Tuesday, October 18, 2016

Reality and Simulation

I read a few weeks ago that Elon Musk, and apparently a lot of other people, think we are very likely living in a simulation.  His argument goes like this:

The strongest argument for us probably being in a simulation, I think, is the following: Forty years ago we had Pong--two rectangles and a dot. That is what games were.
Now 40 years later we have photorealistic, 3D simulations with millions of people playing simultaneously and it's getting better every year. And soon we'll have virtual reality, we'll have augmented reality. If you assume any rate of improvement at all, then the games will become indistinguishable from reality.

Although this seems to be a popular view among the technological elite in San Francisco, there are people who disagree.  Some people think it is more likely that humanity will destroy itself before we get to the level of simulations required.  Another line of argument is that the technology is harder than Musk is giving it credit for.

I tend to agree with the latter criticism.  I am not a connoisseur of VR games, but I am sceptical whether we have anything today that suggests a future where "the games will become indistinguishable from reality."  Musk seems to be committing the basic fallacy of projecting technology in a linear fashion.  He's not literally saying that technology will continue to progress at the same rate -- he even argues that it could slow significantly -- but rather than the developement of better and better simulations will be a matter of simple progress, the way video games have progressed to this point.

There are two things that I would object.  The first is that efficiency tends to drop off rapidly as you approach extremes.  A person projecting the future of travel in 1950 might note that we had gone from horse and buggy to cars to airplanes to jets in a short time.  Why wouldn't we have flying cars now?  Why not travel around the world in minutes rather than hours?  Why not commercial space travel?  Maybe those things will happen eventually, but there are limits to physical laws of weight and inertia that make the leap to air travel more plausible and more efficient than some obvious next steps.  With computers, we may be able to overcome the physical limitations (although I consider that problematic), but the programming requirements seem even larger.  Who is capable of creating a completely realistic world for VR?  Even if we started with a very basic physical universe -- say, the Big Bang -- and let everything evolve from there, so that no one would have to program in all the data points in our present world, there are other issues.  One obvious one is that we don't know all the physical laws of the universe, and any attempt to simulate our universe starting from the very beginning would end up with some very odd results if we got them wrong.  And then, if the simulation really did start with the Big Bang, the chance that it would end up with anything like our present world are virtually zero.  Musk posits billions of VR machines, any one of which we could be in; but even billions of simulations would have virtually no chance of producing us at random.  That even assumes the fact that we could build a machine that could keep track of all the particles in the universe, which would seem to be impossible on the face of it:  how could a machine in the universe have enough memory to keep track of the whole universe?  But anything less than that would imply that someone would have to program in some incredibly detailed parameters to set up our world, in a way that no one could possibly do even if he could somehow access all the data.

You would think that our inability to construct a basic word processor without bugs would be a problem for a completely realistic simulation, but not Musk.

People have come up with other, you know, more exotic ideas, looking for glitches in reality, sort of like in The Matrix when the black cat walks by twice. Remember that scene with Neo? So, looking for mistakes. I don't buy that at all because a very clever error-correcting simulation could simply wipe clean the memory of any such glitch after correcting it.
Maybe a "very clever" error correction algorithm could do that, but I haven't seen any evidence of such a thing in our present world that would lead me to believe that it could happen in a simulation.  Moreover, it's not even clear how the simulation could detect an error.  How would it know what an error means?  We check errors now by doing things like checksums to validate that a file is the same after it has been copied, but a violation of the physical laws of the simulation would imply that the program could keep perfect track of its own rules and note when something violated them.  More likely, it would not notice a violation because it would be due to a bug in the fundamental simulation code.

At base, I think the idea that we're living in a simulation is pretty silly, and I think the evidence Musk gives for it is strikingly weak.  One thing he doesn't address (at least not that I have noticed -- maybe he has elaborated elsewhere) is what exactly we're doing here.  Is he saying that we are "non-player characters" who have been programmed with our own consciousness?  Because that seems about the most implausible of all:  it implies that the programmers have unlocked the key to consciousness to the point where they can not only create a conscious computer, but even conscious elements within a computer. 
Or perhaps we are supposed to be game elements that have evolved consciousness? Again, that seems unlikely. What I thought he meant at first was that we were players in a simulation, physical people who are so absorbed in the game that we don't realize any more that it's a simulation. But how could we possibly become so divorced from our own physical bodies that we could no longer feel them or tend to their needs? Or are we all, like in the matrix, being kept alive by computers while our minds are permanently locked into virtual reality?

There is an old philosophical question that asks how we know we aren't dreaming. I believe a Taoist phrasing of it is, how do we know we're not butterflies dreaming we're humans? Well, it seems pretty unlikely that butterflies could have such a vivid imagination of human lives, but there really doesn't seem to be any way to prove that we aren't humans in a dream state. A more modern formulation is to ask how we know we are not just brains in a vat, being kept alive artificially, all of our perceptions being simulated by some scientist. The simulation hypothesis seems like another variation on this, except for Musk's confidence that his computer simulation theory is extraordinarily likely to be right. I suppose there is no way to disprove it, or any other similar theories, in a definitive fashion. However, I am confident that we will continue to behave as though our physical world is very real, and we will continue to treat those who think it isn't as insane.

No comments:

Post a Comment