Hacker News

The simulation wouldn't actually include all those billions of stars and galaxies. In this kind of simulation, it's all just a skybox. If you point a telescope at a distant star, then the simulation spins up a sub-simulation to produce what output a real star would experience in that situation. You include the stars and galaxies because the real universe contains them, and you want your simulation to be realistic.

It actually does make sense that the simulation would be created for us. Compare the level of computing power required to create a limited human-centric simulation to trying to simulate an entire universe. The whole-universe simulation would be some absurd power of ten times more difficult to simulate. If it is possible to simulate a world convincingly, then it's reasonable to assume that there are many, many more low-res human-scale simulations than giant full universe simulations.

With the whole universe simulation, you also have the problem of scale. It's hard to imagine that an entire universe can be simulated at the atomic level with anything less complex than the universe itself. Unless you have a universe-sized computer, you probably aren't simulating an entire universe with atomic accuracy. A human-scale simulation could be performed in a universe no more complex or different from the one we observe. An atomic-scale universe-spanning simulation would have to be run from a higher level that has completely different physical laws than the one we inhabit. Occams' razor applies. If you want to assume a simulation, it makes sense to assume a type that requires the least exotic higher-order "real" universe possible.