[Swarm-Modelling] LSMs

[Mike Beedle]

http://www.eet.com/showArticle.jhtml?articleID=18902790

After reflecting the better part of the day on 
this article, I have some comments and conclusions about 
the research exposed in it.

1. Self-Organizing equivalents of "regular" Computing

   If LSMs, which rely on self-organizing networks 
   are equivalent to UTM (Universal Turing Machine),
   this could be one of the _most_ significant discoveries
   of Computer Science and/or Cognitive Science in History.

This basically means we have a "self-organizing neural model" 
that is computationally equivalent to a "regular computer".

We already know about some equivalences and relationships
among UTMs, FSMs (finite state machines), programming languages 
i.e. prescriptive/descriptive (systems of descriptions like rules), 
calculi, and evaluators (LISP is equivalent to a UTM) ...  
but we never had a "self-organizing" equivalent 
(as far as I know).


2. Swarm Intelligence and Emergence

Until now, we can model and understand things that 
"emerge" through simple "rules of interaction".
Models, simulations and mathematical treatments abound.

Yet, everyone is mesmerized by the concept of 
"emergent swarm intelligence", but until now, this concept,
as far as I know, has not been explained in terms of 
computational models or even in terms of "equivalent models".
(Please experts out there... correct me if I am 
ignoring important research "out there".)

Might the LSM model also explain some aspects 
of the so-called "Swarm intelligence" and to some degree 
the conditions for "punctuated equilibrium" or 
"stable emergent states" i.e. why self-organized 
"swarms of things" that obey simple rules are seemingly 
"intelligent wholes" like anthills, beehives, or brains?


3. LSM and Theories of Life and/or Intelligence

It is eerie to think of these "controlled positive 
feedback loops", and not think about the relationship(s) they
may have with the "autocatalytic chains" described by 
Kauffman's "life model" -- the similarity and/or consistency 
is significant, it is the same model, basically:  things 
"connected" by closing loops in "controlled feedback loops".... 
ah, but Kauffman makes no relationship with "intelligence" or
"learning" in his revolutionary exposition... i.e. Life 
may be "smarter than we think" ;-)  Could this have any 
relationships with the "coding" and "non-coding" sections
of our DNA?.... where we store our knowledge.  Is the 
DNA more like a configuration management system then, 
rather than a "single version" system?  (Non-coding parts
can be thought to be remnants of genetic material
that previously "made sense" in the generation of
proteins that previously participated in a "older versions"
of self-organized auto-catalytic chains -- no longer 
in effect.)


4. Us, Robots, and "many" Minds around

If LSMs are truly a good "mind model", imagine this:  robots 
and computers that are "computationally equivalent" to us.

The concept of "downloading your mind" to a computer or a
robot, or "transferring your mind" across the network --
possibly to another continent (or planet?), now it is more 
possible, because before, we _did_ not have even a 
computationally equivalent model .... :-)  LSM might not be
the last word, but we seem to be getting closer in this
iteration.  (For you "cognitive experts" in the field out 
there, how significant is the LSM in you opinion?  Is it 
really that "revolutionary"?)

[WARNING: "mind field". Of course, being able to "transfer
minds" or "execute minds" in truly "out of body" experiences, 
begs for answers to some interesting questions: what 
happens to You when you transfer your mind out of your body and 
"execute it" in a computationally equivalent "robot" or computer.
Combining this with some of the new genetic engineering 
staff, will it be possible to "some day" download your 
mind to a "genetically engineered" or "accelerated Morphogenized 
body".... now that we have a "closer model" of "the mind"
we are transferring (after all they can be simply seen 
as "untrained neural nets")  At a practical level, you could 
make a "backup of your mind", just in case you 
"loose it" ;-) .... or you could go to sleep to give "rest to 
your mitochondrial-driven body, while a "backup of your mind" 
thinks all night or learns something new, and then you can 
transfer the thoughts and/or the learned material in the morning. 
Lastly, if we figure how the mind stores knowledge in this 
computational model, will we be able to buy or download 
"Open Source" "knowledge libraries" and upload them to our 
brains?]


5. Intelligence Field Theory?

Also, the "liquid surface" metaphor in LSMs is given to describe 
the interactions of "close neural neighbors" and the overall 
state of the machine is given or controlled by the 
superposition of eigenstates over the network.... so, basically
LSMs are described by a "field theory" over a neural net
that supports "controlled feedback loops".
Might it be then possible to use "standard" field theory
patterns and techniques for its understanding i.e. things like 
generative functionals, gauge fixes, conformal mapping, 
Wilson OPE (operator product expansions), Fadeev-Popov 
techniques, Polyakov-style integrals, ... multi-dimensional 
fields? (scalar, tensor, spinor, twistor, etc.)   

This seems possible ... and it would be certainly be convenient...  
a model of "intelligence" that closely resembles Nature in 
its most fundamental way.  (From a personal perspective, this 
may expand the utility of some programs I am making for 
computing "general Lagrangian interactions" :-)
http://www.physicsissoftware.com/physicspatterns.htm


6. Quantum Intelligence?

Lastly, there is a strong parallel of LSMs with the 
newly explored, or rather newly accepted, "emergent space-time" 
physical theories that describe space-time as a collection 
of "interacting self-organized entities" i.e. Penrose, Finkelstein, 
Lloyd, Zizzi, Rovelli, Smolin, Zuzze, etc.; and how particles, 
(superposition of interacting fields on it) interact over 
this discrete self-organizing "matrix".  (Again, from  a personal 
perspective, this may expand the utility of some programs 
I am using for computing "general interacting self-organizing 
models of space-time".  See the above link. 

[NOTE: A model of "controlled positive feedback" is also 
possible with the standard "(an)harmonic oscillator" 
field theory, but the similarity, or equivalence if 
proven that way, is more compelling by choosing the 
entities and their interactions to describe space-time.]

What this means is that it might be possible for "anything" 
with this "controlled feedback state" to be intelligent 
-- or at least as intelligent as our "brains".

If, for example, "elementary" space-time and field/wave/particle 
interactions are used to induce this state, we might end up 
with something, that for lack of a better term, we can call 
"Quantum Intelligence"...  (This goes way beyond 
"Quantum Computing" which simply makes Nature a 
"network of massively parallel computers" working on 
entangled states.)

So, for example, any system made of atoms, electrons, light,
quarks, gluons, gravitons... that enters this "controlled
feedback loop" field-state could be as "intelligent" as our 
brain!


7. Apologia

Sorry I am dumping all this version of "the LSM
edition of Saturday Night Live" on you.....  but imo
this "discovery" is significant.  If you have a couple
of minutes to reflect on this, I would very much like to 
hear your thoughts.  I am specially interested in
evaluating the true significance of LSMs, and in finding
current parallel research directions.

- Mike

http://www.mikebeedle.com

"In life as in dance: Grace glides on blistered feet."

  -- Alice Abrams