One thing I never understood is why - given the number of different things we could use it for - a supposed future civilization would pick "run simulations of people who are now dead" as "the great task". It seems there's a classic mixup of model vs reality there - you don't learn much from a faulty model, and you have no real way to validate it. Nor do you learn much from going infinite monkeys on it. And trying to sim all variations on the assumption one must be "right". What's the goal?