35
$\begingroup$

A civilization has a "magic" computer with memory and processing capacities far beyond what our physics says is possible. They decide to run a massive simulation of Conway's Game of Life with a random "big bang" initial state. The goal is to see if, given enough simulated time steps and a large enough random initial canvas, intelligent life will evolve.

But here's the question: even if intelligent life did evolve, how would the simulator civilization know it was there, when all they (or rather, their algorithms) can see is a semi-chaotic pattern of blinking dots?

Edit: The "life" bit isn't that important; "intelligent non-life" would also work. And the "signs of intelligence" need not be definitive. What sorts of things would scientists (and philosophers) consider when trying to decide if intelligence had arisen? What sorts of debates might they have?


Philosophical digression (trying to provide a substitute for the best of the comments moved to chat)

Skeptic: What is life? What is intelligence? If you will not define these two, your question is unanswerable. [line quoted from user Molot]

Enthusiast: You don't have to define intelligent life to search for signs of it in astronomy. Why should this be any different?

OP: Both points are good. Consider why you think other people have minds (as opposed to being mindless automatons). The most relevant response here is that minds have explanatory power: there are certain actions people take -- such as having discussions about the nature of minds -- that would be absurd if they did not, in fact, have minds. If we were searching for alien intelligence, we would look for signs like buildings or radio message patterns that can most easily be explained by the existence of an alien civilization. In the same way, we can look for patterns in the Game of Life that -- while not violating the rules -- are virtually impossible to imagine without deliberate design.

Skeptic: Where to begin? First of all, postulating intelligence cannot possibly add any explanatory power to what we already have. The Game of Life is deterministic and we know all the rules.

Second, you can't assume intelligence just because you have an elaborate mechanism for accomplishing a purpose. Many things (like eyes) appear to have been designed for a specific purpose, but were in fact produced by natural selection without any actual intent.

OP: Let's take these points one at a time.

Explanatory power in a deterministic system: Consider the laws of statistical mechanics -- most notably entropy. Technically speaking, if you know a physical (Newtonian) system perfectly, the laws of statistical mechanics are completely unnecessary: you can predict exactly where every particle will be after any specified amount of time. But in practice, applying statistical mechanics will allow you to make certain predictions much more cheaply, such as the average velocity of the particles in one particular spot. So statistical mechanics has explanatory power even though in this instance it adds no information.

Enthusiast: I think your criteria for "explanatory power" is still too stringent. Many explanations have no predictive power whatsoever, even in terms of making cheaper predictions. For instance, if (in real life) we received a radio signal that was the first twenty primes repeated over and over, we would probably postulate an intelligent source even though that explanation doesn't really help us predict anything more cheaply.

OP: That's true enough. Cheaper prediction is a nice illustration of explanatory power in the face of determinism, but it's not the whole concept. Like many philosophical ideas, "explanatory power" is precise enough to be useful but vague enough that no two people have exactly the same definition.

Skeptic: How exactly is that useful?

OP: Let's just agree to disagree on the usefulness of imperfectly defined terms. To be considerate of our readers' time--

Skeptic [aside] That ship has sailed.

OP: --let's move on to your second point from earlier.

Designed technology vs evolved organ: Let's remind ourselves of the earlier exchange:

OP: We can look for patterns in the Game of Life that -- while not violating the rules -- are virtually impossible to imagine without deliberate design.

Skeptic: You can't assume intelligence just because you have an elaborate mechanism for accomplishing a purpose. Many things (like eyes) appear to have been designed for a specific purpose, but were in fact produced by natural selection without any actual intent.

That's fair enough. In our own world, certain kinds of patterns -- such as projectiles or wheels -- seem to be exclusively the domain of designed technology rather than evolution. But even in our own universe these distinctions are hardly intuitive, and spotting them in a completely alien universe like Conway's Game of Life would be all but impossible. Perhaps radial propagation of information might be a sign of designed technology? (In the Game of Life, diagonal, vertical, and horizontal movement are much more natural than movement at any other angle; propagating information in all directions at the same rate could be something highly desirable that would never show up "on its own" (without intent), even through evolution. But who knows?)

Another way to think about it is that artifacts designed with intent tend to be good at things that have nothing to do with their own survival. For instance, if we found something in a simulation that appeared to be a program for playing the board game Go exceptionally well, it would certainly be worth publishing as a "sign of intelligence" whether or not the program itself were considered intelligent.

Another thing to look for would be advanced communication -- especially concerning mathematics, since that is universal. But it's also possible that things that are difficult to compute in our universe, typically requiring advanced intelligence, can be done comparatively simply in a GoL universe, and vice versa.

Enthusiast: Perhaps we (or our fictional protagonists) could scan the GoL universe for patterns that obey differential equations! Clearly no differential equations could show up in such a universe without deliberate intent.

OP: Maybe. But differential equations are so powerful I have a feeling they would work their way in just fine without intelligent help.

Anyway, perhaps if we ask on WorldBuilders.SE they'll have some suggestions.

$\endgroup$
  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$ – Serban Tanasa Dec 4 '16 at 18:41
  • $\begingroup$ Food for thought: Can there really be intelligence given that the game is entirely deterministic? $\endgroup$ – Kevin Dec 4 '16 at 19:01
  • $\begingroup$ Just adding that Douglas Hofstadter's book 'Godel Escher Bach' covers a number of similar topics such as a conversation with a ant colony and a variety of theorems, like Godel and Church, along with recursion. I sense the author may have read it already but throwing out there for anyone's reading list looking to understand this question. $\endgroup$ – Morrison Chang Dec 4 '16 at 22:53
  • $\begingroup$ @SerbanTanasa: Could you please restore at least some of the comments? Some of them are referred to in some of the answers, and I think they included some interesting and useful responses -- for instance, clarifying misunderstandings in a way that a monologue simply cannot. $\endgroup$ – Charles Staats Dec 5 '16 at 9:44
  • $\begingroup$ Any relevant information should be (added) in the body of the question. $\endgroup$ – Serban Tanasa Dec 5 '16 at 12:41

16 Answers 16

25
$\begingroup$

With a large enough canvas of random data and enough cycles, intelligent life would evolve, assuming the Chuch-Turing thesis. Even weaker than that, assuming the universe is no more than non-deterministic.

Rough proof:

Conway's game of life is a Turing-complete game. You can create a general purpose computer in it. This computer may require a large initial state and be fragile and slow, but that just increases the size of the initial seed and cycle time before it can exist.

This computer is magical, so it can handle both of those issues.

In fact, you can have an unbounded number of such computers, running an unbounded number of different programs.

Assuming the Church-Turing thesis, at least one such program is intelligent.

Assuming the universe is no more than non-deterministic, a NDTM can be simulated in a TM, and a NDTM can thus describe the universe.

So given enough space and time and uniformly random initial states, somewhere in the infinite field there will be a turing machine that proceeds to simulate the evolution of every possible universe from an initial big bang with physics like ours.

One of these simulated universes would be our universe (without the magical computer, naturally).

Our universe appears to contain intelligent life.

So that is proof of existence. Now, we should examine what it would take to find this intelligent life.

The Conway's game of life doesn't directly contain it. Rather, it contains a pattern of life that can be interpreted as a computer (actually, unbounded numbers of such patterns, most of which are short lived due to flaws), which simulates a non-determinstic computer, which proceeds to exhastively simulate entire families of universes.

No human being or non-magical computer could, if someone pointed out the section of the system that was doing this computation, even interpret it or confirm if that was indeed the section in question. The parts of this machine's state would be literally larger than the entire universe's data capacity. We couldn't even look at the part in question.

It is so slow that the number of cycles required to describe even the merest fraction of a fraction of a fraction of a ... of a fraction of a ... of a fraction of a fraction of a second of the universe might be a number too large to express in this universe.

But that was just an existence proof. In theory, a much smaller system could actually house inteligence.

We can look at our universe. One of the possible explanations for our physics is that we are experiencing an inverted platonic wall of shadows. Our physics actually plays out on a 2+1 dimensional boundary of our universe, with the 3+1 dimensional interior of the space being a hologram " "projected" by it. One of the motivations behind this possibility is that there appears to be an information limit in our universe proportional not to volume, but to the surface area of a region!

Physicists have created mathematics for toy universes that match this pattern, where there is a N dimensional physics system with a N-1 dimensional system that holographically determines the behavior of the N dimensional system.

Such a technique might easily be the easiest way for conways game of life to create a complex enough universe for evolution to occur in. So the 2+1 dimensional game of life might generete a holographic universe where the "interior" perspective of the intelligent beings does not correspond to the 2+1 dimensional "underlying system" that generates the hologram.

Stepping back again, even if that doesn't happen and the resulting intelligences do end up living in flatland, they might exist on ridiculous scales. The equivalent to a "cell" of our universe is the Planck length, or a Planck patch.

This is about 10^-35 m, or 10^-20 times the size of a Proton.

A proton is about 10^-15 m in size. The observable universe is about 10^26 m.

-35 -- Planck scale
-15 -- Proton scale
-10 -- Atomic scale
1 -- our scale
7 -- planetary scale
14 -- Solar system scale
20 -- Galactic scale
26 -- Universe scale

If we take jumps of a factor of 10 million per step, we get roughly:

-5 -- Planck scale
-4 -- ???
-3 -- ???
-2 -- Proton scale
-1 -- Atomic scale
0 -- our scale
1 -- planetary scale
2 -- Solar system scale
3 -- Galactic scale
4 -- Universe scale

The "cell size" of our universe is ridiculously far away from our scale. Atoms are roughly half way from that to the entire universe.

If we assume this is typical, then even in the regions where there is intelligence, figuring out a reasonable model of physics that they are experiencing so far away from the rules of Conway's life would be beyond our current skills.

They would no more exprience the rules of Conway's cells that we notice the rules of string harmonics when we throw a baseball.

$\endgroup$
  • 5
    $\begingroup$ I'm pretty sure neither Dr. A. Church nor Dr. A. Turing would say "Assuming the Church-Turing thesis, at least one such program is intelligent." $\endgroup$ – Eric Towers Dec 3 '16 at 23:25
  • 1
    $\begingroup$ @eric if humans are incapable of solving any problem that a TM cannot, barring P-Zombies a sufficiently advanced computer can be intelligent. And if any such computer exists in the mathematical sense, it exists within the unbounded game of life. $\endgroup$ – Yakk Dec 4 '16 at 3:37
  • 1
    $\begingroup$ What is the basis for claiming any of: (1) Solving problems (that can or cannot be done by a TM) is unambiguous proof of intelligence. (2) The set of all TMs acting on all possible inputs contains the set of whatever it is humans do. This sounds as hopeless as "the set of all PDA implementing all possible transition functions contains the set of TM computations". (3) Any such computer (perhaps an analog computer, requiring uncountably many internal states) exists within the (countably) unbounded Game of Life? $\endgroup$ – Eric Towers Dec 4 '16 at 4:04
  • 2
    $\begingroup$ The Church-Turing Thesis makes a claim about the limits of computations. It does not place a limit on humans nor does it place limits on "intelligence", whatever that is. Claiming that humans are computations is far more reductionist than either Church or Turing would go. Claiming that intelligence is TM-computable is far more starry-eyed than either would go either. $\endgroup$ – Eric Towers Dec 4 '16 at 4:16
  • 2
    $\begingroup$ I agree with @EricTowers; in particular, the claim "Solving problems (that can or cannot be done by a TM) is unambiguous proof of intelligence." is baseless, at the moment. All CT thesis states is that a number-theoretic function is computable (defined as such) if it can be computed by a Turing machine. Assuming this is akin to intelligence is a massive philosophical and technical leap. $\endgroup$ – socrates Dec 4 '16 at 4:28
19
$\begingroup$

Conway's GoL.

Conway's Game of Life tend to towards two modes: Extreme chaos and extreme order. (To see this, start a simulation of about a million random pixels and see what happens)

Different areas can show different behaviour.

In extreme chaos mode every pixel changes seemingly without patterns.

In extreme order mode every pixel either is stable or changes in a very short cycle, usually 2 time steps, sometimes 3, very very rarely more than 3.

Life

Life would be different. There would be patterns, but they would not be strict. As the life form multiplies, there would be large-scale repetition of approximate patterns.

You would recognize life by this: Large number of "cells" that looks more or less the same, and that seems improbable from a local point of view.

The Environment

Around these cells would be an environment of the usual chaos and order patches.

This environment will be deadly to the life. Simply random changes will kill off the poor innocent little cells.

In the beginning these life forms would follow a very simple strategy: Multiply faster than the environment kills them off.

Advanced strategies, "intelligence"

Eventually, however, they would become smarter.

They would change the environment to be more friendly to them.

They would also find ways to recognize each other. They could either compete, to ensure more space for their own children, OR they could cooperate, to tame the environment.

Some would pretend to cooperate with their neighbours, but secretly preparing to kill them off.

Then recognition would become more complex, to see the difference between close relatives that can be trusted and cooperated with, and strangers that has to be fought.

The life forms would invent war.

At what point should we call them intelligent? That is a matter of definition. I leave that definition up to you.

Looking from the Outside.

For the outside observers, all this looks like more and more complex patterns, over larger and larger areas and longer and longer time intervals.

I suspect it would be quite clear to the observers why things are happening, this cell attacking that cell, that cell cooperating with that other cell to build a wall and so on.

One indication of intelligence would be that the observers no longer understands what happens. At first you will have observers going "Oh, so that is why it did the odd thing 10000 time steps ago!". Later they will say "I have no idea how they achieved that!".

That point is as good as any to say that the life has become intelligent.

$\endgroup$
  • 19
    $\begingroup$ "I suspect it would be quite clear to the observers [..] that cell cooperating with that other cell [..] and so on" Highly doubtful. Any "intelligent lifeform" would likely consist of billions of cells or more. If someone were to an observe a human by watching their (trillions upon trillions of) individual atoms, do you think they'd see intelligence or chaos? $\endgroup$ – BlueRaja - Danny Pflughoeft Dec 2 '16 at 14:34
  • 2
    $\begingroup$ @BlueRaja-DannyPflughoeft They do have the ability to zoom out too. I'm not sure GoL could behave this way anyway, the rules of which cells die and live is hard coded I believe. $\endgroup$ – DonyorM Dec 2 '16 at 15:59
  • 2
    $\begingroup$ Big risk of anthropomorphizing pixels here. There's no planning or foresight going on in this scenario. Using words like pretending, cooperating, preparing, declaring war, or asking about motivations is blindly mapping human behaviors onto something that is simply not human. $\endgroup$ – Nuclear Wang Dec 2 '16 at 17:07
  • 2
    $\begingroup$ @DonyorM As far as we know, the rules of the universe are hardcoded. Apart from certain quantum effects which might be random, the universe is deterministic. $\endgroup$ – wizzwizz4 Dec 2 '16 at 17:16
  • 4
    $\begingroup$ @Matt Big risk of anthropomorphizing molecules here. There's no planning or foresight going on in this scenario. Using words like pretending, cooperating, preparing, declaring war, or asking about motivations is blindly mapping intelligent behaviours onto something that is simply not intelligent. The premise of the question is that GOL might be able to produce intelligent life. $\endgroup$ – wizzwizz4 Dec 2 '16 at 17:20
11
$\begingroup$

A computer scientist here, I will chime in with my educated guess. You will not find them, unless you have billions of people searching for intelligent behavior in this simulation.

Problem is massive, the size of the canvas would be extremely large. I am talking about 10^20 by 10^20 size canvas. The number of atoms in a human body is ~10^28, each atom requires many cells in CGoL. Even if your intelligent beings of automata is much smaller in scale, we are still talking about a pattern that has a size of trillion by trillion. Understanding that this pattern is acting intelligently is a magical task. The interactions at the border of this pattern will be chaotic as well as its internal actions.

I think this problem explores this, if we are living in a simulation, it is most likely that our simulator will never understand that we are intelligent and alive.

Factoring in the discussions from the comments: The size estimates that are done here is for an intelligent system that is not alive. Also the canvas size is not to simulate our universe physics, if you are to do that you will need canvas sizes that are larger than $10^{185}$ cells in the best case.

$\endgroup$
  • $\begingroup$ Question says "A civilization has a "magic" computer with memory and processing capacities far beyond what our physics says is possible" so we could assume their computer could handle this task. $\endgroup$ – user3161729 Dec 2 '16 at 11:55
  • $\begingroup$ This is my own instinct as well, but I'm hoping the WorldBuilding community will surprise me. The one ray of hope here is the "magical" computer, which allows the civilization to run complex algorithms, neural nets, etc. in parallel across the entire vast canvas to "mine" for signs of life or intelligence. That's your analogue for "billions of people searching." The question is what should these algorithms even be looking for? $\endgroup$ – Charles Staats Dec 2 '16 at 16:50
  • $\begingroup$ Magic computer does not mean magic people. Imagining a computer capable of running 10^40 cell CGoL is possible, people magically finding patterns of sizes reaching up to trillions in each dimension is not within the realm of possibility. $\endgroup$ – Cem Kalyoncu Dec 2 '16 at 16:50
  • 1
    $\begingroup$ The problem is that the life forms in that automata will be absolutely alien to us. I mean not their biology is different, their physics different sort of alien. We may postulate which type of life could happen, and we might find it, but probably we will miss many of them. In fact finding a feasible life form is also a hard task. That's why I said billions of people. If you get more people on the task, one of them might come up with a possible life form. Then we could search for that using our magic computer. $\endgroup$ – Cem Kalyoncu Dec 2 '16 at 16:54
  • 1
    $\begingroup$ @CharlesStaats I agree about underestimating the likely required canvas. On that note, regarding the "large, durable region ...": What if the Big Bang in our universe was just the moment when a stable, self-replicating pattern that happened to model our current physical laws first appeared and rapidly asserted itself, somewhere amidst a large canvas of quantum-mechanics/GoL soup? (Just tossing that out there as a fun concept to explore, since you're already going down this path.) $\endgroup$ – mtraceur Dec 4 '16 at 9:19
5
$\begingroup$

Many of the answers here seem to be concerned with whether something like life could develop in a GOL simulation. I will instead begin by assuming that intelligence could develop, and try to answer the question "how could we detect it?" by comparing to our own universe. I'll assume we would know it if we saw it, too -- that after a bit of observation, we could tell when cells are working together and making decisions. My conclusion: it would still be insanely difficult for outside observers to detect -- it's the biggest game of "Where's Waldo?" I can imagine!

To get a sense of scale, if we're trying to compare to our own universe, then we've got to zoom way out. The closest thing we have to a quantized length in our universe is the Planck length. So, imagine each GOL cell in this simulation as a single Planck cube. Now, think about the scale when we zoom out to an entire universe.

I'm not aware of random Game of Life simulations having been conducted and examined on boards larger than a few thousand by a few thousand -- hardly enough to represent a quark, and hundreds of orders of magnitude smaller than what we're talking about here! But a common pattern to most simulations is that pretty quickly, most of space becomes empty, with scattered cells forming what we can think of like "background radiation." In some areas, there is more complex activity taking place, but these areas are few and far between. Our own universe is structured much the same way, so I can image that a search for intelligence in the simulation would be like a search through our universe.

Suppose you had the ability to observe the 100-billion light year sphere of the universe that we can see. Let's even assume you have a computer with enormous memory and processing power to help you. You can look anywhere, zoom in or out, and automate some search algorithms. Would you be able to find the humans?

Given enough resources, I imagine you could make a program that filters out the huge portions of space that are mostly empty. What's left would be the galaxies, which you could recursively search for solar systems and planets.

The first issue is knowing how far you need to zoom. You don't know how big lifeforms will be -- will they contain thousands of GOL cells? Trillions? More? If you're looking at our entire solar system at once, you won't know humans are here -- you have to know to zoom in on the surface of Earth to see our cities. If you're zoomed in to a molecule, you'd have no idea that it's part of a human cell. You have to know to zoom out to see collections of cells working together.

The second issue is that even if you know what scale to look at, there's simply too much to search through. It's estimated that there could be more than a trillion planets in our galaxy alone, and at least 100 billion galaxies in the observable universe. Each of these planets is itself enormous, and would require a close examination to rule out intelligent life (imagine looking at the Earth 50,000 years ago -- if you're going too fast and don't look hard enough, you might not know humans are there!). However you do this, you're not going to be fast enough to process these planets in anything even close to a reasonable amount of time.

The only thing I can think of that could help this search would be understanding the laws of physics in the simulated universe. Of course, we know the microscopic rules; they're just the GOL rules. But I mean understanding the rules on a larger scale. How do the cells tend to group together? Let's call these groups particles -- how do the "particles" interact? On what time and distance scales do they affect each other? From there, we can build up the rules of "chemistry" in the simulation. At the opposite end, we could also start looking at a large scale and discover the equivalent of astrophysics. This would take a huge investment of time and research -- humanity's understanding of physics and chemistry was built up over thousands of years, and was aided by our ability to interact with our world.

If we had this knowledge, it could potentially help filter out planets faster, just as astronomers do in our own world, based on chemical makeup and environment. But it's still not clear that it would help an outside observer much, since even understanding chemistry and physics in the abstract doesn't help you understand what is and isn't needed for life -- you need some examples of life for that.

So I'm left with the conclusion that even if intelligent life develops in your simulation, unless for some reason it's vastly more common than in our universe, you won't be able to detect it.

$\endgroup$
  • 2
    $\begingroup$ There's also the question of time scales, which I think is just as vexing as the spatial scaling question. But do remember the "magic computer" -- in addition to running a ridiculously large GoL board, we can also run pattern recognition of some kind in parallel across, say, all "GoL galaxies" or "GoL solar systems" at once, assuming we can tell the algorithms what a "galaxy" or "solar system" is. $\endgroup$ – Charles Staats Dec 5 '16 at 9:36
5
$\begingroup$

The search for automated intelligence could resemble the search for extraterrestrial intelligence -- it's all about signals. The search for automated life is similarly simple -- it's all about entropy.

The magic computer program could look for regions within the simulation where entropy decreases locally. Those regions may hold life-like constructs. Once those regions are identified, the program could monitor whatever passes between them (gliders, spaceships and so on) and identify those patterns which are complex enough to carry information.

This approach is a heuristic, not an algorithm. It certainly won't identify human-level intelligence. Instead, it might find ant-like intelligence, if such a thing evolves within the simulation.

The researchers running this simulation have a problem that is the opposite of real-life sciences. In our universe, we struggle to discover what is fundamental to physics. We know how things seem at a human scale, but things like quantum mechanics and string theory are far from intuitive. For the simulation, the researchers know the laws of Conway's Game, but they have no idea what a mind inside the Game might perceive. The fundamentals are obvious, but the emergent structures are unknown.

How can you tell the difference between an ant and a person when the things you can directly observe are equivalents to strings and quantum configurations?

$\endgroup$
  • $\begingroup$ My favorite answer $\endgroup$ – user21263 Dec 5 '16 at 15:48
4
$\begingroup$

This could be done in a two step approach:

  1. Identify possible intelligent system
  2. Confirm (or infirm) the presence

For step 1, we could consider the application of the causal path entropy here

When following a causal path, entropy is based not on the internal arrangements accessible to a system at any particular time, but rather on the number of arrangements it could pass through on the way to possible future states.

The article points to similarities between this observable behaviour and intelligence. The super computer running the simulation could also be programmed to detect those macros effects.

For step 2 we could think of a variant of Turing's test. Today it is purely a computer intelligence test of wether a computer can pass as human which limits intelligence to the human intelligence. Once a civilisation has been exposed to more intelligence forms (either biological or not) a more refined test would be available to test those candidates identified in step 1.

$\endgroup$
  • $\begingroup$ To run anything like a Turing test, you have to be able to communicate. How do you propose to do this? $\endgroup$ – Charles Staats Dec 2 '16 at 16:55
  • $\begingroup$ @CharlesStaats since we have a magic computer, I would use a magic mouse and magic keyboard to inject disturbances and assess how the rest of the system reacts. Especially since it has been proven that a game of life can be a full Turing machine, interfacing is possible. It could even be planned from the beginning. $\endgroup$ – Auberon Vacher Dec 5 '16 at 15:53
4
$\begingroup$

As many others have pointed out, you need a definition of intelligence to be able to answer this question. One possible definition of intelligence for the purposes of this question is an entity that you can communicate with. Whether it is "truly" intelligent in a deep philosophical sense is pretty much irrelevant in this case (see the Chinese Room Argument).

Again, as others have pointed out, GOL is completely 100% deterministic. However, communication means altering the state of a thing externally. In the case of GOL, you would communicate by introducing changes to the system that are against the deterministic rules.

The way to test if intelligence has developed as a consequence of some initial conditions is to see how the system reacts to these externally introduced changes. Is it possible to cause a set of changes to which the system responds in a consistent manner? Can we eventually create a language by which we can assign interpretations to outputs? Does the potential intelligence assign its own interpretation to these changes (i.e., does the response to a given input take significantly different forms depending on circumstances, including previous inputs)?

$\endgroup$
3
$\begingroup$

I'm no scientist but as I understand it there are certain things that you can look for that are otherwise improbable.

For instance your simulation might perhaps watch out for movements of dots that are unexpected. Imagine a group of people pushing a big rock up a hill. Normally rocks go down hills or stay where they are. You wouldn't expect one of significant mass to move up a hill. That indicates that something is doing it on purpose. Obviously in this case replace the rock and hills with various patterns.

That might not prove that the life is intelligent but I expect you could keep looking for these changes until you have enough or they become complex enough to be considered intelligent.

$\endgroup$
  • 5
    $\begingroup$ When all you see are quarks how do you even figure out what a rock is or that it usually moves downhill? $\endgroup$ – Charles Staats Dec 2 '16 at 7:29
  • $\begingroup$ It would be a bit more abstract than that. You wouldn't even assume you'd have rocks appear at all in the simulation. Rather you would see that over time certain patterns take place such as entropy. Any differences to the usual pattern should be investigated. $\endgroup$ – user3161729 Dec 2 '16 at 7:36
  • 1
    $\begingroup$ For example, usually if you start with random pixels you get (among other things) a handful of gliders going off in random directions. If you see a bunch of gliders close together and all moving in the same direction that's suspicious. Who arranged for them to all go in the same direction? $\endgroup$ – user253751 Dec 2 '16 at 10:39
  • 1
    $\begingroup$ You said "watch out for movements of dots that are unexpected". What does that mean? Do you mean that the rules to Game of Life were broken? If the rules were not broken, then was it really unexpected? If we program a computer to increment an integer, would we be surprised when it gets to "7"? $\endgroup$ – James Dec 2 '16 at 14:40
  • 1
    $\begingroup$ @James: No. But it is easily possible for a computer to follow a simple rule and for that rule to have unforeseen consequences. The Game of Life itself is a great example – could you anticipate that gliders were possible? Could you anticipate glider guns? Could you anticipate glider gun guns? You can understand everything there is to know about the basic rule, and still be very very far from understanding all of its implications. $\endgroup$ – Ben Millwood Dec 4 '16 at 4:47
3
$\begingroup$

The book Permutation City by Greg Egan explores this concept exactly; the emergence of an intelligent species from within a 3-dimensional cellular automata, evolved over a large time-span. If you are interested in this question, Egan explores it in-depth over the course of the novel. The prose is suprisngly good for an amateur writer, but the concepts explored and thought experiments involved are fascinating.

As mentioned, and similarly to the perspective taken in the book, human observers look for signs of intelligence by observing the symbolic interactions between entities; once they deciphered their "language", the analysis of the complexity and semantics of the language was used to determine whether the species could be considered "intelligent"; i.e. discussing highly abstract concepts, such as mathematics.

$\endgroup$
3
$\begingroup$

Suppose you have a HUGE simulated canvas. Suppose you display an image of it small enough to actually look at. Suppose you observe that the chaos turns into several groups of pixels that each move around, approach others, change direction. Would you consider that "possibly intelligent"?

But note that the original rules of the simulation guarantee that solid patterns of significant size cannot form easily and cannot exist for very long. The centers all have too many neighbors to survive.

$\endgroup$
3
$\begingroup$

Mason Wheeler has the right idea with his answer suggesting that we have to interact with the Game of Life to determine if intelligent life has formed. I wanted to expound upon why.

First off, Mołot has the correct question: what is "life" and what is "intelligence?" These are not easy questions. These are highly philosophical questions which have been batted around for millennia. If anything, they seem to be becoming harder questions as science pushes the limits of the world. I don't think we're going to answer them here, so instead I would like to pose an ever so slightly altered question which I think we can more likely answer here:

How can we tell if something has evolved in a Game of Life simulation that we should treat as though it was intelligent life?

This question is more answerable because it includes human limits. If a human cannot distinguish a GoL simulation from life, is it not reasonable to start treating it as though it is alive?

To get there, we need to be unable to distinguish a part of the simulation from a human. Pesky challenge, really. It's at the heart of the Turing test. At first pass, it's trivial. If it's in the GoL simulation, it's not a human, because it's a pile of bits! Problem solved! Let's all go out and have a cup of tea.

Of course, this really isn't fair. How do you know it's just a pile of bits? The answer is that you peered inside the game of life with your sneaky debugger tools and raped its virtual brain. By the time you were done, you could prove beyond a shadow of a doubt that that section of the GoL space was nothing but a pile of bits. How invasive!

Perhaps more importantly, we know that we can classify a pile of bits as "not alive" because we can easily create a clone of it, or a hundred clones of it. We know that its' hard to create a clone of Charles, or a clone of your favorite dog. It seems to be a dividing line: things that can be perfectly cloned cannot be called "alive." Dolly's DNA was cloned, so it isn't alive. However, Dolly's "children" are not the same as Dolly. Each developed differently. Each child is a different "living" individual.

What if we weakened the rules a bit, and allowed interaction. This means that humans can interact with the GoL space, setting some cells to living and some to dead, against the rules Conway put forth. This introduces a level of unknown into the mix, and creates our first opportunity for confusing the machine with a human. Let's let many humans interact on this same grid in this way, setting bits to 1 or 0. One of them might decide to use this as a communication channel, transmitting data like an old TELNET console. If you came across this bit, you'd notice that it can pass the Turing test, because there's an actual human behind it! While this may be cheating, it confirms that we can communicate with intelligent life using GoL space to do it.

To make things more exciting, let's step up the noise. Let's allow humans to make millions of modifications at each cycle. Make a really big grid, and lots of modifications. Many of these modifications may consist of a human feeding this game of life a random number source (such as de-tuned radio waves). Make enough of these modifications in every cylce that it's really not reasonable for any human to track all of these modifications.

Now we have an interesting situation. If we see something "intelligent" we can't prove that it was GoL rules evolving an intelligence, but we can't prove it was a human interacting either. We have to accept the unknown nebulous reality that there's merely something out there acting intelligent.

We also now know that that intelligence may have power. The intelligence may be someone on the other end, who is offended by what we do, finds us, and does something violent to us. Or maybe they just find our government and raise our taxes. Either way, an "apparent intelligence" found on the GoL field now has the ability to affect our lives.

Now we can answer the question of "When is it reasonable to treat something on the GoL grid as though it is alive." We can set some basic criteria:

  • We cannot reduce it to a deterministic algorithm (defined by a pile of bits). For all we know it may be a deterministic algorithm centered around a single random number generator, but we can't reduce it to just a pile of bits and a rule for how to evolve it.
  • We can communicate with it -- defined as it appears to respond to our interactions in predictable ways, but not so predictable that they could just be a deterministic algorithm.
  • It can communicate with us -- the entity needs to be able to make changes in the world that we can see and interpret.
  • It needs to be capable of helping us further our own goals, and it gives us the impression that it can do so better if we classify it as "an intelligent living being."

The last criteria is my favorite. It's not a criteria for something believe "alive" or being "intelligent," but it is a criteria for being treated as one. If it appears that classifying it as "intelligent" furthers our own goals, why would we not give it such a classification?

$\endgroup$
  • $\begingroup$ I think your first and last criteria are a bit too strong. For the last criterion (usefulness), when we search for signs alien intelligent life we don't require that it be able to help us further our own goals. And for the first criterion, a working clone of the AlphaGo algorithm would also violate the criterion, but would satisfy the requirements of the question in that it's much better evidence of "sentient life" (or pseudo-life) than pulsars ever were. $\endgroup$ – Charles Staats Dec 5 '16 at 9:18
  • $\begingroup$ But the last criteria is certainly interesting. A purely simulated intelligence from a GoL universe could be a useful collaborator in mathematical research since the sorts of intuition available to such an intelligence would likely be very different from the intuition available to any species in our own universe. If useful communication could be established, the different sorts of intuition might turn out to be complementary for solving mathematical problems of interest in both universes. $\endgroup$ – Charles Staats Dec 5 '16 at 9:22
2
$\begingroup$

I was originally having the gut feeling that the rules are too simple to allow for complex patterns to emerge and stay dynamically stable for prolonged periods. But that gut feeling is apparently wrong: Undecidability.

Building on the statement that a universal turing machine can exists as a GoL pattern, the basic precondition question could something resembling life evolve can be answered: Yes.

The next precondition question would be, will something that resembles intelligent life evolve? I have no basis for the answer, but I will pretend that since humanity did evolve from whatever basic life evolved first, it could, and for the sake of the question, it will.

So assuming intelligent life exists in that simulation, will they detect and recognize it?

Presuming they have infinite processing capability with their magic computers, any kind of pattern recognition they could think of is feasible to run. On that basis I would presume they will develop the tools necessary to analyze the games state to find cell clusters that show the most basic behaviors of life: self replication, local entrophy reduction and mutation.

Once life can be identified by these analytical tools, it can be observed specifically and its in the observers judgement if something is considered intelligent behavior.

The question then can be transformed into something else: Can they really develop the neccessary analytical tools? Their magical computer allows them to answer any question about past, present and future of the GoL, as long as they are able to formulate an analytical way of answering the question. It does however not free them of any limitiations they themselves are subject to: Number of scientists working on the problem, limit of their own lifetime and limit of their civilizations time of existence.

So it boils down to what abilities in general the experimenting civilization has. Given enough time to analyze, they will find intelligent life in their simulation, if they destroy the simulation or themselves before they found it, then well...

Its remarkably similar to the question if we will find intelligent life in our universe, really. If it exists and we don't destroy ourselves, we have a good chance of finding it, one day. Thats a big if though.

$\endgroup$
2
$\begingroup$

It could evolve life, so I assume it could even evolve intelligent life, but I don't know if we can know when something is intelligent, so probably we would evolve a sentient life form but we will not be able to recognize it.

Here's the game I played now:

enter image description here

In red you can see actually "living" areas, so there's some storm inside those areas and they are actually evolving. This area act just like few living cells because it continue to move and randomly eat or spawn some "still chunk" (areound living areas you see lot of still chunks, that are equiparable to cell foods).

Basically this game I'm doing I'm interacting the minimum amount of time to give the life some spread: you will learn quickly that clicking in certain areas will cause the food to just disappear, while clicking in certain areas will make life active for a while.

After some time you will find that you find a way to increase number of food, and you see that everytime food amount increase the life will last even longer before eventually coming to a rest: at some point life will continue to spread on its own, meaning there will be a immense space where there are living and still areas, overall food is growing, and living areas are growing too. This is life to me, and after some time maybe this life can become sentient and more autonomous.

By the way interacting randomly (and the minimum necessary I now reached this state):

enter image description here

Which is still continuing to "live" after 10 minutes. It is pretty similiar to real evolution, maybe bits of life evolve and extinguish, but it leaves "traces" that can be eventually be used again if new life evolve again, and slowly a "ecosystem" builds up.

$\endgroup$
  • $\begingroup$ I'm not sure "food" is the right term since the Game of Life does not have anything like conservation of matter or energy so far as I am aware. "Air pressure" might be a better metaphor -- the additional bits help to create a non-hostile environment. $\endgroup$ – Charles Staats Dec 5 '16 at 9:06
  • $\begingroup$ You are totally right, I'm more used to "games" concept of food "something you pick up and consume and continuosly spanw". Your idea of "bits that help to create non-hostile" environment" makes much more sense. $\endgroup$ – GameDeveloper Dec 6 '16 at 17:34
0
$\begingroup$

Has intelligent life evolved in a Game Of Life simulation?

No.

It's really that simple.

Among other things, intelligence necessarily requires the ability to analyze information and make meaningful choices about it. This can't happen in a purely deterministic system such as Conway's Game of Life.

Whether our world truly is nondeterministic, or is simply deterministic at fundamental levels so small that we don't have good models for them yet is a question for philosophers, but intelligent life must be capable of taking actions that appear to an outside observer to be nondeterministic, unpredictable choices. The Game of Life is perfectly deterministic; the entirety of the state of each round after the first is 100% predictable from the state of the previous round. Therefore, it is anathema to the evolution of intelligent life, no matter how large of a simulation is being run.

$\endgroup$
  • 5
    $\begingroup$ This answer assumes that intelligent life can't exist within a deterministic system. Just because a system is deterministic doesn't mean it's easy to evaluate. A sufficiently complex GoL can simulate any computer-simulateable problem. The problem isn't whether a computer can simulate intelligence, it's whether, within a massive and complex GoL simulation where intelligence could involve, it's how to identify whether any portion, and which portion, of the simulation is intelligent. $\endgroup$ – Samthere Dec 2 '16 at 16:37
  • $\begingroup$ @Samthere I specifically addressed your concern in the last paragraph. Please don't downvote me over things I didn't actually get wrong. $\endgroup$ – Mason Wheeler Dec 2 '16 at 16:39
  • 3
    $\begingroup$ I've downvoted because I think how you addressed it is simply wrong and therefore doesn't answer the question well, not because you didn't address it. A deterministic system appearing nondeterministic is simply a way of saying it's complex. Also, you can have a Game of Life appear complex if you can't see the internal workings. As you defined yourself, it appears to an "outside observer", but you're saying GoL can't appear that way because you're an "inside observer". $\endgroup$ – Samthere Dec 2 '16 at 16:43
  • 1
    $\begingroup$ As a further point, assume I implemented a complicated chat bot website that was good enough for you to believe it to be intelligent. Does it matter that the backend, which is invisible to you, is run on a very complex GoL simulation? Does it matter that it's deterministic? Your answer confuses intelligence requiring "appearing nondeterministic" with "actually being nondeterministic". $\endgroup$ – Samthere Dec 2 '16 at 16:45
  • 2
    $\begingroup$ Take a look in Bell's Theorem about if the world is truly nondeterministic (quantum theory) or if world is simply deterministic at fundamental levels so small that we don't have good models for them yet (hidden variable theory). And there were the Bell experiments to try answer this question. You might like. =D. $\endgroup$ – Physicist137 Dec 2 '16 at 17:08
0
$\begingroup$

Intelligent beings would be present right at the start of the simulation, albeit in a non-local form. This follows from the strong AI assumption which says that all computations that renders an intelligence are equivalent; you would have the same consciousness if your brain were replaced by a machine that would simulate your brain perfectly. Suppose then that one could observe by some means that an intelligence has appeared in a CA at time step T. This means that the evolution of the CA from T to T + 1 is such that a local patch of the CA behaves in a way that we would consider to have intelligent presence.

This may involve having to consider perturbing that local patch to see if counterfactual initial states at time T is mapped to a corresponding counterfactual state at T + 1 in the right way. What matters is that a local patch has appeared that acts as some program that takes input from its surroundings, processes that in some way leading to some output. The way the output is related to the input defines the program.

The intelligence in exactly the state at time T then also exists at earlier or later times T', because the CA rules define a mapping from T' to T and back. So, you can use the CA rules themselves to "see" that the intelligence at time T was already present at the time of the Big Bang. This is also true for us: see this article on eternalism and this article on the Rietdijk–Putnam argument.

One can go a step further and argue that running the CA actually doesn't matter. The intelligence would exist anyway; the CA only allows you to see it. The intelligence should be considered as existing in its own universe defined by its own algorithm.

$\endgroup$
0
$\begingroup$

Pattern detection

Life, in it's basic form, is based around the natural selection of things. What exists today is what survived, either specifically or by replication from yesterday. This thing that survives could be thought of as information: Be it DNA, RNA and so on. When there is no life there is no order, no patterns in the matter that emerge from the chaos repetitively.

But life doesn't imply intelligence. What does?

Mental patterns

This is more complex than pattern detection, because it's about finding pattern (or memes) in different hardware (brains). If you see an object with some internal language, and in this language you see those patterns, that may be duplicated in other similar objects, you probably have intelligence.

I think it's obvious that this is a necessary condition for intelligence: A thought without some kind of "words" doesn't seem possible. I argue that it's a sufficient condition. There's an ongoing philosophical debate on it, but assuming you don't think intelligence requires any sort of soul, I think this will do.

Finding the internal language of the object and translating the patterns to the languages of the other objects is an enormous challenge. It's a challenge to even detect the objects. But luckily the computer are magical :)

$\endgroup$
  • $\begingroup$ The memory and speed of the computer are magic, so that we can imagine simulating a system as large as our own universe (or larger). But the algorithms run on this magic hardware should not be that much more advanced than what we (humans) have access to. $\endgroup$ – Charles Staats Dec 6 '16 at 1:54
  • $\begingroup$ @CharlesStaats you're right, I edited the answer. It's actually not that complicated if you have unlimited computing power $\endgroup$ – Dotan Dec 6 '16 at 12:35

protected by Serban Tanasa Dec 4 '16 at 18:41

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.