[Implications of simulated (intelligent) life?] - A New Kind of Science: The NKS ForumA New Kind of Science: The NKS Forum
Pages:1
Implications of simulated (intelligent) life?
(Click here to view the original thread with full colors/images)
Posted by: joe_auricchio
Hello,
(Disclaimer: I haven't read the book. I haven't taken any math courses more advanced than one-variable calculus (and we're only up to differentiation!) I *have* just watched a presentation Stephen Wolfram gave at MIT two months ago, and I'm completely fascinated by what NKS could mean.)
I'd like to start some kind of discussion here, among you people, who are presumably far more qualified to talk about these matters than I am. I just want to see how the discussion goes.
Here are some nice questions to start things off... Can intelligence and sentience be simulated? Can they be simulated in *any* computationally sophisticated system? If Mr. Wolfram could demonstrate how to take a collection of nodes and turn them into a space-time continuum type of thing complete with relativity and elementary particles, then could we start simulating that, and simulate a universe that would contain (based on its own laws of physics) intelligent inhabitants that could figure out Wolfram's ideas on their own and repeat the process of simulation? Could we be in somebody else's copy of Mathematica? (That "somebody else" being outside our Mathematica-defined universe). Would they necessarily need to be at least as computationally sophisticated as our universe? We appear to have a definite "state" at a definite "time"; that's how we define our existence - but how would that work if we were a printout of a cellular automaton? We feel like there's some kind of 'cosmic cursor' that points to a particular line or frame of the automaton or other simulation - but is that only an illusion? Wolfram sort of covers this in his lecture at MIT - he discusses how we couldn't tell if the universe were constantly updating everything, or updating only one cell at a time like a Turing machine. Would there be any way to tell if we were in somebody else's copy of Mathematica? Could we communicate with them? How? Would there be any certainty they'd perceive their reality in remotely the way we perceive ours? Is there a lower limit of what kind of sophistication is needed to be self-aware? (That is, what is the simplest computational model in which the pattern, or small bits of it, would understand its own existence) And a nice big philosophical one... So, if our universe is a computational model, who set the initial conditions?
I hope some good discussion comes from this. I probably won't be able to participate in it, as all of the details of NKS are completely and utterly above me. But it will be fun to read! :)
Posted by: Phillip Craig
You might be interested in a book titled "The Matrix and Philosophy", William Irwin Ed., Open Court Publishing,2002. This is a collection of essays/articles dealing with the underlying philosophy in the Matrix movies which, of course, are all about a simulated existence for humans (except for a few who think they have caught on). One of the articles in the book talks about whether we could really tell if we are in a simulation. The 3rd movie is just coming out this week and we'll see if they can resolve any issues or just dig themselves in deeper.
Here is one possible view. There is no reason to think we could not possibly be in a simulation. If we are the product of a reality based on one or more simple rules running for billions of years and an exponentially huge number of state updates, then this reality could be viewed as a simulation. The pervasiveness of this reality (simulation) would be 100%, meaning there are no detectable seams. Why? Our brains are the product of this reality and only function according to the rule(s) of this reality. If there were any seams we wouldn't see them or, having seen them, reason them away. Mostly, though, they would be invisible to us.
But, does the fact that you or I can speculate about simulated existence mean that we can detect it after all? Maybe the unpredictability of the simulation has created a biological brain that can ask the right questions and get into the right state to see how the simulation works - and also maintain some level of sanity. If this is a simulation, where is the "real" reality? Someone would have to have some way to detect the "real" reality, or possibly another simulation in progress, in order to know that we are not in it.
Posted by: KarlGamer
To some degree being part of some gigantic computer just makes sense.
A common thought that is almost held as a law is the idea that everything is made of something.
People seem to always agree to this without dispute.
Then someone brings up the old paradox of shooting an arrow.
You shoot an arrow form point A to point B.
Before the arrow gets to point B it must reach halfway between A and B. Point C.
But before it gets to point C it must reach halfway between A and C. Point D.
But before it gets to point D it must reach halfway between A and D. Point E
But before it gets to point E it must reach halfway between A and E. Point F.
But before it gets to point F it must reach halfway between A and F. Point G.
…
If the arrow has to travail thought an infinite number of points how does it ever get to point B.
Is there a smallest unit?
It’s hard for me to think that there is. Then I think “Ok now what is it made of.” Funny.
Well maybe we are in a big three dimensional monitor. The bounds of which we are not able to see as Mario is not able to see the screen he is on.
Posted by: Jason Cawley
I'll take the original questions one by one and address each.
"Can intelligence and sentience be simulated?"
The Turing test of an AI is whether ordinary humans can't tell whether the responses they are seeing are coming from a human operator or an AI. Various trials of various systems have been made and they can sometimes do OK in limited areas of expertise. But with any generality allowed to the tests they typically fail in obvious ways. This shows that the full generality of human communication has not yet been simulated by an AI.
After that one starts getting into definitions. Chess programs play chess well enough that they deserve to be called intelligent. Without that needing to mean "sentient". If intelligent means good at something because able to understand it in all its variety and implications, then expert systems can be intelligent. If you like you can call that "clever" and distinguish it from "intelligent", letting the latter include some sense of self-awareness or consciousness.
Then one immediately gets to an interesting problem, which is that nobody knows the real relation between consciousness or self awareness on the one hand, and cleverness on the other. It is entirely possible consciousness in biology may arise early, easily, even in quite dumb creatures by our standards. You can't teach a dog to play chess, and it'd lose to a shareware chess program every time if you could. But the evidence that dogs are conscious in much the same way we are, is pretty good.
The point being that they are not arranged as one after another on a single hierarchy of "more". Instead they are distinct things. You can have clever things that aren't conscious (at least, we typically think so), and conscious things that aren't clever. Of the two, cleverness has been easier to get out of computers. Consciousness seems easier to reach in animals.
We probably simply don't yet know what natural trick is used to create consciousness. It is also possible there isn't one, that instead the barriers are to communication rather than to consciousness (i.e. maybe there are lots more conscious things than we typically suppose, but communication between them is fantastically hard). But the more likely possibility, it seems to me, is that there is some trick to it which we simply haven't figured out yet.
"Can they be simulated in *any* computationally sophisticated system?"
Since we don't know the trick of consciousness, we can't say for certain. If that trick requires system characteristics AB and C, which exceed the requirements of computational sophistication alone, then presumably there can be computationally sophisticated systems that cannot simulate consciousness. If they are universal they can simulate arbitrary contents of (at least, any finite) consciousness - that follows from the definition of universality.
But there is in principle a difference between recording everything some other system has done, in some encoding, and carrying out the same physical process. The encoding properly captures every relation in the original, considered as relations or as an abstract computation. But it doesn't use the same media, the same marks, etc. If any of those matter for the hypothetical consciousness trick, you might get a recording of contents of consciouness but not consciousness (say).
"take a collection of nodes and turn them into a space-time continuum type of thing complete with relativity and elementary particles, then could we start simulating that, and simulate a universe"
Correct. We'd have a little toy universe running the same rule as computer code. The relations in that toy universe between element and element, would mirror exactly the relations between element and element in the real universe. The relations in the toy case would use instances encoded as sets of gate positions or charges on much larger hunks of silicon. But those huge hunks would stand to each other in the same, mirrored relations as tiny elemental relations do in the real universe.
You should understand though just how "toy" it'd actually be. We'd be talking about a few dozen to a few hundred nodes. We'd only mirror the relations seen at the most elemental level, therefore. Because to make something like an iron nucleus (to pick something at random) would take gobs more. The universe is bigger than our computers. Its nodes - if it has nodes - are a lot smaller than our hunks of silicon. It processes fast; our fastest computers are hunks of rock just sitting there by comparison. And its got a 10 billion year head start. It is not like we are going to catch it.
"would contain intelligent inhabitants"
All the intelligent inhabitants we know about take rather a lot of stuff to make. Far more than we are going to be able to simulate. It is an amusing philosophy exercise, I understand, but it is not a practical question. If we succeed in making intelligent AI, it will be because we figure out how to make intelligent AIs, or figure out the trick nature uses to produce consciousness. It will not be because we've simulated the universe's own rule for the equivalent of 10 billion minus a few thousand years and some such has independently evolved inside our whole universe sim.
"Could we be in somebody else's copy of Mathematica?"
Sure. Many varieties of philosophical idealism (and not a few religious traditions) have had that idea, one way or another. Not that they've insisted on using the right program (lol). As a "could we", it is easy to answer. As an "are we", it is distinctly harder to answer, and most consider it an open question whether one could tell.
My own take on that, is that real means what it means to us, because we are the ones that mean it. If reality is a sim then sims are real, and don't deserve to be sharply distinguished, as "mere" sims, because in that case there is obviously nothing "mere" about them.
"need to be at least as computationally sophisticated"
Yes, certainly, but that isn't hard. At least, we are certain it isn't hard if there aren't any funky continuous infinities under it all, that aren't just our methods and approximations. If there are, then we don't know whether it'd be hard. But computational universality is not hard, and that is enough to emulate any finite algorithm.
"We appear to have a definite 'state' at a definite 'time'"
That one is actually a fun subject. That is our psychological intuition, yes. Our models used to be built that way, too. Lately they haven't been. Instead we've made time something much more slippery, in physics that is, to fit things we see and make traditional mathematical descriptions of it elegant. If a casual network model for the universe works, then it will probably restore much of this intuition. Updating a network uses a definite sense of "state", and a sense of time that is close to our intuitive one. Though presumably the time we can measure and that we sense will be an emergent, not a primitive underlying, consequence of the universe's rule. Because we measure only within the sequence of updates the rule provides, not outside of it.
"Could we communicate with them?"
Something of a conceptual problem, especially with the pronouns ("we", "them"). We are a "we" to us, and to each other. To somebody's Mathematica, we are some internal subsystem. Presumably a complicated enough one that knowing exactly what we are going to do is an experimental matter - though theologians doubtless have their various opinions on that one.
I know the rule 30 rule, and I can set it up to be simulated. It does not however follow that I know what it is going to do, before I see it do it - and being a dim finite simpleton, I don't. So, I set up the rule. The rule is purely deterministic. I simulate the rule. But the rule's execution has to tell me what it is going to do, right when it does it. So, is my computer - or for that matter the rule - telling me what it is doing? I'd say "sure", since "telling" or "communicating" means transfering information to me that I didn't have before.
If you get definitional and metaphysical enough, however, you can easily tie the common sense understanding of communication up into such knots you wouldn't see it as possible. If communication only arises if something spontaneous does the completely unforeseeable, then whether there can be communication turns on whether there is spontaneity. If by hypothesis there isn't any then you've defined communication out of existence. Personally, I prefer to leave words with common meanings, and let the metaphysical chips fall where they may. So I'd say there is communication within our reality whether it is deterministic or not, and therefore there is no reason to require spontaneity to call something communication.
"Would there be any certainty"
No. Certainty is a straw man, in anything remotely like this context. Ask instead whether there could be evidence for or against something, or whether something could or could not be known. Certainty has a definite, very strict meaning in philosophic circles. And nothing here remotely reaches its criteria. (Indeed, philosophers regularly argue about whether anything does).
"perceive their reality in remotely the way we perceive ours?"
Remotely? Sure, remotely enough. It is a taffy-ish word. I can stretch it and stretch it until there is some overlap. More to the point, though, presumably both the hypothetical "they" and the phenomenal "us" are way above the threshold of universality, and the premise or "if" inside which this entire discussion takes place says at least our universe is computable. Then anything that can happen in our universe has some possible encoding for us, and for them. They don't have to be the same encoding, but they both have to be "preservative" or faithful encodings of the same finite process.
Otherwise put, one universal "mind" can think any finite algorithm, so can another, therefore they can in principle think the same set of things. Since these are large possibility spaces and actualizations are always vastly smaller than possibilities, there is no requirement that they actually will coincide, or "think the same thoughts". But they could.
"Is there a lower limit of what kind of sophistication is needed to be self-aware?"
Are earthworms self-aware? I don't pretend to know. Dogs probably are. I don't think it is such a high standard. Now, computational universality is a known threshold, and it is also very low. Elementary CAs can pass it. But we also know that (at least up to funky continuous infinities etc) it is all there is on the computational sophistication scale. Is universality necessary for self-awareness? Probably. But it isn't hard to get, so...
"if our universe is a computational model, who set the initial conditions?"
It would be interesting to see how simple they are, before trying to answer that. If they are about the simplest possible, I think it argues likely universe uniqueness. But also not much, if anything, in the way of setting anything. If they are the 12 billioneth possible initial condition for the 4 millioneth simplest rule, then the universe we know is probably a speck of flotsam in a lot of other stuff. If they are the 5041st simplest initial conditions for the 37th simplest rule, Plato call your office.
My opinions on your fine questions. I hope they are interesting.
Posted by: Phillip Craig
This idea seems a little difficult to get across, but here goes. Treat this as speculation, so replace "is" or "are" with "might be".
An underlying causal network is producing patterns with varying degrees of complexity. Some of the patterns look like rocks. Other patterns look like sentient beings. Extending this further, some patterns look like protons and electrons, and some patterns look like galaxies.
The difference between these things is complexity and aggregation, although the differences are pretty big (to us).
Sentient beings are complex patterns (or aggregations of multiple interacting patterns) that have characteristics that behave just like thought processes and memory. Our belief that we think thoughts and have feelings is a function of the causal network firing through it's current set of patterns and interactions.
Our memory of experiences is based on patterns in the network that persist for varying amounts of time.
The speed at which the network fires is incredibly fast compared to our sense of time, so huge numbers of state changes could occur before a single thought occurs to us. Thoughts are sort of ripples in the network that arise due to the indeterminant behavior of the network, or because of ripple effects from previous causes (thoughts). No sentient creature would be independently thinking or reacting. In fact there isn't any such thing as an independent sentient creature (or rock for that matter).
Does this mean no free will? Not necessarily. The network behaves in a manner that gives the more complex patterns the appearance of free will. Sentient beings are a kind of visible proxy for the network. In fact, so are rocks and protons - just on different scales. In practical terms, saying we have free will seems to be correct. Rocks don't seem to have free will because they just aren't complex enough.
But talking about visibility or "seems to be correct" is a problem in this scenario. If a human is a somewhat persistent and complex pattern in the network, then all functions such as seeing and thinking are in the network. There is no escape from the seamlessness of this system.
This type of speculation reminds me (possibly) of one of those pictures that looks like one object, but with a slight shift of perspective, looks like something completely different. This seems to be one of the effects of NKS - to think about whole new sets of possibilities that don't build on previous assumptions.
Posted by: Philipp Michel
This might be slightly off-topic, but some might find this sort of neat. It concerns the Turing test.
I always had the feeling that many people underestimate the difficulty of the test. In a recent lecture, Dan Dennett gave a great example that gives you some nice intuition of just how hard it would be to make a machine provide satisfactory answers to some questions.
I don't know where this example originated from, but here goes:
"Imagine the letters D and J. Now, rotate D counterclockwise by 90 degrees. Now place it on top of the J. What's the weather like?"
So simple, yet so hard.
Thanks for the interesting posts above. They are a joy to ponder.
Posted by: dan miller
Lots to think about here. Let me start with one important point.
Originally posted by Jason Cawley
[quoting original thread poster:]
>>"take a collection of nodes and turn them into a space-time continuum type of thing complete with relativity and elementary particles, then could we start simulating that, and simulate a universe"
[JC:]
>Correct. We'd have a little toy universe running the same rule as computer code. The relations in that toy universe between element and element, would mirror exactly the relations between element and element in the real universe. The relations in the toy case would use instances encoded as sets of gate positions or charges on much larger hunks of silicon. But those huge hunks would stand to each other in the same, mirrored relations as tiny elemental relations do in the real universe.
You should understand though just how "toy" it'd actually be. We'd be talking about a few dozen to a few hundred nodes. We'd only mirror the relations seen at the most elemental level, therefore. Because to make something like an iron nucleus (to pick something at random) would take gobs more. The universe is bigger than our computers. Its nodes - if it has nodes - are a lot smaller than our hunks of silicon. It processes fast; our fastest computers are hunks of rock just sitting there by comparison. And its got a 10 billion year head start. It is not like we are going to catch it.
[dbm:]
I'd like to examine this hypothesis. You're saying we could only hope to model a few dozen or hundreds of nodes. First of all, you're not accounting for "Moore's Law" (the idea that computers double in speed every 1.5 - 2 yrs). If the trend continues, in 20 years we will have 6 orders of magnitude more computing power. Instead of a few hundred nodes, we would then be able to effectively simulate up to a billion nodes or so. How many atoms does that represent?
But this all presupposes a couple of things. First, do we really know the computational complexity of say an iron atom? It would be interesting to get that information before we make predictions about what we can do.
Second, is it really necessary to simulate everything down to the lowest level? A good understanding of fundamental physical processes may enable us to take some massive shortcuts and still retain the important chemical and electrical properties that account for almost all interesting behavior at the typical size and temperature of living organisms. And of course, the simulation we're thinking of isn't *required* to have exactly the same laws of physics as our Universe. Perhaps a much simpler and easily simulated subset or alternative set of physical laws would suffice to allow artificial evolution to a high level of complexity.
Third, beyond Moore's law, I would like to propose the following hypothesis: Computational capacity of our computing technology will increase to the point where it starts to approach the underlying computational capacity of raw matter and energy. Why? Because as we understand the physical workings of our universe better, we will become more and more efficient at bringing its fundamental properties to bear on arbitrary algorithmic problems. To take a concrete example, if it turns out that at at the quantum level we really do have some sort of massively parallel processing (ie 'many worlds' theories), then Quantum Computing will eventually be able to take advantage of it. Alternatively, if the Universe is 'only' capable of computation proportional to the mass (energy) involved, we will get down to the atomic level (quantum dots, nanomachines, molecular computing, etc) and do as much computation as the substrate can handle.
So either way, in my scenario, there will come a time when we can simulate say a cubic foot of room-temperature matter at maybe 1/100 of real time, or something like that. That's a lot of computing power.
Anyway, just some food for thought - dbm
Posted by: Kovas Boguta
Concerning Wolfram's fundamental theory: One thing to remember is that the amount of nodes in an area of space does not correlate 1-1 to the amount of "matter" present.
What we percieve as matter is structures present in the already-existing network of nodes. It probably takes a lot of nodes to support our 3 dimensional space. The fact that it takes quite a bit of matter to bend that space suggests that the amount of extra nodes needed to create some "matter" may potentially be quite small.
However, I do agree with the sentiment that in time we should be able to simulate a piece of universe on the scale visible to the naked eye.
Posted by: Gunnar Tomasson
Re. the following:
The fact that it takes quite a bit of matter to bend that space suggests that...
Comment:
There is no such "fact".
As for 'space bending', a photon passing by its limb would have been deflected towards the sun on strictly Newtonian principles.
Here is an extract from an article on the subject matter:
"Curiously, the idea that light rays may be deflected by solar gravitation had been examined a century before Einstein. The 1804 volume of the 'Astronomische Jahrbuch', edited by J. E. Bode in Berlin, contains an article by Johann Soldner (1777-1833), who later became director of Munich Observatory. He pointed out that light should be influenced by gravitation if light consists of material particles emitted by a source. When these corpuscles come near to a center of gravitation, they must move around that center in an orbit which, owing to the high velocity of light, will be hyperbolic. Consequently, the direction of the light ray will be altered. According to Soldner, a light ray that just graces the sun should be deflected 0.875 seconds of an arc.
"Soldner's paper was forgotten, for about 20 years later A. Fresnel's investigation convinced astronomers and physicists that light consisted of waves rather than particles, and there seemed to be no reason why waves should be influenced by gravitation. Long afterward, when Einstein reexamined the matter in connection with relativity theory, he obviously did not know about Soldner's paper.
"According to the theory of relativity, light rays move along geodesic lines in the four-dimensional continuum. These lines are curved in the neighborhood of masses, with consequences similar to those phenomena that are called gravitational in classical theory. Therefore, light rays must move in a curved line when passing near a gravitating body. The deflection of a light ray just grazing the sun was predicted by Einstein's theory to be 1.75 seconds of an arc -- twice Soldner's value." ('The Einstein Shift - An Unsettled Problem' Schmeidler, F.; Sky and Telescope, 27:217-219, 1964, reprinted in 'Mysterious Universe - A Handbook of Astronomical Anomalies', compiled by William R. Corliss, The Sourcebook Project, Glen Arm, MD 21507, 1979, pp. 674-675)
Gunnar
Posted by: Tony Smith
dan miller said:we would then be able to effectively simulate up to a billion nodes or so. How many atoms does that represent?
Given that we seem to be talking about a causal network at the Planck scale which is around 10^-20 the diameter of a proton, the answer will always remain "an insignificant fraction of an atom" and for a vanishingly small time, given that Planck time is around 10^-43 of a second.
Even if we can agree with the thrust of Kovas Boguta's assertionthat the amount of extra nodes needed to create some "matter" may potentially be quite small.
or at least "small" relative to the number of nodes needed to create ever expanding space itself, that still does not get us near to Kovas's optimistic conclusionHowever, i do agree with the sentiment that in time we should be able to simulate a piece of universe on the scale visible to the naked eye.
We should not oversell ourselves on the prospect of A New Kind of Science providing a persuasive model of space time energy matter any time soon. I have started to argue elsewhere that space time energy matter is a harder problem than the origin of life or of (human) consciousness, and we should all be aware just how slow the progress of the revolutions in biotech and info tech (AI) has been on those lesser problems.
The value of NKS is to open our minds to new ways of understanding that might serve to show us that a natural explanation for all the wonders of the universe is at least conceptually achievable, even if for reasons of irreducibility it might not be modelled by any system within reach of human patience.
I don't see that as in any way pessimistic, but rather as underlining the limitless opportunities we have to do what we can in the world we find ourselves in.
Posted by: KarlGamer
The other day my brother explained to me that .99999 (repeting)
is acutaly 1.
33/99 = 1/3
66/99 = 2/3
99/99 = 1
and quickly I thought that must be how the arrow gets to point B
Originally posted by KarlGamer
You shoot an arrow form point A to point B.
Before the arrow gets to point B it must reach halfway between A and B. Point C.
But before it gets to point C it must reach halfway between A and C. Point D.
But before it gets to point D it must reach halfway between A and D. Point E
But before it gets to point E it must reach halfway between A and E. Point F.
But before it gets to point F it must reach halfway between A and F. Point G.
…
If the arrow has to travail thought an infinite number of points how does it ever get to point B.
My brother agrees that the two ideas are related.
Could one of you explain it to me?
Posted by: Mike Lin
If you truly believe in continuous space then it is not problematic that for any two points, there are infinitely many points in between. They are separated by "infinitely small" - that is, infinitesimal - distances. Therefore the sum of the distances between them may be finite.
The properties of the infinitesimal are established in the field of mathematical analysis. It is not an easy thing to understand. The mathematician considered the father of this field, Weierstrass, eventually suffered a nervous breakdown.
By the way, a slightly different way to say the point you raised is the following: 1/9 = .1111..., 2/9 = .2222..., ..., 8/9 = .8888..., so we should expect that 1=9/9 = .9999....
Posted by: Jason Cawley
I'll start with Dan Miller, Moore's Law, and how much we may be able to simulate. Yes computing power might expand by a factor of a million in 20 years or so. But this will not translate into the ability to simulate a million times as many nodes. The reason is that possibility spaces blow up much faster than linearly. That is the reason a network simulation with present computers - which can plow through trillions of possibilities - would still have only a couple orders of magnitude worth of nodes.
To help think about this, consider chess. There are 64 squares and 32 pieces of 6 different types, each duplicated once in a different direction. A very small finite number of generating possibles to arrange, in space and in sequence. And there are more possible games of chess than there are particles in the universe.
With Go, there are only 3 possible states for a space on the board, and spaces almost always change in only one direction, only once each. Yet chess is a small enough problem that computers can be taught to play it at the master level, while nobody has succeeded in writing a decent Go program, for an ordinary 19x19 board. The number of possible Go board states on a 19x19 board is 3 to the 361st power. The number of possible games of Go is higher, because you can reach any of those states by multiple paths.
Now you may think, "OK sure, but we also get the same explosion on the computing side". Yes. As memory increases we have far more elements we can arrange to track whatever the system we are modeling is doing. As processing speed increases we can examine more of those possible arrangements per unit time. The problem is you still have to compare one to the other. Comparing two large sets for correspondance of internal arrangement is astronomically more to do than comparing two small sets for such correspondance.
When searching for a candidate physics rule, moreover, we must search not only on the arrangement but also on the rule. Effectively this means considering a slice of the possible paths through the arrangement space. Some sequences of arrangements correspond to simple rules; others do not. The problem would be completely intractable if one had to consider all possible paths to all possible arrangements.
The simple rule idea drastically reduces the task required. That is part of the point. You can arrange the possible rules in an ordering from simple to more complex. You can arrange the possible arrangements (states) from simple initial conditions to slightly less simple ones. Then you can exhaustively examine only the few billion simplest on each score, and if necessary randomly sample from billion size slices of more complicated rule or initial condition possibilities, much farther out in the possibility tree. But you do not have to consider every possible path to every possible arrangement.
Moreover, if you know something about what you are looking for you can stop the examination of possibilities relatively quickly, when that particular possibility can't give what you are looking for. You can see if the rule would have too many dimensions, or no ordered sense of time, or violate special relativity, or lack this or that other basic experienced, emergent property of real physics. As soon as you can see it will, you throw it out and move to the next.
Getting these search filters right is always a tricky business, because you don't want to throw out interesting things too soon, just because you couldn't see beforehand how the system might manage to create property X. But these tricks prune the possibility tree early, to make the task computationally managable. Raw computing power is necessary certainly, but alone will not get it done.
Now, what about physically realizable limits on the computing power we manage to achieve, and what about the question of how low level the rule matching has to be to get physics out of it? First on the quantum computing (QC for short) side. The view suggested by NKS is that much or all of the phenomenal "continuity" we see is an emergent property of underlying discreteness, while at the bottom level the universe is finite state. If this is so, there will not be continuous infinities available from QC. Maybe multiple paths exploited, but not continuous infinities worth.
If on the other hand continuous infinities are not abstractions of ours but are real, then the rule emulating problem is intractable even in principle, regardless of the level of computational resources available. You can't span a continuous infinity's worth of possibilities for every bit. QM works because a finite basis is always sufficient to answer a particular finite question (giving measured probabilities, not determinate answers). But if you don't pose a finite question you can't pick a basis. Concretely, you can ask in QM what the probability is that an electron will be in this finite region in this definite time period. But you can't even ask where it is, without specifying such a region to ask about.
The idea of finding the universe's rule just by guessing it, in other words, is only going to work if the universe is finite state at bottom. Indeed, it is only likely to work if it is in addition a relatively simple rule. Even a complicated enough finite state rule, we'd never be able to guess in practice.
What is encouraging about the idea is that all of this complexity blow up stuff is working the other way if you ask "how simple could the universe's rule be, and still give rise to all this phenomenal variety we see?" Just as Go manages to make a rich game out of extremely basic elements, and not even all that many of them, the universe might manage to make enourmous variety out of a trivalent network with a simple and fixed evolution rule.
Note I did not say "must", I said "might". If it is simple enough we may find it just by searching. That it could be simple enough to find just by searching, and still give rise to a whole universe, is the possibility space problem working in reverse.
Now, can we eventually get to the level of computing power of the universe itself, meaning the inherent information capacity of the laws of physics themselves? If we had the rule, perhaps. But that won't find it for us. We can only arrange our computers to exploit the elements of physics we already understand. Yes, the point of increasing our physical understanding is in part to make more of the computational resources of the universe available to us.
We can compute with wafers of silicon because we understand what electrons do in a crystal, and have developed (quite difficult) engineered processes to manipulate them in fine detail. We could compute with underying nodes if we (1) understood what the underlying nodes do (we don't and won't until we have the rule) and (2) could engineer vastly harder processes to manipulate them in fine detail. The latter might be practically impossible even with the former.
There are plenty of things even our present physical theories predict which we can't even test directly. E.g. various quarks don't live long enough to be observed rather than inferred, no matter how we set up experiments. As a consequence we can only see what they do in situations that arise "naturally" in the events by which we produce them, and can't see what they do in others we might want to contrive.
There is no reason to expect such practical limitations to vanish. The lower the scale we move to, the larger they will probably become. The point is merely that theoretical knowledge and practically exploitable information need not coincide. Still there is no question large advances in the former will lead to advances in the latter.
The reason any bit of the universe can be used to predict the behavior of other, larger portions of the universe is the one Kovas identifies, though he spoke of one particular example of it. The larger portions are often doing relatively simple things. Or they differ only minutely from a simplified state or background. Abstraction works because one can then track only the evolution of those differences. Also, we can detect the differences without knowing all the details of the background (though when we know those, it certainly helps immensely). This is basically what enables all of our present physical theory to operate as well as it does.
I hope this helps.
Posted by: Jason Cawley
Next I want to make a basic philosophic point for Philip Craig. Patterns in things, arrangements or forms, are the real varying element from one object or state to another. This is really a point about "looks like" and "is". Appearance should not be mentally rendered "mere appearance"; appearance means data, the phenomena. Which are not "unreal", but are the standard against which theories are tested.
He said "Some of the patterns look like rocks. Other patterns look like sentient beings." Right. But some of the patterns are rocks, and some of the patterns are sentient beings. Yes, they are both patterns in a network, by hypothesis, and a network with the same underlying rules. But that simply means that "rock" and "sentient being" describe emergent phenomena of that network, or apply to appearances up at a higher level of aggregation. It does not mean that "rock" and "sentient being" do not apply or are "illusions".
A theory either lets rocks be rocks and sentient beings be sentient beings, or it does not. If it does not it is just plain wrong. That rocks are rocks is data. Notice, I did not say "no rock is sentient" - that is quite another question and in principle an empirical one. But rocks appear to be rocks, with all the properties of rocks, because they are rocks. You can have a network theory that shows the cause of the properties of rocks, as emergent consequences of an underlying structure or set of rules, that in themselves may have nothing that particularly reminds you of a rock.
That theory then tells you the cause of an appearance. But it is an appearance. Appearances are real, they are events, they are phenomena. Theories are either adequate to them or they are inadequate as theories. A theory of rocks can explain the appearance of rocks, that they have properties AB and C. It cannot explain those properties away, without becoming false as a theory.
"The difference between these things is complexity and aggregation, although the differences are pretty big"
Yes those differences are pretty big. They are also (1) real and (2) the only game in town, when it comes to differences. Otherwise put, when we learn or think everything is a pattern in an underlying network, we are noticing or positing a unity behind appearances, a set of properties they have in common. Nothing distinct can be correctly so posited. Distinct means different means particular. The idea of "form" is simply that arrangement is the cause of distinctness or particularity, that it is what differs from one instance or particular to another.
On this understanding, one thing differs from another in its arrangement. It would be somewhat misleading to say "one thing differs from another only in its arrangement", because there really is no "only" about it. Differences are traced to arrangements. Whenever there is a phenomenal difference, a different arrangement is posited to account for or track it.
Take the case of sentient beings as complex patterns or interacting aggregated patterns (which really fits inside the previous phrase). You say "have characteristics that behave just like thought processes and memory." I'd say, "have thought processes and memory". "That behave just like" is all one can ask. Yes, on this view it is an appearance or a phenomena, therefore to be explained by an underlying level of arrangement. But it is a real appearance, and that is all one needs to start using the word "is" rather than "just like".
The point is to not restrict "is" to the underlying network level, and relegate everything else to a different category (of "seems", say). The higher levels of aggregation are every bit as real. Real is not a synonym for "underlying" or "basic cause" or "dissolved below all arrangement". Real is the opposite of unreal or false, not of aggregated or arranged. Yes, higher level aggregated appearances are arrangements or forms. So is every definite thing (as distinguished from "everything taken together"). When everything is an arrangement, arrangement ceases to be mentally preceded by "mere".
It is not an NKS point or a science point, but a philosophy point. It is about how the "form" view of things arranges its categories, where it puts things. I hope it is useful.
Forum Sponsored by Wolfram Research
© 2004-2008 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