Tuesday, January 10, 2017

Why we don't live in a simulated universe

There's an oft-repeated theory that our universe is a simulation -- a computer program (or equivalent) created by some higher form of intelligence. The modern popular form of this claim is generally attributed to philosopher Nick Bostrom, although the basic idea goes back much further.

And the only problem with this idea is that it is completely wrong.

The basic "simulation argument" goes like this: imagine that an intelligent race becomes intelligent enough and powerful enough that they could simulate a universe in a computer. (We already do this ourselves, in a very crude sense, when we create computer models to simulate weather or traffic or any other aspect of the real world.) Our hypothetical aliens are able to build such rich simulations, they actually contain simulated intelligent beings of their own -- and those simulated beings would (somehow) perceive themselves to be conscious.

And here's the clever twist: according to the argument, those simulated beings could become intelligent enough to build their own simulated universes, with simulated intelligent beings of their own, who in turn... Eventually, there would be an enormously large pyramid of simulations-within-simulations. And from a simple probabilistic perspective, it's enormously unlikely that we happen to be in the topmost and only real universe (and sometime in the future will ourselves start simulating universes) rather than one of the vast number of simulations.

And this is completely mistaken.

The problem with the argument is that the universe we find ourselves in is enormously complicated from the point of view of having intelligent beings in it. For a start, you could discard the other one hundred billion galaxies in our observable universe and it wouldn't make any difference to us. So it's enormously more likely that the simulated universe we are in would be much simpler than this one. (How much more likely? Borrowing an argument from Roger Penrose, possibly something of the order of 10 to the power [10 to the power 100] -- a 1 with [10 to the power 100] zeroes after it -- more likely.)

So the simulation argument turns on itself: the exact same argument that leads to the conclusion that we live in a simulation, i.e. that there are many more simulations than real universes, also inevitably leads to the conclusion that this universe isn't simulated, because there would be hugely many more simpler simulations we would be more likely to find ourselves in.

5 comments:

Unknown said...

Other than the problem of needing sufficient computing capacity to handle the enormous universe we see, I'm not sure I get your argument. If reductionism is true, then all this complexity we see reduces down to an extremely simple, elegant set of formulae. It hardly matters complexity-wise how many galaxies developed, or whether intelligence resulted.

Do you really mean complexity, or do you actually mean quantity?

Until we have some evidence that we are in a simulation, which is likely impossible unless the programmer made an error, I agree this is all philosophical speculation, not physics. But I don't understand how your argument sheds any light on the matter.

The Czar Dictates said...

Well, I had to simplify a great deal to get this down to the size of a blog post rather than a book :-)But to (try to) clarify a little... when I talk about complexity I'm really referring to entropy. And the entropy of our universe depends not only on the simple, elegant forumulae (plus of course, in our current limited theory, a bunch of apparently arbitrary constants); it depends overwhelmingly on the starting conditions.

Penrose argues eloquently that the entropy of our particular universe is very peculiar, and goes to show how neither inflation nor multiverse theory is a satisfactory explanation for this. (I really can't do his argument justice, I'm afraid). I simply realized that the "simulations all the way down" philosophical speculation is really just another flavor of multiverse, which falls for the same reasons. (And yes, I realize that risks becoming an appeal to authority: "because Penrose said so"; but it's the best I can do in a brief blog comment).

So why didn't I just say "entropy" instead of "complexity" in the first place? Basically, because the average person really doesn't grok entropy, especially in the cosmological sense, and I worried that it would send people on the wrong mental track.

So basically, if you are reading / have read Penrose, you can mentally substitute "pyramid of simulations" where he says "multiverse" in chapter 3; or you can treat this whole blog post as a bit of throwaway fun :-)

Unknown said...

I guess I should read your Penrose reference. I thought entropy was a measure of disorder, not complexity. I suppose those two concepts are related, but I can think of highly disordered systems that are not complex (cosmic microwave background radiation for example). Am I confused?

I confess I'm interested in the discussion because the simulated universe speculation actually makes a lot of sense to me. Of course, I'm a software engineer by trade, so I suppose it would. :) But when I look at things like granularity at the Planck length, all I can think is "pixels". When I see all the elaborate mathematical objects that we're trying to use to model behavior of systems, all I can think is "algorithms". In principle, there is no reason that a programmer could not create a simulation that models our universe to the last decimal point, if only we knew the exact physical laws (algorithms) and had sufficient capacity for all the data and computations. No multiverse necessary, and all the fine tuning is exactly that - fine tuning done by the programmer to put bounds on the sim. Makes a certain sense, even if completely impossible to prove and useless scientifically.

The Czar Dictates said...

Thinking of entropy as a measure of disorder works fine in most circumstances... except cosmologically. Penrose really opened my eyes to how to think properly about entropy. For example, the CMB is high entropy because it is "disordered" i.e. almost entirely "evenly and randomly spread out" -- but Black Holes have the highest possible entropy for a given volume of space because they are as compressed as possible. Penrose spends a lot of time on this dramatic contrast between the thermodynamic entropy and gravitational entropy of our universe.

As a former software engineer and mathematician myself I understand the appeal of the speculation. I'm also fascinated by the apparent complexity of physical laws. For much of fundamental physics, the exact equations (as far as we know) are intractable (at our current level of computation) and the best we have are perturbative approximations. Our best simulations of particle collisions are laughably small. And yet somehow the universe manages to "compute" the right answers. To me that suggests one of two things: either the real laws of the universe are much simpler than we think (which would make us like pre-Copernican astronomers, achieving impressive feats of accuracy at great computational effort through an entirely wrong model of epicycles); or that the universe at that level is fundamentally irreducible, and can't be computed by something smaller than itself, a kind of minimal algorithm so to speak.

One last thought, just for fun: if we are in a sim, it be possible to discover it. If some of the things we speculated about above are true we might be in a sim that has some approximations to make the calculations tractable -- or if our programmer was careless, outright contradictions. Some things might make us suspicious: for example, what if the search for dark matter continues to turn up empty (and alternative explanations don't pan out)? We might begin to wonder if our programmer forgot to divide by 4 in the algorithm for cosmic-scale gravity... I don't know whether it would ever be possible to find an outright contradiction, but it's fun to speculate. Maybe our programmer never expected anything intelligent to develop in his sim, so got lazy at the edges?

Unknown said...

A fun conversation.

The thing is, if we are in a sim, then the rules of the sim are created by the programmmer. She doesn't have to painstakingly discover the laws of physics, go down blind alleys, deal with intractably difficult math, etc. - all because she knows the laws of nature. She wrote them all! So to me the fact that we are still a long way from a theory of everything, which *might* be close to the original programmer's code, really only says that it's easier to engineer than reverse-engineer. :)

Yes, I agree our only hope for empirical evidence is discovering a bug in the software. Something that shows a clear logical inconsistency, something that does not obey the rules, and is irreconcilable with all known rules, in a reproducible way. Maybe that would be enough. However, it might be extremely hard to distinguish such phenomona from mere "new physics".

As I mentioned, fun speculation, probably hopelessly unfalsifiable, and scientifically useless. Until, just maybe, someday, it's not ...

A Brief History of Trumpistan

January 21: A coalition of eighteen states led by Texas announce their succession from the United States, forming a new country reviving the...