Registered: Feb 2004
More on Dialectical CAs
As I said in one of the posts above, I thought that the actual perceptual schemes performed by these dialectical CAs may be off-topic. Yet, I observe with amusement that my ID on this NKS forum is exactly 256!... which kind of impels me to be true to this peculiar coincidence and therefore elaborate some more...
I have had some inquiries, all in private. Following excerpts are
typical: (also, some of Jon Awbrey's thoughts in his
"Variety and Regulation" notes may be relevant here).
Joel, do you have pictures of the simulations that you can send me to print and bring to the meeting. Including 2-D and 1-D for comparison.
I'll fax you a collection of figures from a technical report to ONR/SDIO of 1987.
CAUTION: These 2-D patterns are not always easy to interpret, especially at first sight . By analogy, consider bubble chamber or cloud chamber photography -- you don't expect an untrained eye to draw the necessary deductions/implications just like that... it takes some doing...
Some of the interpretations have taken me years to figure out.... of course, now that I can point my finger at what's going in there, it is a different story... except that my finger is not long enough to reach from St. Louis to Tel-Aviv...
Another issue we didnt have time to talk about is stability. How
stable are the converged patterns to added noise to the rules?
These processes are incredibly stable, or robust. There is a manifest, unusual feature that I called in the patent "autonomic error-correction". See columns 25 to 27 in the patent:
Sorry... I am told that the isss server is down (09-01-04)
for re-organization. Patent document can be found
at http://www.uspto.gov Search for Patent No. 4,286,330
I show there three types, with examples, but there are also more. These are:
(1) next-step correction
(2) re-generated cycle correction
(3) correctness without correction
The underlying reason for these is that each of these dialectical
processes is destined to converge on a "basin of attraction" (or limit cycle) and it is generally hard to escape that 'attraction'. Now, even when escaping, a new basin is formed, which provides for a form of 'creativity'. Let me explain.
In 1-D (also in 2-D, but it is much more complex to explain here), under a vast multiplicity of recursive tetracoding (RT), the universe of all possible strings (regardless of (finite) length, and regardless of 'alphabets' they are written in -- dubbed "StringLand"; do you remember FlatLand?) is divided up into a relatively small number of equivalence classes.
Exposing a system comprised of a multiplicity of RTs to a variety of input strings 'trains' the system by establishing various cycles, each of which becoming a 'recognizer' for all external strings belonging to the same class.
Occasional noise or malfunction may alter an established cycle into
a cycle never 'experienced' before via input of external strings! So, the system now has a recognizer for strings (or events) it has never experienced before!
It is like forming a hypothesis that something 'out there' ought to exist, and be ready to confirm the hypothesis upon encountering that thing in future inputs. (Almost something like pre-cognition.) Hypothesis-formation (also called 'abduction' by C. S. Peirce) is one of the highest forms of creativity.
Without going much further, I think you may get a feel for 'creativity' here, but also for notions of creative 'random mutation' that may be inherent in these kinds of systems. A lot more in this regard to be discussed later.
(now I am on a safer ground of going from philosophy to specific questions). Have you tried to look at elements that have internal states and if so did you look at a case that the states are like quantum states namely each of the elements is in a superposition of its internal state and the rules on the number
(changing the number of the elements ) affects the expectations of the internal states of each . Or alternatively it can be viewed as virtual particles.
The question above is not sufficiently focused for me to answer directly. For one, I can't see clearly the 'internal states' and 'quantum states' point in that question. But I can say a number of things that may be related to the gist of your
First, bear in mind that I never intended to do modeling in physics, or particle physics. I had my hands full trying to figure out visual perception, awareness, cognition, and things of that sort; then build artifacts that are as close as possible to the actual *natural* thing. (Called it 'True Machine Intelligence' (TMI) in a majpr NASA project -- got some traditional AI people very upset, but was able to prevail for the duration of the project...)
The connection to the baryon octet, hence to particle physics, has been unexpected, unintentional, and purely serendipitous. I don't fancy myself a particle physicist, so, what I'll say below is based on a naive point of view, as far as physics is concerned; but these descriptions are tied directly to what can be objectively observed by others in the behavior of the processes under discussion.
We have definite things like "vacuum" with stuff popping in and out spontaneously. We have "sea quarks" and "valence quarks". We have "interferences", "oscillations", wave-like propagations. We have convoluted enfoldments that may amount to "superpositions". We also have virtual particles. And so on and on... sometimes it looks to me as though people working in theoretical particle physics are reporting on their own perceptual processes, as opposed to something 'out there'. For the similarity is so very incredible...
Now, something to note in regard to enfoldment and "hidden" domains. When you'll look at those 2-D simulations you'd need to know that these are essentially generalizations of 1-D. Effectively, thru each pixel you have 4 recursive tetracoding (RT) working simultaneously, crisscrossing vertically, horizontally, and the two diagonal. At the same time, each such RT captures all other pixels along a straight line. So, the whole image is globally tied up, while the rules are local distinction rules. Also, instead of only 4 states in 1-D, we now have 256 states in 2-D!!! These is enormous for any CA -- just humongous! But the stuff I'll fax you is implemented via neural nets. So, all features in the graphics are composed of individual neurons firing/not-firing. There is NO WAY IN THE WORLD that you (or any one else, for that matter) would be able to look at patterns of firing neurons (in the simulations, or in real nervous tissue) and read-off any connections to elementary particles -- just impossible!
Things are too complex, too enfolded, too dynamic, etc., etc.
Yet -- here one can freeze the action at will; tell exactly which
particular neuron fires when and why; where/what the interference pattern is; where are the sea quarks and where are the valence quarks; etc. etc., and bring out the goings on in terms of patterns of elementary particles dynamics. Also, fantomark strings (thru their 'streaks') can be analogous to 'guiding waves' or carriers of unexpected field effects.
Last edited by jdi on 09-08-2004 at 05:06 PM
Report this post to a moderator | IP: Logged