Brain in a Vat and Computers

[Minty7]

So, guys.

You all know the classical skeptical problem that many people refer to as the "brain in a vat" problem (or BIV). If you don't, the idea is that we could all be brains in vats, just stimulated to perceive a reality that actually doesn't exist by some kind of computer simulation. It was most famously expressed by Descartes in his Meditations on First Philosophy, where he explained the idea of the Evil Genius, basically the idea that it's possible that only you exist and this Evil Genius is tricking you into dreaming a false reality. Now we tend to express the idea as the BIV problem, because after learning a lot about the universe we realized it's not really possible under any circumstances to exist without having a brain. But the idea of BIV's tends to hinge on a computer existing to stimulate our brains into being in this virtual reality.

So my idea is this: If we are BIV's, then we are probably inside a computer-created virtual reality. If this is true, then we can relatively safely assume that this computer uses some kind of a code to run the program to create the virtual reality. So, if we were to somehow make the computer compute something that disrupted its normal proceses, then we might crash it and all wake up. Therefore, what we should do is try to get the computer to compute something disruptive and make it crash the system. If we devoted a lot of time and effort to this, but nothing happened, then we would be able to say that we are probably not all brains in vats. (This idea also works for the possibility that we are in a computer simulation of the universe built by our future selves.)

Since the computers could (and in fact probably would) run some kind of unknown code, we would probably have to make it delete a directory or something using only hex code, or possibly even binary. I don't know much about coding, so I came to post this here as an interesting problem to you guys who know about this kind of stuff. The thing is, how do we make it process something as a 1 or a 0? We could make a sculpture that read out a big binary string that we thought could crash it, but it probably computes each individual particle as a certain combination of variables that describe its nature and momentum. So then how do we make it screw that up? It's very interesting, so I wanted to pose this question to you guys to see if you could come up with anything.

Looking forward to your ideas!

[gmalivuk]

Interesting question, but I moved it to Fictional Science because I think the discussion is going to be pretty purely speculative without much in the way of actual testable hypotheses.

[scarecrovv]

...So, if we were to somehow make the computer compute something that disrupted its normal proceses, then we might crash it and all wake up die. Therefore, what we should do...

This outcome seems equally likely, if not more so. However, if we assume that we are (or at least, if I assume that I am) living in a computer simulation, and we/I decide that crashing it is a desirable outcome, the first step is to consider the nature of the simulation. Is the computer computing every wave-function of every particle in the universe once every Planck time? Is it more like the Matrix, where the computer simulation runs at the level of everyday objects and people, so that people and animals could be controlled by AIs, and a desk is represented as a desk, and not a bunch of atoms? Is the computer really simulating the rest of the universe at all, or are the stars and galaxies just painted on the wall, and appear indistinguishable from actual stars and galaxies, complete with simulations of redshift and all the rest built right in?

The answer to these questions determine the likely points of attack for crashing the simulation, with the possibilities getting easier and more accessible as the computer takes more and more shortcuts to avoid simulating every particle in the universe. For example, if the computer is, in fact, simulating everything down to the last detail, then crashing it would seem stupendously difficult, since the results of practically any action we could take are determined by laws that have been running quite handily since at least the Big Bang, and have likely seen quite a lot of stress testing during that time. The only way to expose bugs that haven't popped up before is to create conditions that have not occurred before, likely involving energies much higher than those experienced in the Big Bang. So building a particle accelerator powered by a dyson sphere around a supernova seems like a plausible plan, if anything were going to work at all. But I wouldn't count on it.

However, if the computer simulation really only cares about us, they might well have decided to throw away the 99.999...% of the universe we'll probably never get to visit, and are just simulating our solar system. In this scenario, only our solar system really exists, and the rest of the universe is just smoke and mirrors. If this is so, perhaps all we need to do to crash the computer is send a probe (or actually travel) to another star. Or another galaxy, depending on how far the author of the simulation thought we'd get. This assumes, of course, that the simulation doesn't expand whenever we visit new places, which is a distinct possibility.

Things get even more interesting when we consider that the computer might not be simulating quantum physics at all, but is simulating objects at the human scale, and only fudging the results of all our quantum physics experiments, or something like that. If this is so, the only limit for how to crash the universe is your imagination. Maybe by capturing Agent Smith and brainwashing him we could cause a segmentation fault in his AI, which brings down the computer. Perhaps if we exactly collide 3 bullets in mid-air all at once we'll break the assumption in the Bullet.Collide() method that only two bullets would ever hit each other simultaneously, causing the program to complain about an incorrect number of arguments, and die. Maybe by committing genocide on a vast scale, I could tease out a bug that makes the program crash when numHumans <= 1. The possibilities are endless. Who could possibly know where to begin?

[Minty7]

Maybe by committing genocide on a vast scale, I could tease out a bug that makes the program crash when numHumans <= 1. The possibilities are endless. Who could possibly know where to begin?

Well, there's only one way to find that out.

But seriously, I think it's probable that we would die if we were to crash our theoretical virtual reality. There's a chance we wouldn't, because we will have been surviving in the same conditions since we were "born" and the only thing that would change would be our sensory reception. But that in itself would probably kill us, because it's a pretty huge shock to suddenly have no body or sensory organs.

But the problem of objects vs. atoms is interesting. I would be inclined to say that the computer would calculate things based on the properties of each individual particle, because we've observed phenomena associated with single-particle objects. We haven't seen atoms, but we have reason to believe that they exist in our reality. That makes our job harder, but there's still the possibility that everything is run on an object basis. And I would say that the computer probably wouldn't be computing only our solar system, because we've launched the Pioneer and Voyager missions pretty far out into space. Of course, we haven't yet reached another star, but it should crash pretty soon if that was the case.

If things are calculated based on an object basis, then I think the sculpture of a disruptive string of bits would work. That would be the easiest thing to do.

And then there's the problem of computing power. If you want to simulate the universe down to the quantum level, would you need a universe-sized computer? Regardless, even if you wouldn't, you would probably only want to simulate the solar system, and even that would be a massive undertaking. So in that case, we probably only need to wait for the Voyager and Pioneer probes to get out of the solar system.

Very interesting arguments about this. It feels good to stretch your theoretical muscles.

[Jplus]

Maybe I'm simplifying things too much, but why not just do something that is deemed impossible by the simulator?

[Minty7]

Maybe I'm simplifying things too much, but why not just do something that is deemed impossible by the simulator?

That would probably work, but what would you say is impossible? What the simulator deems to be impossible is probably what we observe to be physically impossible, which means that we would have to do something like violate causality or make something with significant mass exceed the speed of light. If we did that, the simulation would most likely crash, because it's not programmed to let something like that happen. But we've been trying to do things like that for years now. It's possible that the simulator considers something impossible that is actually not impossible to us, but God know what that could be.

[Xanthir]

Your idea of what constitutes a "computer" is quaint. Almost adorable, even.

Most simulationists I know of assume that the simulation is probably running as an incidental part of some other process that a godlike intelligence is running. For example, the intelligence might be attempting to solve some problem via Solomonoff Induction (for example, you can predict the future by simulating all possible pasts, up to some complexity bound). Our universe could just be one iteration of that algorithm. It doesn't even need to officially be a "universe" - assuming physics is computable, our universe is equivalent to some Turing Machine, and so a simple automated theorem-prover may eventually run our universe in an attempt to find a proof to some statement. There's a lot you can do when "iterate over all turing machines" is a possible strategy. Hell, if we ever uncover a mechanism for hypercomputation, *we* might very well end up doing that (there are somewhat disturbing moral implications, of course).

My point is that it's highly unlikely we're literally a brain being fed simulated information in an attempt to trick us, and much more likely (relatively) that our entire universe, including our brain, is being simulated.

[Ptharien's Flame]

You all seem to be assuming that the simulation has left itself open! I agree its not a very interesting question if that's not true, but I'd like to point out that I've often had to actually work to put a flaw in my small simulations (don't ask why I needed to) as opposed to just restricting them enough so that they cannot be "broken" this way. For example, why would the "computer" calculate the result of a three-bullet collision as a call to Bullet.collide(b1,b2,b3)? Isn't it significantly more likely for it to calculate b1.collide(b2).collide(b3)? If it were at all designed properly, as it certainly seems to be, the category of possible entities (objects) in our universe should be fully closed under all possible actions (morphisms)? Or am I not thinking about this the right way?