I’m currently making my way through David Chalmers’ new book: Reality+: Virtual Worlds and the Problems of Philosophy. Chalmers explores the simulation hypothesis, using it as a portal into a wide ranging selection of philosophical issues, including discussions on God, information theory, structuralism, and a lot of other topics I haven’t gotten to yet. His overall thesis is that we can’t be sure we’re not in a simulation, and even if we are, we should regard everything in it as real.
Chalmers, showing his usual skill at describing and categorizing the conceptual landscape, discusses different types of simulations.
He uses the word “biosim” to describe a biological being that is in a simulation, with Neo in The Matrix as an example. Neo is a biological human hooked up to a system that provides a simulated world. This is in contrast to machine characters like Agent Smith or the Oracle, who are “pure sims”, completely simulated beings.
Simulations that contain only pure sims are “pure simulations.” While those that include biosims are “impure simulations.” There are also “mixed simulations” with a mix of biosims and pure sims. The matrix is a mixed simulation since it has both biological humans and AI entities.
He describes a number of other categories. A “global simulation” simulates the entire universe, while a “local” one only simulates some portion of it, like maybe a city. (The simulation in the movie, The Thirteenth Floor, is a local simulation, one an inhabitant can reach the edge of if they attempt to leave town.) “Temporary simulations” are ones where people enter and leave them, while “permanent ones” are lifelong simulations. “Perfect simulations” reproduce the laws of physics completely, while “imperfect” ones cut corners in some fashion.
That last category is important, because it gets to whether there would be any way for us to detect that we’re in a simulation. If the simulation is imperfect, then that gives us something to find, some way to discover clues that we’re in the simulation. The matrix, for example, is pretty imperfect (green tint, noticeable time jumps, etc), leaving lots of clues for people to realize that something’s not quite right with the world, ultimately culminating for some of them in the red pill vs blue pill decision.
Chalmers discusses the idea that a perfect simulation is, in principle, undetectable. If perfect simulations are possible, then, the argument goes, there probably are or will be lots of them, meaning that from a probability standpoint, we’re much more likely to be in a simulation than any base reality. This is pretty much the Nick Bostrom argument, and if we buy it, it’s hard to avoid concluding we’re in a simulation.
But I find the idea of a “perfect” simulation strange and in need of much stronger justification. It doesn’t seem like anything engineered is ever truly perfect, much less the idea of a simulation perfectly representing its host universe in any kind of usable time frames. If perfect simulations aren’t possible (or practical), then it seems to undercut all the subsequent reasoning based on them.
On the other hand, a global simulation of a universe that is simpler than the host universe (think Conway’s game of life) might effectively count as “perfect”, at least in the sense of there being no way for the inhabitants to detect they’re in it. In that case, the simulation isn’t representing something else, but is simply the results of seeing how certain laws or initial states play out, and any inhabitants are a result of those laws and initial states. There may be no way to detect such a reality is simulated.
And I don’t know that a simulation has to be perfect to be undetectable, at least not for pure sims. Remember, pure sims are themselves completely simulated, which means the simulation controls their thoughts, perceptions, and memories. Such a simulation could essentially make its pure sims incapable of noticing any imperfections. If we are pure sims, we could be coming across imperfections all the time and simply be incapable of noticing them, realizing what they mean, or remembering it if we did.
Even if we can notice an imperfection, how do we recognize that it is an imperfection? Is our inability to reconcile general relativity and quantum mechanics, or resolve the discrepancies in measurements of the Hubble constant, or solve any of a number of other scientific conundrums, because we don’t understand things yet, or the result of the simulation cutting corners?
So while I disagree with some of his reasoning, I agree with Chalmers that the simulation hypothesis is more difficult to dismiss than most of us would like. But I do think the more parsimonious conclusion, at least currently, is that we’re not in a simulation. A reality, as revealed by our observations and theories, seems more parsimonious than the same reality that is in fact a simulation in a host universe.
It turns out that this is similar to an argument made by Bertrand Russell. Chalmers argues that it doesn’t hold as well if we have reason to believe there are many perfect simulations. But as noted above, I find the idea of a perfect simulation dubious. However given the caveats I mentioned, this isn’t nearly as comforting and categorical a conclusion as the quick dismissals many want to make on this question.
I do agree with Chalmers that if it turns out that we are in a simulation, then, because it appears to exact painful consequences for not taking it seriously, the simulation is effectively our reality.
What do you think? Am I hasty in dismissing perfect simulations? If so, what are reasons to regard them as plausible? Or are there reasons I’m overlooking to summarily reject this whole simulation business entirely?