Right, but the simulation is nonetheless bottlenecked by whichever system requires the greatest fidelity, and the more technology advances, the more of a problem that becomes. For instance, medieval times would likely be far easier to simulate than modern times, because the latter requires simulating our entire computer infrastructure.
And I think that infrastructure is harder to simulate than you'd think: I could use a solver on a big NP problem (we'll assume that P != NP), get a solution after an hour, and in theory, with some practice, I could probably check if the answer is correct in my head. So the simulator can't simply give me what I "expect". It has to actually compute the thing, and then it's clear that the faster our computers get, the slower the simulation has to run.
Alternatively, the simulation could mess with our minds so we never notice anything out of place, but at that point I'm not sure I understand the point of it. Might as well wonder if this is all a dream.
Agreed.
> because the latter requires simulating our entire computer infrastructure.
Maybe, that isn't clear. There are probably plenty of optimisations here too if given some thought.
> I could use a solver on a big NP problem (we'll assume that P != NP), get a solution after an hour, and in theory, with some practice, I could probably check if the answer is correct in my head.
Yes, but note that we very rarely solve NP or EXPTIME problems exactly due to the costs. We often solve them heuristically or approximately, which wouldn't pose a problem for a simulation either.
Then there's also the possibility that we are simply not free to choose the problem to solve. When running a solver for an NP problem, we just need to input any kind of NP problem, and a simulation could easily have large sets of precomputed solutions available.
> Alternatively, the simulation could mess with our minds so we never notice anything out of place, but at that point I'm not sure I understand the point of it
Depends on whether any such changes affects the point of the simulation. If the simulation is to test world-scale economic models, then isolated tribes wouldn't have much influence on those outcomes.
Then again, maybe the point is simply entertainment. Maybe we're just The Sims for post-humans, in which case there's no point anyway.
Idle curiosity is drives a lot of human behaviour, particularly in philosophy!
No, they are saying the opposite. The argument that simulating the universe requires more atoms than the universe says that a later civilization would not simulate the entire universe. IE, #2 of the refutations really true.
The simulation only needs to produce observations that are consistent with the knowledge of the first observer. Sometimes bit even that, as I describe in the blog post, because eyewitness testimony is known to be quite unreliable.
I'm not sure what sort of history you're thinking of specifically.
The amount of information science could infer on many questions is strictly bounded and in those cases we could only reason stochastically. The data presented at time of first observation can then be generated randomly from the set of answers consistent with what's already known.
The place to implement an optimization like this would be at the items with the biggest O in your world, which is usually the smallest building block--what you'd have the "most" of, which would drive the largest memory and processing demands.
I think you've expressed a number of confusions.
First, I think you contradicted yourself. The line you quote says that posthuman civilizations are unlikely to create simulations, but you say this is false because a universe simulation requires more power than available in the universe. So you're agreeing with the outcome while saying you're disagreeing.
Second, I suggest reading the the paper fully, because Bostrom explains that we don't need full universe simulations, we need only consciousness simulations (kind of like the Matrix). The very premise of a post-human civilization is that they have knowledge sufficiently advanced that they have algorithms to simulate human minds.
Much like how video games only render the part of the world that is visible to the players, so a consciousness simulation only needs to simulate minds and their perceptions of a macroscopic, classical world, they do not have to simulate a full quantum universe. Our brains are great at filling in information that we expect to be there, so even the parts that we directly perceivedon't need to be simulated with complete fidelity.
Frankly, I don't think you've given the argument sufficient thought, but by a happy accident you picked exactly the outcome that I think is most likely, and I elaborate on why here:
https://higherlogics.blogspot.com/2021/02/why-we-are-likely-...