["If there is intelligence everywhere..."] - A New Kind of Science: The NKS ForumA New Kind of Science: The NKS Forum
Pages:1
"If there is intelligence everywhere..."
(Click here to view the original thread with full colors/images)
Posted by: Constantino Constantini
I read Wolfram's book and what I consider as its central idea hit me as a blow on the head.
If there is intelligence everywhere and there is nothing special about ours, then everything human crumbles to insignificance, not only freewill (a problematic concept anyway, even Tolstoy in “War and Peace” argues forcibly against it), but also purpose, meaning, cause, even space and time.
Well, if this is the case, it wouldn’t make any sense to complain, would it? But there are some clarifications that I would like very much to receive.
1. I take it that the underlying idea of the book is expressed in a sentence of the note “Physics as intelligence” of page 1191:
“…perhaps intelligence could exist at a very small scale and in effect have spread throughout the universe, building as an artifact everything we see.”
Now, is there a single “cellular automaton”, breathing life, as it grows, into everything? Where, in this case, do the different rules proposed for crystals, fluids, plants, animals, and all fundamental networks come from? All the examples given in the NKS book obey the same rule all the way down. None, unless I missed something, produces countless “daughter” and “granddaughter” rules. Or, on the contrary, does every single speck in the universe follow its own simple program? Surely, then, their interactions are much more important than the single programs, since it seems evident that arbitrarily small sub-worlds cannot develop independently. But the book states that intrinsic complexity is substantially more relevant than external influence. Are there other alternatives?
2. Is complexity in itself an almost sure sign of creativity?
The NKS book says that creativity is not a human prerogative; so it appears as unavoidable that there is no inherent difference, from an neutral point of view, between—say—a scan of a painting of Piero della Francesca, the summit of the Italian Renaissance, and a random collection of 0s and 1s. Also, if the Principle of Computational Equivalence is right (if it is wrong, it is wrong with an admittedly very good case) any class 3 or 4 automaton can, given enough time, produce by Poincaré recurrence “the Legend of the Cross” or any other work of art, even all those that have not been created. It is as if the mind had lost any significant link with reality, and was floating in a space totally separated from matter and brains: a Cartesian nightmare. (Descartes himself was apparently uncomfortable about it and decided that the two substances must meet in our pineal gland!)
3. If, having studied these systems for more than twenty years, Wolfram is quite sure that there is no way to deduce the subjacent rule from the system’s behavior, it must be so. In this case, what objective is proposed for investigations of the new kind of science? Science—new and old—turns up to be as irrelevant as old painting and everything else. For if all we can do to check the performance of a cellular automaton is to watch how it develops, its explanatory power is about null, and it would be as rewarding to watch and see nature itself unfold. But can we really drop our Ptolemaic bias? No matter how atypical or marginal the standard patterns established so far, they are still more precious to us all than mere tautology. The meaning that our brain assigns to the world might be a fantasy, but we naturally look for it, and can’t help doing the old kind of science as we go.
Posted by: Jason Cawley
If there is intelligence everywhere
There is computational complexity everywhere. Whether there is anything more to "intelligence" (e.g. consciousness, perhaps shared with other animals but not e.g. by the weather) than computing ability is something NKS leaves open. Formal ability "to be 'clever'" is not exclusive to human beings. It arises much sooner than that.
NKS certainly plays with the idea that anything across the threshold of universality may be called "intelligent" in some meaningful sense. It also addresses the question of how one would know (e.g. the discussion of animal intelligence), regarding it in the end as a difficult problem.
We successfully recognize human artifacts mostly through their simplicity, implicitly appealing to a large area of shared history, specifics of human engineering, etc. There is no good reason to expect artifacts of a fundamentally different intelligence to be "stamped" with the same markers, if that background history is absent.
there is nothing special about ours, then everything human crumbles to insignificance
If you think significance arises only from uniqueness, I suppose. But our intelligence, with all its baggage of history, is certainly well across the threshold of universality. Which means it can in principle encode any sequence or formal system. In a sense, anything past that threshold potentially has characteristics of "microcosm". If significance is read a little more literally, universality can be seen as a marker for "significance", certainly of ability to signify.
Otherwise put, if the issue is ability to get the universe into our heads on some level of abstraction, we easily pass. Is this made "insignificant" if it is in principle shared by other systems? Why should it? It is not regarded as a drawback to the significance of one person's thoughts that other people might think them. In fact, sharing them is usually regarded as sort of the point.
also purpose, meaning, cause,
At some level these are of course human concepts slicing up our experience for the benefit of our own understanding. But that is hardly any objection to them, or to their "significance". Purpose ascribed to systems we see acts as part of a short hand description of their behavior. Some systems can be described more easily by specifying "attractors" they tend toward than by specifying a rule of transformation from one step to another. Meaning to something may involve that something's characteristic processing or abstracting of some aspects rather than others. Cause as a concept is most solid when a bit of behavior is more or less repeatably but with variations, and then refers to overall changes resulting from modified initial conditions.
Would some other intelligence slice up experience in the same ways? Probably not. But that is again no objection to our methods of doing so or their "significance". Our systems of representations can capture what is going on in any finite system, following the rules of our own systems of representations. Including, in principle, any other universal system's different encoding - though that might be as complicated the original problem all over again. Understanding something else's encoding of the world might be as hard as understanding the world. Which means in practice, arbitrarily hard, but in principle not impossible.
the underlying idea of the book is expressed in a sentence of the note Physics as intelligence of page 1191
Rather than "the underlying idea of the book", I'd call that a fun, speculative throw away line about a possible consequence of easy universality and difficulty recognizing artifacts. It echoes animist or pantheist or theistic notions, that there are little intelligences running around, or that the universe as a whole thinks, or that the universe is an artifact. Certainly the universe is capable of supporting universal computation. The only open question is where that ability first arises. Since that is a matter of putting together simple components a certain way, nothing special in the way of underlying components is needed.
In the past, people have thought a very elaborate system of hierarchical biological "engineering" was necessary to support anything like thought. Of consciousness in some suitably limited biological sense that may still be true. But of bare ability to calculate it isn't remotely true. That comes in much earlier - so much earlier that it is easy to imagine it being successfully accomplished with "nano" components.
is there a single cellular automaton, breathing life, as it grows, into everything?
NKS does not suggest a cellular automaton in particular, as those have too much built in structure. The fundamental physics chapter instead suggests casually invariant formal networks. But yes, the idea is there could be one underlying rule, that produces all the other arrangements seen on all higher levels of abstraction as internal states. A universal system can in principle encode other universal systems, which then evolve according to their own rule at their level of abstraction, as the underlying or encoding rule evolves according to its rules.
See the construction of the universal CA, starting on page 644, to see what this involves. Essentially, subgroups in the state of the underlying rule correspond to elements in the higher level system. If you look at the behavior of such a system on the lowest level of abstraction, it is following only one rule, that of the underlying system. But if you look at it "course grained", only able to "see" groups of elements above some larger size, you will see only sub-groups of the whole system, acting as elements and interacting with one another. The rules of the interaction of these grouped elements are not be those of the underlying rule, but those of the rule being emulated.
We see this in artificial cases, where we have explicitly engineered the whole thing to produce that effect. But in principle, there is no reason a higher level rule can't "emerge" within the behavior of underlying rules chosen at random. For some simple resulting rules, this can be seen in numerous NKS examples. Class 4 systems give rise to groups of sub-elements that move about and interact as elements, sometimes with much simpler interactions than the underlying rule.
does every single speck in the universe follow its own simple program?
Not the idea NKS is proposing. Rules are relations among, not specific to each bit. In fact, the idea of NKS is that special bits with their own special "natures" are not necessary at all to explain any of the complexity we see. Instead, arrangements alone, among generic bits, are quite sufficient.
their interactions are much more important than the single programs
The NKS idea is that their interactions are all you need, and not even interactions according to tons of different rules. There are rules rich enough that you can stuff any arrangement you can imagine inside them, so to speak. That was already known, as the phenomenon of universality. But the rules in question were thought of as special, artificial, highly complicated. This was mostly because they were engineered deliberately to produce the universality effect. NKS shows that much simpler rules not specifically designed to be universal are universal. And suggests this is very common, occuring just about everywhere we see behavior that is not obviously simple.
2. Is complexity in itself an almost sure sign of creativity? The book says that creativity is not a human prerogative
Actually, NKS says little about "creativity" as a specific subject. But the ability to produce striking and unexpected phenomena is certainly not a human prerogative. In talking about forms seen in art, NKS notes that repetition is favored and even nesting arises late and rarely (see in particular the discussion of ornamental art in the note on page 872, and the following notes through page 875). Most other "creative" art studies natural forms, copying them with a greater or lesser degree of idealisation. This is sign that we learn about creating interesting form from methods nature uses all the time, not the other way around. Utilitarian artifacts usually have simple shapes, mostly because we must be able to predict their properties to use them for intended purposes.
any class 3 or 4 automaton can, given enough time, produce by Poincaré recurrence "the Legend of the Cross"
I wouldn't wait around. Specified complexity explodes very fast with the number of elements involved. The space of possibilities for even quite simple rules quickly exceeds the number of particles in the universe, and associated times can quickly exceed the universe's age. The point is not that eventually anything universal produces everything.
A given painting was probably constructed according to vastly simpler rules. Meaning there is no need to run through a giant possibility space, like that of a bitmap of that painting. Data compression regularly reproduces master art works to a level sufficient for reasonable visual satisfaction in a few dozen kilobytes. And the elements the painter himself used - a realistic human figure e.g. - unify gobs of even that reduced amount of data around a few simple and reproducible forms. Instead of a class 3 CA and monkey typing, go to an art school and watch the development of simple techniques.
as if the mind had lost any significant link with reality
Surely the ability of the brain (or mind) to model reality is a link with reality. Brains can because many systems can, of which they are one. That there are others does not mean brains don't or can't; they obviously do. Instead it helps explain how it is possible for anything like brains to succeed in modeling even arbitrarily complicated things.
there is no way to deduce the subjacent rule from the system’s behavior, it must be so.
Except by applying the rule, of course. Or by watching what happens. In simple enough cases you can say more than that just from looking at the rule. In some less simple cases you can say rough or approximate things, after seeing a small bit of the systems behavior (e.g. that class 3 results will look random, for all practical purposes). And sometimes limited "speed up" methods may be possible, that effectively calculate the result of applying the rule twice in a single step. But the basic amount of computational work involved in figuring out exactly what a complex system will do is, in general, irreducible. You do the same amount of computational work with a model as complicated as the system itself, or you let the system do that work and watch what it does.
what objective are you proposing to the investigations of the new kind of science?
Modeling the universe, as opposed to predicting the universe. NKS in effect says that the world as a whole is extensively model-able, but only predictable in pockets of simplicity. In the past, it tended to be assumed that if you could model something you could predict it. It was implicitly assumed that the behavior of the model itself would always be predictable. And that simply isn't the case.
So instead of an "inside outside" knowledge gap, between simple models and a sometimes slippery reality, you get a kind of "vertical" gap between rules and behaviors, both inside (in models) and outside (in natural systems). Even if you know the rules, the behaviors arising from them have to be looked at experimentally. But you can match up the overall behavior, and the sort of rules giving rise to such behavior. Then you have a model, and can watch either the model or the system itself. The model is successful if it captures enough of what is going on in the natural system, that you can learn as much watching the model as you could watching the natural system.
it would be as rewarding to watch and see nature itself unfold.
Even if you watched nature unfolding with complete incomprehension? Watching both evolve the same way, you see how nature is doing the trick. In addition, the model may be more tractable to "play with" than natural experiment. You can do millions of computer experiments. For some processes, much more rapidly than the natural system evolves. In some cases, it is obviously impractical to play with the natural system by conventional, empirical experiment. Think of galaxy formation, for example.
The meaning that our brain assigns to the world might be a fantasy, but we naturally look for it
Why should it be a fantasy? If it is a successful encoding of aspects of real world behavior, there is no reason to distrust it simply because some entirely hypothetical alien intelligence might cut up the world into different formal categories. Who cares? Let the hypothetical alien understand the world its way, but understand it, yourself. An encoding of the natural world either reproduces the main features of your experience of it, or it does not. If the former there is no reason to mark it as "fantasy". If the latter, you just need to improve your models. That it is possible in principle to capture what you see is assured by the universality of our own brains.
I hope this is interesting.
Posted by: Ken Zweibel
I found it interesting, and it coincided with some of what I took away from the book - which was heartening.
I also thought that the idea was a universal rule and emulations within the universal rule that appear as either objects interacting or worlds of perception. If we perceive ourselves as separate, that is part of some higher level emulation; but at some lowest level, we are all together, whatever that might be like. Is this what the book is implying? Or what I and others, no doubt, are constructing from it?
I guess from a purely mechanistic view, we always were part of the universal unity, separated only by our mobility and the repetitive structure of our cells, induced by DNA and our environment's ability to support us. So being part of the unity from a 'rule' standpoint shouldn't be a surprise. The question is, then, whether we should project consciousness into the mix, too.
I think the idea of universal intelligence is quite intuitive; and adding it to the mixmaster of trial and error evolution is also excellent. Somehow, the way things are evolves faster and 'better' than simple trial and error. It's good to think that along with the hardwired instructions of DNA we have access to this greater intelligence.
Posted by: Gunnar Tomasson
Constantino asked:
what objective are you proposing to the investigations of the new kind of science?
And Jason replied:
Modeling the universe, as opposed to predicting the universe. NKS in effect says that the world as a whole is extensively model-able, but only predictable in pockets of simplicity. In the past, it tended to be assumed that if you could model something you could predict it. It was implicitly assumed that the behavior of the model itself would always be predictable. And that simply isn't the case.
So instead of an "inside outside" knowledge gap, between simple models and a sometimes slippery reality, you get a kind of "vertical" gap between rules and behaviors, both inside (in models) and outside (in natural systems). Even if you know the rules, the behaviors arising from them have to be looked at experimentally. But you can match up the overall behavior, and the sort of rules giving rise to such behavior. Then you have a model, and can watch either the model or the system itself. The model is successful if it captures enough of what is going on in the natural system, that you can learn as much watching the model as you could watching the natural system.
Comment:
Re. Jason's point that NKS is concerned with "Modeling the universe, as opposed to predicting the universe."
Let me play Devil's Advocate.
If it aims to "model" the Universe while NOT "predicting" it, the NKS view of the Universe is akin to that of the Hubble Space Telescope - it provides us NOT with "answers" but food for thought.
And, re. the following:
"In the past, it tended to be assumed that if you could model something you could predict it."
Within the Solar System, the predictive success of Newtonian Orbital Mechanics was a function of some NON-OBSERVED and unspecified STABLE underlying aspects of Solar System Structure that gave rise to observed orbital patterns.
Thus, a "model" was neither more nor less than a means of converting past observational data into "predictions".
Finally, re. the following:
"The model is successful if it captures enough of what is going on in the natural system, that you can learn as much watching the model as you could watching the natural system."
If "watching the natural system" cannot in principle teach us anything about the NON-OBSERVED aspects thereof, the same must be true of any NKS "model" of natural phenomena.
Gunnar
Forum Sponsored by Wolfram Research
© 2004-2009 Wolfram Research, Inc. | Powered by vBulletin 2.3.0 © 2000-2002 Jelsoft Enterprises, Ltd. |
Disclaimer
vB Easy Archive Final - Created by Xenon and modified/released by SkuZZy from the Job Openings