Avoiding doomsday: a "proof" of the self-indication assumption — AI Alignment Forum