if we’re currently in a simulation, physics and even logic could be entirely different from what they appear to be.
I have another obscure shortform about this! Physical vs metaphysical contingency, about what it would mean for metaphysics (e.g. logic) itself to have been different. (In the case of simulations, it could only be different in a way still capable of containing our metaphysics as a special case, like how in math a more expressive formal system can contain a less expressive one, but not the reverse)
I agree a metaphysically different base world is possible, but I’m not sure how to reason about it. (I think apparent metaphysical paradoxes are some evidence for it, though we might also just be temporarily confused about metaphysics)
Just physics being different is easier to imagine. For example, it could be that the base world is small, and it contains exactly one alien civilization running a simulation in which we appear to observe a large world. But if the base world is small, arguments for simulations which rely on the vastness of the world, like Bostrom’s, would no longer hold. And at that point there doesn’t seem much reason to expect it, at least for any individual small world.[1] Though it could also be that the base world is large and physically different, and we’re in a simulation where we appear to observe a different large world.
Ultimately, while it could be true that there are 0 unsimulated copies of us, still we can have the best impact in the possibilities where there are at least one.[2]
By the way, I’m also somewhat skeptical of a couple of your assumptions in Mutual Anthropic Capture. Still, I think it’s a good idea overall, and some subtle modifications to the idea would probably make logically sound. I won’t bother you about those small issues here, though
I’m interested in what they are, I wouldn’t be bothered (if you meant that literally). If you want you can reply about it here or on the original thread.
If we’re instead reasoning over the space of all possible mathematical worlds which are ‘small’ compared to what our observations look like they suggest, then we’d be reasoning about very many individual small worlds (which basically reintroduces the ‘there are very many contexts which could choose to simulate us’ premise). Some of those small math-worlds will probably run simulations (for example, if some have beings which want to manipulate “the most probable environment” of an AI in a larger mathematical world, to influence that larger math-world)
In other words: “Conditional on {some singular ‘real world’ that is somehow special compared to merely mathematical worlds} being small, it probably doesn’t contain simulations. But there are certainly many math-worlds that do, because the space of math-worlds is so vast (to the point that some small math-worlds would randomly contain a simulation as part of their starting condition)”
And there’s probably not anything we can do to change our situation in case of possibilities where we don’t exist in base reality. Although I do think ‘look for bugs’ is something an aligned ASI would want to try, especially when considering that our physics apparently has some simple governing laws, i.e. may have a pretty short program length[3], and it’s plausible for a process we’d describe with a short program length to naturally / randomly occur as a process of physical interaction in a much larger base world—that is to say, there are plausible origins of a simulation which don’t involve a superintelligent programmer ensuring there are no edge cases)
I have another obscure shortform about this! Physical vs metaphysical contingency, about what it would mean for metaphysics (e.g. logic) itself to have been different. (In the case of simulations, it could only be different in a way still capable of containing our metaphysics as a special case, like how in math a more expressive formal system can contain a less expressive one, but not the reverse)
I agree a metaphysically different base world is possible, but I’m not sure how to reason about it. (I think apparent metaphysical paradoxes are some evidence for it, though we might also just be temporarily confused about metaphysics)
Just physics being different is easier to imagine. For example, it could be that the base world is small, and it contains exactly one alien civilization running a simulation in which we appear to observe a large world. But if the base world is small, arguments for simulations which rely on the vastness of the world, like Bostrom’s, would no longer hold. And at that point there doesn’t seem much reason to expect it, at least for any individual small world.[1] Though it could also be that the base world is large and physically different, and we’re in a simulation where we appear to observe a different large world.
Ultimately, while it could be true that there are 0 unsimulated copies of us, still we can have the best impact in the possibilities where there are at least one.[2]
I’m interested in what they are, I wouldn’t be bothered (if you meant that literally). If you want you can reply about it here or on the original thread.
If we’re instead reasoning over the space of all possible mathematical worlds which are ‘small’ compared to what our observations look like they suggest, then we’d be reasoning about very many individual small worlds (which basically reintroduces the ‘there are very many contexts which could choose to simulate us’ premise). Some of those small math-worlds will probably run simulations (for example, if some have beings which want to manipulate “the most probable environment” of an AI in a larger mathematical world, to influence that larger math-world)
In other words: “Conditional on {some singular ‘real world’ that is somehow special compared to merely mathematical worlds} being small, it probably doesn’t contain simulations. But there are certainly many math-worlds that do, because the space of math-worlds is so vast (to the point that some small math-worlds would randomly contain a simulation as part of their starting condition)”
And there’s probably not anything we can do to change our situation in case of possibilities where we don’t exist in base reality. Although I do think ‘look for bugs’ is something an aligned ASI would want to try, especially when considering that our physics apparently has some simple governing laws, i.e. may have a pretty short program length[3], and it’s plausible for a process we’d describe with a short program length to naturally / randomly occur as a process of physical interaction in a much larger base world—that is to say, there are plausible origins of a simulation which don’t involve a superintelligent programmer ensuring there are no edge cases)
(but no longer short when considering its very complex starting state? ig it could turn out that that itself is predicted by some simple rule)