Notes on Caution — AI Alignment Forum