Self-modification is the correct justification for updateless decision theory — AI Alignment Forum