If, as I think you are assuming, reality is a benign simulation, then we're probably safe from the dangers of unaligned AGI. We would also be safe from a lot of other dangers if we're in a benign simulation. And that would be awesome. I think we might be, but it's far from certain. Therefore I would like to solve the alignment problem.