> What do they mean by "simulate the past"?
It's a simplification of the argument for the sake of making the first two prepositions easier to swallow as possibilities.
Consider that if we can simulate consciousness, then we could conceivably simulate a near infinite set of variations of consciousness.
This would still affect our likelihood of being in a simulation, as we don't have any outside knowledge that lets us know which type of entities we "should" be. It doesn't need to be human-type civilizations simulating their ancestors, it could just as well be some aliens playing a hyper-advanced version of Spore and making disgusting aliens to creep out their friends.
But it is easier to get people to picture e.g. Sim2,000,000 simulating full brains based on actual humans, set in a world based on "our" earth as having relevance to our existence, and it makes it more relevant to speculate on proposition #1 and #2. E.g. is there anything in human type society that will prevent us from getting there or wanting to do it?
> On the other hand, if they mean "simulations of some hypothetical past with loose connections to the actual past, Matrix-style", i.e. "they" have the freedom to generate the scenario we're in, why on earth would they choose this one with all the troubles and suffering in the world? Why not make one where everyone's a pony and we all eat rainbows and poop butterflies?
I don't know about you, but most of my Civilization/FreeCiv/Civ:CTP games ends up with things like nuclear war or brutal conquest - it's a lot more interesting to play the games in extreme directions. It's a good question, and one reason to explore this problem space.
Even so, in some sense it's not quite the right question: The point is that with high values for propositions #1 and #2, we could just happen to be in one of a potentially nearly infinite number of simulated realities. Perhaps they're running vast quantities of them to figure out of different choices would have different outcomes, for example. Or perhaps we're some kids afternoon entertainment.
We don't have a way of telling.
Unless we know that the possible set of "realities" is low, however, it makes little sense to speculate in whether there is anything "special" about "just this" simulation, because we have no basis for determining whether or not there is anything special or if the number of simulations is just so great that "ours" is just one tiny variation in a near endless sequence.
> They are running into exactly the same debates that Christian philosophers have been struggling with for millenniae - why is there evil in the world created by God / our hyperintelligent great^{n}grandchildren? Which indicates to me they are much closer to religion than science.
I'll repeat myself: It's not meant to be science. It's a philosophical argument. Nicholas Bostrom is a philosopher. It's his job to be a philosopher, not to set out scientific theories. Scientific theories could arise from speculation around the choices in question, such as e.g. whether or not there could be possible ways of determining if we're in a simulation, that might yield hypotheses that could be tested and potentially falsified.
But until that time, what separates this argument from religion is that the argument is not making any claims for either possible outcome - it is setting out a set of possibilities and asking questions around what we can learn from it.
As I've said elsewhere, if someone picks one on belief and sticks to it in the face of contradictory evidence, that'd be akin to religion. But asking questions and seeking ways of refining those into testable hypotheses is not religion.