of time.
Data mining answers business questions that traditionally were too
time-consuming to resolve, finding predictive or categorical information 
that
experts may miss, either because it is too subtle and complicated, or 
because it
lies outside their psychological expectations. It provides answers to 
questions
such as, "Which clients are most likely to respond to my next promotional
mailing, and why?" and "Which direction is the price of the Deutschmark 
likely
to move over the next twenty-four hours?" and "What kind of person is 
hitting my
Website?"
What differentiates Webmind from other data mining packages is, first of 
all,
its ability to ferret out subtle patterns in data. It is able to pick up
regularities that other, less sophisticated algorithms miss. Just as neural
networks and genetic algorithms go beyond linear statistics, Webmind goes 
beyond
neural networks and genetic algorithms. Furthermore, Webmind's sophisticated
Internet implementation allows it to deal with large databases in a highly
efficient way. Unlike other data analysis algorithms, Webmind is 
intrinsically
designed to deal with tremendously large amounts of information, stored 
across
distributed networks. As demonstrated in the financial experiments reported
above, it is also able to analyze textual and numerical information in a 
unified
way -- a fact of the utmost importance given the vast amounts of 
conceptually
interrelated textual and numerical data on the contemporary business 
computer
network.
On a more conceptual level, what differentiates Webmind is the fact that it
approaches a body of data in roughly the same way that the human mind 
approaches
the outside world. It takes the body of data into itself, just as perception
brings the world into the mind, and it "brings the data to life," enabling 
the
data to transform itself and study itself. The core, low-level pattern
recognition routines used by Webmind are not necessarily different from 
those
used by other data mining systems. The difference lies in the way the 
low-level
recognized patterns are combined to form an holistic understanding of the 
entire
body of data. This holistic understanding is embodied in Webmind's link
structure, after a period of evolution, and it results in Webmind being able 
to
recognize extremely subtle and sophisticated patterns in its input data, in 
an
efficient way.
In practical terms, the application of Webmind to data mining is fairly
straightforward. Data is read into Webmind and forms itself into a 
population of
nodes; at which point the Webmind dynamics goes to work, bringing out the
structure implicit in the data. Questions posed to the system are then 
resolved
in terms of the information that Webmind has already discovered, on its own. 
If
one's information is stored on an intranet, then absolutely no 
pre-processing is
required, as Webmind is optimized to process distributed intranet data. If 
one's
information is stored in a conventional database, or some other format, this 
is
not a serious problem either, as the addition of simple additional software
modules allows Webmind's to speak SQL and other relevant database languages.
Webmind carries out the two basic tasks of data mining -- prediction and
categorization -- in a flexible and unified way. Categorization is taken 
care of
by link update methods, as described above. Categories among, say, customers 
of
a business, or stocks in the market, are found in the same way that the 
Webmind
search engine isolates categories among Web pages. As for prediction, this 
is
essentially a matter of recognizing patterns in one data set (representing,
e.g., the past history of a financial market), and then extrapolating these
patterns to another, perhaps unknown data set (e.g. the future of a 
financial
market). The recognition of patterns is the crucial part. Neural networks 
and
genetic algorithms pick up patterns that statistics and decision trees miss; 
and
Webmind, with its self-organizing hierarchical/associative structure and its
ability to read text and internally relate text with data, picks up patterns
that neural networks and genetic algorithms miss.
As a single example of the application of Webmind to commercial data 
analysis,
outside the finance industry, consider the problem of "targeted marketing." 
A
business maintains a list of its customers over a period of time, which 
contains
a fair amount of information about each customer. By studying this list, one 
can
determine the qualities that typify a "good customer" for this business. The
business then obtains a large database of names, e.g. census data, and 
wishes to
determine which people from among this large database are likely to be good
customers.
The specifics of this problem will vary from case to case, depending on the
nature of the database. But the basic principles underlying the application 
of
Webmind to this type of problem are quite general and flexible. Most simply, 
one
may begin by building a Webmind database consisting of the elements of the 
large
database. In the case of census data, each Webmind node might correspond to 
a
single block, or even a single household. One then sets Webmind to work,
building informative links between the nodes. Different links specify 
different
types of relationship. For instance, there will be "geographical links," 
which
link two nodes and are weighted by the distance between the blocks 
represented
by the nodes (two nodes will be linked by a geographical link only if they 
are
relatively close together). There may be links tied to particular 
parameters,
e.g. "age structure" links or "income distribution" links. The different 
types
of links are not chosen haphazardly, but are chosen based on the outcome of
pattern recognition operations within the nodes. Each type of link is
constructed by swarms of agents equipped with an appropriate criterion 
function,
an appropriate measure of similarity between nodes. Higher level nodes are 
then
formed, representing clusters amongst nodes -- natural groupings of 
"mutually
similar" nodes. Finally, the elements of the smaller database are then 
entered
into this network like queries into a text search oriented network. They 
find
their most natural place in the network, extend links, and are activated. 
The
most active nodes are then the "similar" nodes within the larger database.
This differs from typical statistical or AI approaches to market analysis in
many essential respects. In terms of performance, the biggest difference is
Webmind's ability to carry out highly targeted market analysis. Existing
algorithms are effective at weeding out terrible customers from the large
database, but only mediocre as isolating the best customers. Webmind, on the
other hand, has a deeper understanding of the structure of the large 
database,
because it begins by allowing the structure of the large database to 
discover
itself. This deeper understanding allows it to target precisely those 
customers
who are most similar to the business's best customers -- just as, in the 
Webmind
search engine, a query finds those Web pages that are the most similar to 
it.
Similar approaches have been worked out for diverse other data mining
applications -- no business intelligence problem lies outside its domain. 
The
architecture encompasses the most effective aspects of previous approaches 
--
such as statistical clustering, neural networks, and genetic algorithms -- 
into
an overarching framework. It integrates the two main objectives of data 
mining,
categorization (carried out by higher-level nodes) and prediction (carried 
out
by pattern-recognition methods within nodes), in a seamless way. As with Web
search, it is not a panacea: there is still a certain amount of work 
required to
tailor the prediction and link update methods to fit the nature of 
particular
application domains. And searches for extremely subtle patterns in huge data
warehouses can still take time, even with Webmind's sophisticated
internetworking capabilities. But Webmind represents a significant step 
beyond
what is out there; and it has the flexibility and reflexive adaptability to 
take
advantage of improvements in Internet technology and data mining algorithms,
without straying from its basic framework, which is, after all, nothing else 
but
the structure of mind.

Knowledge Management and Knowledge Creation
"Information retrieval" and "data mining" are concepts that were created to
describe the behavior of software much more limited in its behavioral 
repertoire
than Webmind. Webmind, with its ability to understand context and to ground
concepts in its own experience, also has the ability to do much more than 
any
mere information-extraction tool. It is able to serve as a true knowledge
management and knowledge creation system, integrating information from the
document archives and databases of interest to a business, and allowing
exploration of this information in all its facets, each piece considered in 
the
context of the whole.
If Webmind is installed on a company's intranet, then real-time queries
regarding relationships between textual, numerical and other data to do with 
the
enterprise may be posed by any employee with computer access at any time. 
The
result is that Webmind's intelligence is integrated with the social 
intelligence
of the organization, and the individual intelligence of the employees. Each 
time
an employee accesses a document, he may ask Webmind for related documents, 
and
Webmind will carry out this task with an understanding of the role of that
employee in the company, that employee's particular needs and interests, 
etc. It
will be able to make its own creative suggestions, based on its autonomous
thought processes.
Furthermore, the social dynamics of the different Webmind servers residing 
in
different parts of the company's intranet (governed by Webmind's highly
innovative, proprietary server-server communication methods) will grow to
reflect the social dynamics of the individuals using those parts of the
intranet. Each Webmind server will respond most effectively and rapidly to
queries involving information which it stores locally; but the information 
that
a certain server stores locally may change over time, depending on user 
needs
and internal Psynet dynamics. Thus, while providing easy access by all users 
to
all information at all times, Webmind will nevertheless nudge the 
information at
the readiest disposal of individual humans and divisions in certain 
directions,
based on its inferences and its own emergent understanding. Webmind will do 
more
than just provide an understanding of structures and processes; it will be a
participant in processes, in the formation of emergent human and 
informational
structures.
And, as various Webmind units in various organizations exchange 
non-proprietary
information, in the interest of increased mutual intelligence, Webmind will 
be a
participant in the formation of human and informational structures on the 
global
scale. This is an exciting new vision of artificial intelligence, in the
business context and beyond -- not AI as something separate from humanity,
providing us with answers to our questions, but AI as something interacting
symbiotically with humanity, participating in our communications, goals, and
social structures and processes.
Enterprise Analysis, Workflow and Business Process Reengineering
Using its extremely flexible structures and dynamics, Webmind can easily be
trained to perform a variety of interesting "enterprise analysis" functions. 
For
example, Webmind will be able to:
    produce pictures of the amount of information passing between different
    people and groups
    make causal inferences: "Because Bob finishes his balance sheets on 
time,
    Joe is usually able to give meaningful information to the production
    department."
    isolate the most important concepts being passed between a particular 
pair
    of people or groups
    track the evolution of concepts through written communications
    discover the tightly connected clusters of people or groups, among which
    large amounts of information commonly passes
    discover the boundaries in the network of people or groups, which
    information, or particular concepts, have difficulty passing through
    produce a map of the clusters of people in an organization, separated by
    boundaries, and the amount and type of linkage between the different
    clusters
Collectively, these functions amount to making a map of an organization as a
self-organizing system, and identifying the emergent structures in the
organization. These functions will play an important role in enabling
organizations to become more effective learning systems.
Achieving these enterprise capabilities will not involve adding any new core
structures to Webmind, but only creating new types of query nodes, and 
tuning
the categorization, visualization and concept-evolution processes to work on 
the
type and volume of data found in enterprise applications.
Financial Decision Support
As I write these words, a Webmind is running in the  office,
producing daily judgement as to which of four major markets are going to go 
up
or down the next day. Each day, it downloads the news from the Net, extracts 
the
important concepts from the news, and uses these, combined with various 
purely
numerical strategies, to make its predictions. Currently the system is doing
prediction on a daily scale, not on a second-by-second scale, but this is 
purely
for simplicity's sake; by the time you read these words the principles
underlying our analyses of daily financial data will have been extended
throughout the financial domain.
The system reads the news, not in a musing kind of way, but with a 
particular
financial data set in mind, say the Dow Jones. It comes up with concepts 
with
the following property: The relevance of the concept to the news on a given 
day
tells you something about what the market's going to do on the next day (or, 
2
weeks later, etc.). The financial meaning of the text is thus boiled down to 
a
collection of numbers – one for each concept extracted, representing the
relevance of that concept to the text on a certain day. The numbers
corresponding to the concepts can be computed anew every day, and used for
financial analysis purposes just like numbers coming from any other source.
These numbers, representing the relevancies of text-derived concepts to the 
news
at a particular time, are what we call text indicators.
The extraction of text indicators is the crux of Webmind's financial
intelligence. It relies on Webmind's ability to intelligently judge 
relevance,
which draws on all of Webmind's abilities at reasoning, language 
understanding,
conceptualization, and so forth. But text indicator extraction is not the 
end of
the story. In addition to the extraction of financially relevant concepts 
from
news, there is an additional process of learning optimal trading models for
particular financial markets. Just knowing the news concepts that tend to
correlate with a certain financial market (say, knowing that trouble in 
foreign
countries tends to drive the Dow) doesn't tell you enough to make accurate
predictions. You have to get at the nonlinear interrelations between news
concepts and numerical patterns in the data. This has to be done differently 
for
the Dow, for IBM stock, for the Yen, for 30 year bonds, and so forth. For 
each
market, Webmind derives a trading model that embodies the best way of
incorporating text based information into decisions about that particular
market.
Webmind's trading models are what computer scientists call "boolean 
automata,"
simple logical decision rules, just as Webmind uses for making any kind of
decision. They're basically the same kind of rules that are used, within the
natural language system, to decide which sense of "Java" is intended in a
sentence; or that are used more generally for learning of abstract concepts 
and
categories. As an initial test of the system, we asked Webmind to learn 
optimal
trading models for five major markets, first with and then without 
news-derived
information. When it included the information derived from news archives, 
the
performance increased tremendously, often by a factor of 3 or more. 
Financial
analysis has traditionally been based on fundamental analysis (looking at
qualitative, economic and political factors), technical analysis
(number-crunching; extrapolation based on past numerical values of behavior
within the market), or, more recently, psychological analysis (exploiting
commonalities in behavior of the traders who actually take market actions). 
The
approach here contains elements of all of these, since it seeks out and 
exploits
both causal and correlative relationships from many sources within each of 
the
three perspectives. Beyond this, it clearly represents a new, fourth 
approach,
in its use of text resources for integrated decision-making. By linking vast
reservoirs of numerical- and text-based data sets in a sophisticated
state-of-the-art intelligent learning system, capacities for financial
decision-making are greatly enhanced.
These results are also of fairly obvious importance outside the financial 
arena,
because the dynamics underlying Webmind 's performance on these sample tasks 
are
the same ones underlying its more general performance. In addition to
demonstrating that Webmind can produce highly profitable trading systems 
using
textual information, these tests also show something more general: that 
Webmind
can meaningfully extract concepts from large amounts of text data, relate 
these
concepts to trends in numerical data, and use the relationships it has 
inferred
to provide highly significant actionable information. This capability is of
tremendous value beyond the world of finance, in areas such as risk 
assessment,
demand planning, supply chain management, enterprise management, document
retrieval and analysis, primary market analysis, medical diagnosis and 
research
-- the list could be continued almost endlessly.
Wide implementation of this kind of technology will have a globally dramatic
effect. The financial component of the World Wide Brain will become vastly 
more
intelligent than it is right now. I would hate to see the World Wide Brain
develop with an overly financial bias – the finance industry, for all its
breadth, does after all represent a fairly narrowly biased view of the human
race and all the information it has to offer. But the financial markets are
already a globally integrated, perceiving and acting self-organizing system, 
and
so they are a natural place for Internet intelligence to start.
The precise mechanisms underlying Webmind's text-based market prediction 
are,
obviously, proprietary (and are also patent-pending). But the basic 
character of
the process is nothing but Webmind intelligence, applied to one particular
domain. Recall the basics of Webmind architecture. Webmind, internally, 
consists
of a collection of software objects called "nodes," each of which contains 
links
to other nodes, representing inter-node relationships. Some nodes contain 
raw
data such as text or numerical time series; others are more abstract and 
consist
entirely of links to other nodes. The design is superficially similar to a
neural network, but is fundamentally different in several respects. A 
Webmind
node is more like a "neuronal module" in the brain than it is like a single
neuron. In Webmind , unlike in a neural network, link construction is 
carried
out by a variety of intelligent software agents, and nodes and links are
frequently created and destroyed as part of the learning process. Webmind 's
internal intelligent agents use a variety of techniques such as genetic
algorithms and statistical language processing. Different types of nodes are
used for representing different types of data. The node types most directly
relevant to simple financial applications are:
    DataNode, which refers to a numerical data sets, e.g. financial time 
series
    TextNode, which refers to a text document, e.g. a market news report for 
a
    particular day
    TimeSet, which refers to a series of time-indexed nodes, e.g. a series 
of
    market news reports over several years
    ConceptNode, which refers to a concept either extracted from text or 
learned
    by Webmind
    TradingRuleNode, a particular kind of finite automaton representing a
    financial trading rule
The process of learning textual indicators to aid in market prediction, 
trading
and analysis is a natural outgrowth of Webmind 's intelligent 
self-organizing
dynamics. Consider, for simplicity, the case where the market being trading 
is
the Dow Jones Industrial Average, and the text being used is market news 
which
is issued on a daily basis. In this case the central data structures are a
DataNode pointing to the DJIA, and a TimeSet pointing to the series of 
market
news articles. The text indicator learning process is carried out by an 
agent
associated with these two nodes. The goal of this agent is to isolate a
ConceptNode, representing one of Webmind 's internal ideas, with the 
property
that the relevance of the ConceptNode to the TimeSet, on a given day, is a
useful indicator for trading the data in the DJIA DataNode on that day. 
Having
carried out this learning process, it then uses the ConceptNode it has 
isolated
to create a new "text indicator" DataNode, each entry of which indicates the
relevance of the ConceptNode to the market news TimeSet on a certain day. 
This
text indicator DataNode can then be supplied to the user as, quite simply, a
series of numbers. Each day, on reading the news, Webmind can supply a new 
value
for the text indicator -- just as, on seeing the DJIA itself, it can supply 
new
values for numerical indicators such as moving averages and the like.
The text indicator obtained by this process of internal conceptualization 
can be
used for a variety of different purposes. It can be used by humans to form 
their
own intuitive judgements, or incorporated in any computational trading 
framework
that is sufficiently flexible to incorporate arbitrary numerical indicators.
Most of our work so far, however, has involved incorporating text indicators
into Webmind 's own intelligent trading processes, which are ideally suited 
for
incorporating text indicator information. A TradingRuleNode, in Webmind ,
encapsulates a trading rule, which is a logical combination of information
deriving from a numerical predictor and a number of indicators, some 
numerical
and some textual. These details, however, have more to do with computational
finance than Internet intelligence, so I will omit them here. The point is 
that
Webmind's intelligent text analysis feeds into a Webmind subsystem that 
carries
out integrative numerical-textual financial prediction, according to methods
that are original but build on standard practices in financial analysis.
Finally, what exactly are these "concepts" that Webmind derives, by reading 
the
news with financial data in mind? Within Webmind, they are represented as
abstract data structures. They are associated with various words and 
documents
and numerical data sets, but they are not truly described in any way except 
as
abstract Webmind-internal data structures. Just as, while we can try to 
describe
our intuitions about the world in words, we can never quite capture them. A 
good
boxer, or financial trader, or mathematician can try to explain the 
intuitive
ideas that guide her work, but the articulation never quite matches the 
reality.
For the period 1996-1997, for instance, the most prominent concepts driving 
the
Dow and the Yen had to do with trouble in Asian economies, and Asian banks
particularly. "Asian bank trouble" as a focus of the news was highly 
predictive
of downturns in the Dow. One useful concept had to do with the relation 
between
the British government and the Asian financial situation. Another had to do 
with
European unification. But the important thing to remember is that the 
concept
that Webmind found to be related to the market wasn't just "European
unification" itself – it was, rather, a certain slant on European 
unification, a
certain sense it got from the news, that had to do with European 
unification.
These results are fascinating, and are sure to get more so as Webmind 
becomes
more sophisticated, as all the different modules of the code become 
integrated
with each other (as of now, natural language and reasoning are not used
"full-strength" in the finance applications; and the self and social aspects 
of
Webmind are not used at all). Ultimately, they will only be a small part of
Webmind's utility. However, they will never become trivial: whatever else it 
may
be, the future of the Internet is going to be commercial, which means that 
the
World Wide Brain will have to be financially intelligent!







                              Chapter 8

                        Semiotics and Autonomy

1. Introduction
We have seen that the Internet contains ample resources to support the 
learning
of language, in its syntactic and many of its semantic aspects. But language 
is
not a closed system. Discussion of the automatic understanding of language 
leads
naturally to the question: how can an emergent Internet intelligence 
understand
what concepts "mean"? A Net-embedded AI system will have access to vast 
amounts
of text, for use in language understanding, but recognizing patterns of 
words'
appearances in text can only bring you so far. To achieve true intelligence, 
an
AI system needs to go beyond this and relate concepts to its own experience.
This is where the vast amounts of nonlinguistic data accessible to an 
Internet
AI system becomes essential. Grounding concepts in experience requires a 
vast
sensory universe, coupling reduncancy with endless variety; and the Internet
provides precisely this.
To take a very simple example, if an intelligent information retrieval 
system is
asked by a user to "find me some big novels about love and politics" it 
should
not merely look for occurences of the phrase "big novels" in the texts in 
its
memory, it should look for texts containing novels that are actually big. 
I.e.,
it should "ground" the symbol big by associating it with the "physical"
parameter of file size.
What I call "symbol grounding" -- expression of linguistic entities as
combinations of nonlinguistic elements -- is typically classified under the
label "semiotics." And semiotics ties in naturally with the concept of 
autonomy,
a relationship that was drawn to my attention by Cliff Joslyn (currently of 
Los
Alamos National Laboratories). A system that grounds its own symbols makes 
its
own meanings, and is hence autonomous in linguistic space. This kind of 
autonomy
in linguistic space seems to be a prerequisite for intelligent autonomy in
physical space.
2. Meaning
In order to rigorously explore the question of semiotics in regard to AI, 
one
requires a precise definition of "meaning." The definition to be used here 
is
the one supplied in Chapter 5 of Chaotic Logic (Goertzel, 1994), where it is
suggested that the meaning of an entity is simply the set of all patterns
related to its occurence. For instance, the meaning of the concept "cat" is 
the
set of all patterns, the occurence of which is somehow related to the 
occurence
of a cat. Examples would be: the appearance of a dead bird, a litter box, a
kitten, a barking dog, a strip dancer in a pussycat outfit, a cartoon cat on 
TV,
a tiger, a tail,....
It is clear that some things are related to "cat" more strongly than others.
Thus the meaning of "cat" is not an ordinary set but a fuzzy set. A meaning 
is a
fuzzy set of patterns. In this view, the meaning of even a simple entity is 
a
very complex construct. In fact, as is shown in (Goertzel, 1994), meaning is 
in
general uncomputable in the sense of Godel's Theorem. But this does not mean
that we cannot approximate meanings, and work with these approximations just 
as
we do other collections of patterns.
This approach to meaning is very easily situated, in the sense of situation
semantics (Barwise and Perry, 1981). The meaning of an entity in a given
situation is the set of all patterns in that situation which are related to 
that
entity. The meaning of W in situation S will be called the S-meaning of W. 
The
degree to which a certain pattern belongs to the S-meaning of W depends on 
two
things: how intensely the pattern is present in S, and how related the 
pattern
is to W.
These ideas are not difficult to formalize. Consider a pattern in S as a 
process
p whose result rp is similar to S. Let s be a "simplicity function" mapping 
the
union of the space of mental processes and the space of situations into the
nonnegative real numbers; and let d be a metric on the space of mental
processes, scaled so that d(rp,S)/s(S) = 1/c represents an unacceptably 
large
degree of similarity. Then one reasonable definition of the intensity with 
which
p is a pattern in S is given by the formula

IN[p|S] = [1 - c d(rp,S)/s(S)] [s(S) - s(p)] / s(S)

The term [s(e)-s(p)]/s(e) = 1-s(p)/s(e) gauges the amount of simplification 
or
"compression" provided by using p to represent e.
Next, let MW,S(p) denote the degree to which p is an element of the 
S-meaning of
W (S is a situation, W is an object). Then one might, for instance, set

MW,S(p) = IN[p;S] * corr[W,p]

where corr[W,p] denotes the statistical correlation between W and q, gauged 
by
the standard "correlation coefficient." The correlation must be taken over 
some
past history of situations that are similar in type to S; and it may 
possibly be
weighted to give preference to situation which are more strongly similar to 
S.
The determination of similarity between situations, of course, is a function 
of
the mind in which the meanings exist. Webmind contains built-in 
methodologies
for gauging similarities between situations and other entities. But the 
point
for now is that the meaning of an entity is a fuzzy set of patterns: the 
degree
to which a process belongs to the meaning of X is determined by the 
intensity
with which that process is a pattern in situations correlated with the
appearance of X. This mathematical definition implies that meaning is a 
fuzzy
set of patterns.
Finally, the relationship between meaning and intelligence is worthy of 
brief
comment. Intelligence, as I have argued in (Goertzel, 1993) and reviewed 
above,
should be considered as the capability to achieve complex goals in complex
environments. A mathematical definition of "complexity" may be founded on 
the
basis of the definition of "pattern" given above, the gist of which is that 
an
entity is complex if it has a large number of different patterns in it. An
environment is complex if it has many patterns in it; and a goal is complex 
if,
when considered as a function mapping situations to degrees of goal 
achievement,
its graph has a large number of different patterns in it. The collection of
patterns in an environment or goal, which makes that environment or goal
complex, is a fuzzy set of patterns. Since intelligence has to do with
processing and enacting complexity, it requires manipulating fuzzy sets of
patterns. Meanings of specific elements in the world may thus be viewed as
subsets of the overall field of complexity in which intelligence operates.
Isolating meanings is a way of breaking down complexity into parts -- i.e., 
it
is an example of problem solving by division and reunification, the most 
basic
problem-solving strategy of all.

3. Symbol Grounding
A consequence of this approach to meaning is that, in order for an 
intelligence
to understand the meaning of a word, concept or other entity, this entity 
must
be "grounded" in the system's own experience. Consider, for example, the
position of Webmind or another similar intelligent text processing system on
encountering the word "two." In order to deal with "two" in an intelligent 
way,
the system must somehow understand that "two" is not only a pattern in 
texts, it
also refers to the two computer users who are currently logged onto it, the 
two
kinds of data file that it reads in, etc.
Through processing of text alone, part of the fuzzy set that is the meaning 
of
"two" can be acquired: that part which is purely linguistic, purely 
concerned
with the appearance of "two" in linguistic combinations with other words. 
But
the other part of the meaning of "two" can only be obtained by seeing "two"
used, and using "two," in various situations, and observing the patterns in
these situations that are correlated with correct use of "two." In general, 
the
fuzzy set of patterns that is the meaning of a symbol X involves not only
patterns emergent among X and other symbols, but also patterns emergent 
among X
and entities involved in physical situations. Recognition of patterns of 
this
type is called symbol grounding (Harnad, 1990), and is a crucial aspect of
artificial intelligence. Symbol grounding is, in essence, the difference 
between
semiotics and formal language.
One approach to the task of symbol grounding would be to encode thousands or
millions of specific rules, explaining which items in which situations 
correlate
with which symbols. However, this kind of approach is obviously doomed to 
fail,
because even if one could accomplish such a tremendous feat of rule encoding 
for
all known situations, one would still face the problem of enabling one's 
system
to deal with unknown situations. Fortunately, in a self-organizing 
intelligent
system, it is possible to take a different approach: to supply a small 
number of
symbol-reality mapping rules and allow the remainder to emerge spontaneously
from the network's own dynamics. The general principle here is that the
subjective reality of an intelligent system is necessarily self-organizing: 
only
a subjective reality which "understands its own structure" will be able to 
adapt
and grow, and adaptation and growth are prerequisites of intelligence.
In order to make this concept more concrete, consider an AI system composed 
of
multiple agents, each of which has the ability to look inside itself and 
report
various aspects of its current state. Textual statements and other data, 
once
read into the system, are decoded into collections of agents. A "symbol
grounding" is represented by a special kind of mobile agent, a symbol 
grounding
agent. A symbol grounding agent is associated with another agent, which it
"tracks." Its goal is to mimic the behavior of the agent which it tracks, 
but
without looking into that agent itself, only by looking into other agents. 
In
this way it represents the agent that it tracks as a relationship among 
other
agents. Given the collection of information-bearing agents as a fixed
information base, the symbol grounding agent attempts to become a pattern in 
the
node that it tracks. A given agent may be tracked by a number of symbol
grounding agents, each of which is operative in certain situations, and some 
of
which are more effective than others. The idea of meaning as a fuzzy set of
patterns is thus embodied concretely: the meaning of an agent is enhanced by 
a
fuzzy set of meaning-gathering agents. Situation specificity of
meaning-contributing patterns is reflected as situation specificity of 
agents.
These are intelligent agents, as recognizing patterns in the occurence of
linguistic symbols or other mental entities is not an easy task; but the 
machine
learning problems involved here are not insurmountable, and are in fact 
similar
to those encountered in other areas of AI, e.g. financial prediction.
The implementation of symbol grounding agents involves many detailed and
difficult issues not touched on here, but the conceptual aspects of the 
design
should be clear. The point is that the population of symbol grounding agents
associated with a piece of information expresses that piece of information,
explicitly, as a fuzzy set of patterns. The usefulness of the symbol 
grounding
agents for generating meaning depends on two factors: the intelligence of 
the
symbol grounding agents, and the coherence and informativeness of agent 
system
used as an information base.
5. Philosophy of Symbol Grounding
The examples considered above -- "long", "two", etc. -- involve texts and
numbers, things that Net-based AI systems have fairly direct knowledge of. 
They
don't involve, for example, horses, or rooms, things that a such a system 
could
know of only via indirect reference. This is not coincidence of course: an 
AI
system will be able to ground only those symbols that refer to something it 
has
direct experience with. This is a fundamental fact of mind and there is no
getting around it!
Does this mean that Internet AI systems won't be able to reason about 
horses? By
no means. If such a system is told "the horse is big," it will be able to
interpret this statement by using its intuitive understanding of the word 
"big"
-- which is gained via its grounding of big in the context of texts, data, 
and
so forth. But, it won't be able to ground the concept of a horse for itself,
unless it is given some perceptual basis for this grounding. If it were 
trained
to recognize horses in .gif files, this would allow it to develop a visual
grounding for the concept of "horse", which would still be very partial 
compared
to the grounding that most humans have.
Humans experience the difference between grounded and ungrounded symbols 
very
vividly when visiting, for the first time, a city that they had previously 
only
read about. I understood a lot about Prague, before I had ever visited 
there,
based on reading novels set in Prague, tourist books about Prague, etc. I
reasoned about Prague by reference to my large battery of relevant
already-grounded symbols: city, building, street, etc. etc. I constructed a
fuzzy, shifting default image of the ungrounded symbol "Prague" as a fuzzy
combination of already grounded symbols -- bringing other ungrounded symbols
such as "Moscow" to bear on it too, no doubt. But when I visited there, the
symbol "Prague" was rapidly grounded, which brought my thinking and feeling
about Prague to a level it could never have achieved without the symbol
grounding. One can make great discoveries about a city without ever visiting 
it
-- pick up historical and cultural trends, etc. -- but there are some things 
one
will necessarily miss from not really "knowing" it. (A more potent example 
than
cities would be sex. Compare the understanding of sex possessed by someone 
who
has never had sex, but has looked at a lot of pictures and read a lot of 
books
about it, with someone who has had the experience.) Similarly, there are 
some
things Webmind will always miss about horses until it is given multisensory
organs to interact with horses and ground the symbol "horse" for itself. But 
it
will likely pick up a lot more than us humans about computer networks, 
numerical
data, Internet traffic, etc., because in these areas, it has more grounding 
than
us, rather than less.
From a philosophical perspective, the basis of the notion of grounding is as
follows. Symbols emerged to represent recognized patterns. But a symbol can
never capture all the nuances of a pattern. It can guide the mind toward the
recognition of the same patterns that others have recognized before; or it 
can
serve as a shallow, limited "proxy" for the pattern, when the mind has not 
yet
recognized the real thing. But there is no substitute for recognizing the
pattern oneself. And no mind can operate entirely on the basis of proxies. 
The
percentage of ungrounded proxies must be below a certain "critical level" or
reason will necessarily fail. This is perhaps the main reason why all 
attempts
at AI so far have failed.
What we do when we don't have a real-life grounding is to use a formal
definition. For instance, what is a horse? Well, a horse, of course, is

A hoofed quadruped of the genus Equus; especially, the domestic horse (E.
caballus), which was domesticated in Egypt and Asia at a very early period. 
It
has six broad molars, on each side of each jaw, with six incisors, and two
canine teeth, both above and below. The mares usually have the canine teeth
rudimentary or wanting. The horse differs from the true asses, in having a 
long,
flowing mane, and the tail bushy to the base. Unlike the asses it has
callosities, or chestnuts, on all its legs. The horse excels in strength, 
speed,
docility, courage, and nobleness of character, and is used for drawing,
carrying, bearing a rider, and like purposes.

A high-quality NL system will be able to parse this definition and use it to
construct a number of different things. It will build is-A links from 
"horse" to
"quadruped" and "Equus", and part-of links to "hoof", "molar", "incisor",
"tooth", etc. It will construct trilinks to (4,horse), (long, mane) and
(flowing, mane). It will construct "quality" links from "horse" to 
"strength",
"speed", "docility", etc. This is useful information, but not the same as
grounding. These definition-embodying links simply "outsource" questions 
about
horses to other concepts: to figure out if X is a horse, you check if it has
teeth, hooves, four legs, etc. Then, to check if X has "teeth", "legs", 
etc.,
the definitions of these other terms must be appealed to. This is a 
reasonable
way of proceeding, but yet, it will never give true understanding. At some
point, the system must be able to move outside the network of formal, 
relational
links into a deeper kind of understanding: an understanding grounded in
experience. This understanding grounded in experience is implicit, not 
explicit.
It may be represented as a neural network, as a high-dimensional polynomial 
or
logical combination of various simple terms, or in any sufficiently flexible
format, but it is unlikely to be transparently comprehensible in the way 
that an
explicit dictionary-like definition or associated collection of relational 
links
is. Implicit symbol groundings cannot be derived from formal definitions, 
and
cannot be communicated in language in any reasonably compact way, but can 
only
induced or evolved based on experience.
Whether a grounding is constructed in terms of rules or in terms of neural
networks or something else is not important, the important thing is the
complexity of the framework used to represent the grounding. Groundings must 
be
constructed from a framework which is flexible and adaptive -- if they are 
to be
built from rules, they must be built from large numbers of rules according 
to a
non-restrictive rule-combining algebra. Every single grounding need not be
complex -- e.g. the grounding of "big" given above is not complex -- but the
framework for generating groundings must be capable of fluidly and compactly
representing complexity.
One of the important qualities of symbol groundings is this: they are 
compact
processes which enable a mind to make category membership decisions 
regarding
categories that are of complex shape in "idea space." The most common
statistical clustering algorithms are optimized to recognize categories of
spherical shape. Simple collections of logical rules, on the other hand, 
often
recognize categories of polygonal shape. Any generalized system of
categorization is going to have a bias toward some particular shape of 
category.
Generalized categorization systems have an important purpose: they help a 
mind
to form a rough map of the conceptual landscape, to get an initial picture 
of
what things belong with what other things. But in order to arrive at an
effective working understanding of the world, it is necessary to go beyond 
any
generalized system and create specific processes for determining category
membership, and hence refining the mind's categorization system. This 
process is
too computationally expensive to use as the sole means of categorization, 
but is
important as a method of deepening the understanding initially arrived at by
simpler, statistical means.
But, important as it is, flexibly-shaped categorization is not the defining
characteristic of a symbol grounding. The word "grounding" is significant 
here
-- it suggests attaining firmness by appealing to a lower level. In a human
context, this lower level is the body and physical reality. For an Internet 
AI
system, this lower level is its own body and physical reality: the lengths 
of
files and strings and tables, etc., stored in its various component agents,
which correspond directly to the sizes of chunks of RAM absorbed by 
different
parts of its internal network. The lower level is not more absolutely 
meaningful
in a philosophical sense, but it is more broadly meaningful in a social 
sense:
it is a realm that is common to a host of different minds. Symbol grounding, 
in
essence, may be defined as the identification of concepts with compact 
processes
constructed as potentially complex combinations of socially-real phenomena, 
and
demarcating complex shapes in idea space.
5. Inductive, Deductive and Abductive Groundings
Crystallizing these philosophical observations into a framework more 
suitable to
guide engineering, one may think about three types of grounding:

    Inductive: observing that a symbol is correlated with certain phenomena, 
and
    inducing a machine that predicts which phenomena the symbol is 
correlated
    with
    Deductive: observing that A is a subset of B, and hence that the 
grounding
    of B is an initial guess at the grounding of A. (e.g. a=bird, b=animal)
    Abductive: creating a virtual grounding for something never directly
    perceived (e.g. a quasar). Here we look for transformations from things 
we
    have perceived, e.g. quasar = f(star), quasar = g(comet), etc.
Abductive grounding is the most complex of the three, and the one that gives 
the
perceived world most of its texture. To carry out abductive grounding, a 
system
must evolve transformations from unknown to known enities by a GA or similar
method. Then, it must compose these transformation with the groundings of 
their
arguments, e.g.
        possible-grounding-for-quasar = f(grounding-for-star)
This is how we create "mental movies" grounding things we have never had
experience with.
The real subtlety of symbol grounding lies in the fact, pointed out to me 
with
particular force by Julian Troyanov, that even groundings that seem to be
inductive also have an abductive, imaginative component. This abductive
component of even familiar concepts is is what gives them their flexibility 
and
living nature. An Internet-based AI system will have to deal abductively 
with
many common elements of human physical reality; just as we have to deal
abductively with things like packet flow over the Net, which it will be able 
to
deal with inductively ... it will be able to "feel" the packets flowing and
hence ground the concept of packet flow inductively.
With these concepts in hand, we may revisit the issue of grounding numbers. 
It
seems that the straightforward evolution of inductive symbol groundings will
suffice to ground numbers from specific examples: e.g. to cause the "2" node 
to
recognize entities that are sets of 2, if it is trained by being given 
examples
of things that are sets of 2 and things that are not sets of 2. This is not
difficult if it is assumed that the agents comprising an Internet AI system 
will
be introspective enough to numerically measure their own sizes.
But, what about a number like "three thousand, four hundred and six"? 
Clearly,
the system can't be given explicit training on every number it will ever 
see. Of
course, a specific formula for grounding numbers could be hard-coded into 
the
system, but that isn't the point: the point is, how could it learn the 
correct
way to ground numbers, as human children do when learning mathematics? This 
is
abductive grounding, and it must be done in such a way as to provide not 
only
precision but flexibility. For instance, what does "three zinkum, four 
hundred
and six" ground to -- merely a vague impression of a big number? Yes, but 
every
one knows that it's one more than "three zinkum, four hundred and five." 
Unless
of course one discovers that a zinkum is a kind of infinity, in which case
"three zinkum, four hundred and six" and "three zinkum, four hundred and 
five"
are equivalent! This is all a matter of adjusting the transformation used in
constructing the abductive symbol grounding.
6. Semiotics/Autonomy Feedback
In closing, let us return to the relation between semiotics and autonomy. 
This
relationship is, in an intelligent system, necessarily bidirectional: on the 
one
hand, a system uses its shared and individual semiotics to maintain 
autonomy;
and on the other hand, a system requires its autonomy in order to gather and
maintain the information that is the basis for the formation of its 
semiotics.
The feedback dynamics between semiotics and autonomy is very subtle and can 
be
expected to manifest phenomena of chaos, complexity and emergence.
As autonomous intelligences, we experience this feedback from the inside; 
and as
observers of other humans, we experience it from a combined 
internal/external
perspective, due to our empathy for other humans and our embeddedness in the
shared social semiotic dynamic of human culture and society. In engineering
artificial intelligences, however, we get a unique opportunity to observe 
the
semiotics/autonomy feedback from the outside, a perspective that yields many 
new
insights.
In particular, based on the above discussion, we may enounce a few general
points about meaning and mind:
    An intelligent system's strategies for meaning extraction are 
necessarily
    inseparable from the overall intelligent dynamics with which its dynamic
    knowledge base is endowed.
    The meaning of a word, phrase, sentence, text or concept, to an 
intelligent
    system, is encapsulated in the relationships between this entity and 
other
    entities which it contains in its dynamic knowledge base.
    In order to truly understand the meaning of text or speech, an 
intelligent
    system must be able to map the meanings recognized in linguistic items 
with
    aspects of own data structures and of non-textual data presented to it.
In an autonomous intelligent system, such as Webmind or the human brain, 
each
entity is a symbol for a tremendously large fuzzy set of other entities in 
the
system. Meaning thus emerges as a self-organizing web of fuzzy patterns. The
task of building a subjective reality is carried out by this web; and the
subjective reality constructed by this web is a key contributor to the web's
growth.
Because the evolution of symbol groundings is derived by usage of the nexus 
of
relationships in a self-organizing memory, semiotics depends on autonomy,
inasmuch as autonomy is what allows an integral, useful nexus of 
relationships
to evolve. On the other hand, a self-organizing AI system's autonomy relies
equally much on its semiotics, in the sense that a system's understanding of
itself hinges on its understanding of its relationship with humans and other
computers, and its understanding of this relationship would be feeble indeed
without symbol groundings of basic relational concepts. Essentially, we 
conclude
that, in a self-organizing Internet-embedded AI system,
    Semiotics supports autonomy via the social definition of self
    Autonomy supports semiotics by supplying a coherent body of raw material 
for
    the evolution of symbol groundings
It is tempting to conclude that this characterization of the 
semiotics/autonomy
feedback relation holds generally, beyond the context of the Net; but this 
is
not the place to enter into such issues.

                              Chapter 9

                 Toward an Agent Interaction Protocol

1. Linguistics versus Pragmatics
Intelligent software agents on the Internet will, like humans, need to
communicate via language. What language this will be, however, is not 
entirely
clear. It might be English, or a computer-ized dialect of English; it might 
be a
form of logic or mathematics; or it might be an evolved language totally
incomprehensible to humans. Standardizing the discourse of intelligent 
agents is
not something we should be trying to do at this stage. Standardizing the
pragmatics of intelligent agent discourse, on the other hand, is an 
important
task, and one that may well be a prerequisite for the emergence of Internet
intelligence.
Humans communicate via a variety of different languages, all of which exist
within the shared pragmatic/semantic space of physical discourse. If I were
placed on a desert island with a person who spoke only Quechua, we would
gradually learn each others' languages, by reference to physical objects 
(point
at a tree and say "tree", etc.). On the other hand, if I were strapped 
against
the wall in a dark room, and a person who spoke only Quechua were strapped
against the wall on the other side of the room, we might, after months or 
years,
begin to make progress -- again, by virtue of the shared physical reality. I
would whisper "quiet" and yell "loud", etc. It would be difficult, but 
possible:
this is the power of pre-linguistic, pragmatic reality, the common stream of
perceptual data shared by various linguistic and nonlinguistic beings.
The notion of an Agent Communication Language, as presented e.g. in Fikes 
and
Farquhar (1998), embodies this distinction between linguistics and 
pragmatics.
Agent communication is broken down into two layers: an outer "agent 
interaction
protocol" (AIP) layer, and an inner "content communication" layer. The AIP 
layer
plays the role of pragmatics, giving a basic framework in which linguistic
communications are framed. The content communication layer plays the role of
language. The dissociation of these two layers is important, because right 
now,
different commercial and academic AI developers radically disagree on what 
kinds
of languages should be used to communicate between intelligent agents. 
However,
agreement on an outer layer of pragmatics, an agent interaction protocol, is
more closely within reach.
Taking a slightly different tack, the importance of agent interaction 
protocol
may be seen by analogy to the birth of the Internet. The emergence of the 
Net as
we know it required two main things, engineering-wise:
    ARPANET, the first packet-switching network
    TCP/IP, the protocol allowing machines on differently architected 
networks
    to communicate with each other via "gateways"
The same may be true of the World Wide Brain. Proprietary AI engines acting 
over
intranets and extranets will be the equivalent of ARPANET, but to achieve a
truly global brain we will need an equivalent of TCP/IP to allow the 
emergent
intelligent network to grow beyond the bounds of the customers of a 
particular
firm.
2. An Agent Interaction Protocol
The most promising agent interaction protocol proposed to date is a very 
simple
message-handling protocol called KQML (Knowledge Query and Manipulation
Language; see Fikes and Farquhar, 1998). In this section I will describe 
KQML,
and suggest some revisions that would simplify KQML and render it more 
general,
less tied to the formal-logical reasoning systems that gave rise to it.
KQML contains the following basic "performatives":
	tell
	deny
	untell
	evaluate
	reply
	ask-if
Each of these takes a certain number of parameters, drawn from:
	:content <expression>
	:language <word>
	:ontology <word>
	:in-reply-to <expression>
	:force <word>
	:sender <word>
	:receiver <word>
This is a fairly general and powerful framework, but it is not quite 
perfect,
and its imperfections reveal its roots in logic-based AI. KQML was designed 
to
work with KIF, a knowledge representation format based on first-order logic,
with the addition of some object-oriented primitives (class, subclass of, 
slot,
slot-value-type, etc.). Though in principle KQML can work separately from 
KIF,
it contains a number of features that are really not meaningful in the 
context
of AI agents based primarily on self-organization rather than logical 
deduction.

The presupposition of KIF is that knowledge is represented in terms of 
logical
formulas, and that "Sets of expressions necessarily bottom out in 
expressions
containing undefined (primitive) symbols" (Fikes and Farquhar, 1998). 
However,
this is not a universal presupposition of intelligence: in a later chapter 
we
will see that, in some AI systems, sets of expressions bottom out in highly
complex combinations of nonsymbolic, nonlinguistic data. Just as, in the 
human
mind, "chair" bottoms out in a collection of perceptual data and
perceptual-action schema. This means that chair is a "primitive" symbol, but 
it
does not mean that nothing can be communicated about "chair." Rather, it 
means
that communication about chairs may possibly take place on a pragmatic,
pre-linguistic level as well as on the linguistic level. KIF does not permit 
for
pre-linguistic discourse; it operates only on the level of logic. However, 
if
intelligent agents are to communicate in a truly intelligent way, they must 
be
allowed to interact prelinguistically and prelogically as well as on the 
level
of logical formulas. Imagine trying to teach a child to speak, think and
interact using only logical formulas!
For instance, how, in the KQML language, would one agent point to a picture 
and
say "pretty"? There is a primitive for "tell" but no primitive for "show" -- 
a
glaring omission given that showing is perhaps the most essential aspect of
pragmatic, pre-linguistic communication.
Also, the inclusion of a parameter for "ontology" is fairly artificial. When 
two
humans communicate, they do not specify the ontology to which they are
referring: ontology is implicit for humans, and it is implicit for many 
types of
AI systems too. A better name for the "ontology" parameter would be 
"context." A
formal-logic-based ontology would be one example of a context, a document 
would
be another.
To explore this issue a little more deeply, let us observe what is really 
meant
by an "ontology" by members of the research group involved with designing 
this
protocol. Fikes and Farquhar (1998), advocating the OKBC ontology protocol,
which fits in naturally with KIF, state that
"We consider ontologies to be domain theories that specify a domain-specific
vocabulary of entities, classes, properties, predicates and functions, and a 
set
of relationships that necessarily hold among these vocabulary items"
Clearly, this is a highly restrictive notion of an ontology, and not all
intelligent agents are going to operate in such terms. In fact, any system 
that
does operate according to such a restrictive notion of ontology is 
guaranteed
not to be very intelligent. A useful theory of a given domain of knowledge 
is
not going to consist of a set of necessary logical relationships, but a 
fluid,
self-organizing network of possible relationships. Recognizing this, Koller 
and
Pfeffer (1998) are seeking to generalize the logic-based notion of ontology 
to
deal with uncertainty; but this is really only a tiny step in the direction 
of
psychological plausibility.
The inclusion of "force" as a parameter in the agent interaction protocol is
probably worthwhile. The forcefulness of a statement is important, witness 
the
importance of voice volume in human speech. Of course, there are many other
parameters besides forcefulness that are important in establishing the 
pragmatic
context for communication. Even in the case of myself and the 
Quechua-speaker
strapped at opposite ends of a dark room, we have a lot more than force of
expression to go on! But force has a very pragmatic use in the case of
questions: it indicates how badly the questioner wants an answer. 
Furthermore,
it is something that has meaning in a variety of different AI architectures:
neural network based systems, for example, will lead to different 
conclusions
with different "activations", activation essentially being a version of 
"force"
in the sense of AIP.
"Evaluate" might also seem to be a good thing to have -- one agent will 
often
want another to evaluate something for it, although "evaluation" may not 
always
be meant in the sense of logical evaluation. Something sent along for 
evaluation
may be evaluated in many different ways, the result not necessarily being a
truth value or logical expression, but possibly a whole body of information.
But, evaluation in this sense is really a form of asking. The distinction
between "evaluate" and "ask-if" is only meaningful in the context of 
logic-based
AI systems.
Finally, "untell" is clearly not a generally necessary part of an AIP. What 
does
it mean for a person to "untell" someone something, as opposed to denying 
that
thing? This is a distinction that is meaningful only in the context of 
specific
logical formalisms, not in general intelligent discourse.
On the other hand, since we have statements (tell) and questions (ask) in 
the
protocol, it seems odd not to have a way to command an agent to do things.
Commanding is clearly a much more natural behavior than untelling! Telling 
is
informative; commanding is what the central machine in a local network does 
to
the other machines in the local network. Computers do a lot of commanding in
networks now, so an AIP should support this.
Finally, and more controversially, I would argue for one more additional
performative in the Agent Interaction Protocol: mate. We have seen above the
potential that exists in the "Internet as evolving ecology" concept. The 
best
way to encourage this along would be to have genetic-algorithm-type 
crossover as
a standard operation of agent interaction. Of course, each agent is 
responsible
for determining what it can mate with and what the result will be like.
Radically different agents will not be able to mate.
In conclusion, my preferred AIP would have the following performatives:
	show
	tell
	ask
	reply
	command
	mate
The omission of a few performatives from the KQML specification is not
important, as it was not the intention of the designers of KQML that every 
agent
use every performative. However, the addition of the "show" and "command"
performatives is crucial. Without these, the protocol is really not 
complete.
Show, tell, ask, reply and command are very basic features of communication, 
and
the omission of showing and commanding in favor of untelling and denying 
really
makes KQML inadequate as a general agent interaction protocol. Furthermore, 
the
addition of "mate" is an explicit support of emergent intelligence that is 
not
at all present in KQML.
Basically, KQML only allows agents to talk together and to evaluate logical
expressions together. But this is not enough. We interact with each other in
many ways beyond conversation, and artificial Net agents should have the 
same
ability: at minimum, they should be allowed to show each other things on the
Net, and to mate with each other. Showing, talking and mating are the basic
operations out of which a society of intelligent agents may be formed, and 
may
form the substrate of an emergent electronic mind.
Next, along these same lines the parameters that I think should be generally
used are:
	:content < expression>
	:data <array of pairs (data type, data item)>
	:language <word>
	:context <array of URL's>
	:in-reply-to <expression>
	:force <float>
	:chance <float>
	:sender <word>
	:receiver <word>
Note the changes here. "Ontology" is replaced by "context", and "content" is
augmented by "data", which is not merely an expression but may be any 
collection
of data (e.g. any MIME type). "Force" is supplemented by "chance" -- the 
amount
of randomness that is desired in the response. In a mating operation 
"chance"
gives the mutation rate; in a query processing context, "chance" gives the
adventurousness of the reply. These changes are small but important. Showing
does not work unless content can be drawn from beyond the realm of logical
expressions. And, context must be expressible as a list of multiple relevant
items, not merely as a pointer to a formal, explicit "ontology." The URL's 
that
"context" points to may be ontologies, or anything else; and the recipient 
may
do with them what they wish. In general, it seems reasonable to allow 
specific
agent types to use custom parameters within messages. As with tags in SGML, 
if
an agent receives a parameter type that it does not know how to process, it 
may
simply ignore it.
KQML with performatives and parameters thus modified I propose to call SAIP, 
for
"Simple Agent Interaction Protocol."
In the following section I will explore some interesting uses of SAIP, but 
first
it should be observed that SAIP also has down-to-earth applications. It is
general enough that it could be used as a wrapper around very mundane 
"agent"
interactions such as Web browsing. What a Web browser sends to a Web server 
is,
on the AIP level, a "show" message. In a distributed search system, on the 
other
hand, a search engine might send an "ask" command to a Web server, asking 
for
information about particular topics. An e-mail is a "tell" message; an 
e-mail
with attachments is a "tell" message bundled with a "show" message. The
inclusion of the "show" command makes apparent the continuity between the
sophisticated agent interactions of the future Internet, and the simple but
highly useful interactions between the programs that serve on the current 
Net as
"agents" for human desires.
3. Four Levels of Communication
To make these ideas more concrete, I will now explore some specific types of
interactions that agents on the Net will need to make in the near future. 
This
list of projected interactions is systematized according to the four levels 
of
being as introduced above. It is not intended too be an exhaustive list, 
merely
to indicate some very-near-future possibilities.
Note that the different levels of being do not correlate in any 
straightforward
way with the different performatives of the SAIP. Interactions on any level 
of
being may involve ask, tell, show or reply messages, in various 
combinations.
Only the mate message is special: it lives on the level of wild being.
STATIC BEING
Agents interacting on the level of static being means agents simply 
exchanging
information. The action of a Web browser or an e-mail client falls into this
category. Information is being pushed from here to there. Standard search
queries also fall into this category -- they involve a search for what is 
there,
what is present. Finally, "push technology" in its simplest form is of this 
type
as well. One agent is pushing information that it judges relevant -- pushing 
it
to an agent representing a human, such as an e-mail program or an "active
desktop," pushing it to an AI system, or wherever.
Note that the static-ness of these interactions is dependent on the breadth 
with
which one defines the "system", the overall context of consideration. These
interactions are static when considered purely on the Internet level. An 
e-mail
is a static object transferred from one Internet agent to another, and when 
it
is received, it just sits there, rather than entering into any particular
dynamic. On the other hand, if one defines the overall system to include 
human
computer users as well as computer programs -- justifiably, as humans are 
the
most flexible and powerful Internet agents existing today -- then things 
work
out quite differently. The human who reads the e-mail may enter the e-mail 
into
his mind, where it may well trigger off all sorts of complex dynamics. For
instance, it may cause him to send more e-mails, in which case it was a 
"static
being" interaction with the e-mail client, but a process being interaction 
with
the human/e-mail client joint system.
PROCESS BEING
Agents that trigger processes are quite common today. The Web browser, when 
it
activates a CGI script, is triggering a server process. Standard 
client-server
data mining software resides on this level as well. If one is simply 
accessing
database data, then one is in a sense triggering a process -- a lookup 
process
-- but the essence of the interaction is static, accessing what is there. 
But if
one is asking for a clustering to be performed, or for the future value of a
time series to be predicted, then one is asking for the on-the-fly creation 
of
new information: it is really the process one is paying for, as much as the
information itself. As online databases become larger and larger, 
information
retrieval necessarily becomes more a matter of process being than one of 
static
being. The abundance of information means that the premium is not on mere
possession of information, but on filtering of information, an intelligent
process.
HYPER BEING
Hyper being agent interaction goes beyond triggering of processes -- now we 
are
talking about the seeding of creative self-organization. If the querying of 
a
large database has a focus on the "mining" process, the querying of an
intelligent, self-organizing database has a focus on the complex,
information-building processes seeded by the query. Asking an intelligent
information system for knowledge about "AI technology as related to bovine
menstruation" doesn't merely trigger off a search process; it triggers off a
learning process within the AI system, which may potentially change the AI
system's dynamic memory for good.
Hyper being queries may be specific requests for the creation of new
self-organizing systems -- as in the case of a manager who wants to build a
simulation of his organization within an Internet AI system, knowing that 
the
running of this simulation will structure the way the AI system gathers
information from other systems, etc. Or it may be implicit, as in the case 
of an
ordinary query to a self-organizing database. The point regardless is that,
rather than just a process, one wants to trigger off a complex process 
resulting
in a system of interlocked informational causes and effects to create a new
subsystem of the target system.
And of course, hyper being interactions need not be queries. Telling another
system something can seed off self-organizations on its own. Preachers, 
teachers
and marketeers are always in search of that magic formulation that will seed 
off
self-organization in the minds of their listeners. But, one thing they lack 
is
the ability to directly implant ideas into their audience's brain. Instead 
of
pushing data to another agent, one Internet may push a small agent, intended 
to
interact with the recipient agent and encourage it in certain directions. 
The
"data" parameter in the Simple Agent Interaction protocol supports this: 
part of
the data passed along may be an agent itself, e.g. in the form of a Java 
applet.

WILD BEING
What "wild being" agent interactions means is nothing short of mind meld. 
This
is not so much a query as an undirected open-ness: "Allow my brain to merge 
with
yours and let's see what happens." The "show" command is essential here. One
agent may show another portions of its own mind, and in return obtain 
portions
of another agent's mind. The self-organization processes on either end may
result in an agent-transcending feedback that binds the two agents into a
combined emergent intelligent system.
Among appropriately sophisticated agents, wild being interaction could be
triggered by a simple "ask" message: Do you want to mind meld? If the 
"reply"
were "Go for it", then a repeated exchange of data would ensue. The data
exchanged would not be static information, but rather networks of complex
processes. Each agent, having taken some of the other's complex processes 
into
itself, would absorb the other's tacit ways of thinking as well as its data. 
Of
course, this kind of interaction can only occur among agents that have
sufficiently similar internal workings to be able to exchange interior 
segments.

And mating, of course, falls into this same category. Mating is a process by
which parts of two or more agents come together to form new agents. This is 
an
extreme version of mind meld: instead of one agent taking in a small part of
another agent's mind, an agent is formed of roughly equal parts from two 
others.
Again, mating implies that the two agents mating are of the same "species", 
that
they are close enough that combinatory operations are defined. But this
specificity requirement is no different from what is found in the biological
world: different species coexist, but do not interbreed.
If we count humans into the system, wild being interaction happens from time 
to
time on the Net already. E-mail listservers and discussion groups, and 
Usenet
groups, at their best, are a kind of collective thinking. Most often they 
are
collective stupidity, but on rare occasions they can achieve a true emergent
intelligence; I have experienced this myself on several occasions.
Purely electronically, wild being agent interactions do not seem to be 
happening
yet, because the agents out there are too unsophisticated. But, there are no
fundamental obstacles in the way of this happening. All that is needed is 
the
adoption of a sufficiently flexible agent interaction protocol, and the 
creation
of various special-purpose-driven agents using the protocol to communicate. 
This
is something that will arise in the next decade purely due to market 
pressures,
rather than out of any desire on the part of agent engineers to create 
emergent
global intelligence. Just as static being querying gradually gives way to
process being querying, as databases become large enough that filtering is 
more
important than raw information; so process being interactions will gradually
give way to hyper being interactions, which will gradually give way to wild
being interactions. All four levels will remain important, of course, just 
as
they do in biological systems. But wild being interaction will occur in what 
now
seem the most unlikely places.
For example, an e-mail/web-browser type client will in the future be a
mini-mind, a self-organizing network of user-relevant information. The 
process
of querying a database will be a process of melding the mind of the user's
client with the mind of the database, and then submitting simple process or
static queries into the resultant emergent information store. And the 
relevance
of this data will be assured by the accuracy with which the client's 
mini-mind
reflects the user's mind, an accuracy obtained through routine daily 
wild-being
interaction between the user and his/her computer.