Clarifying some key hypotheses in AI alignment — AI Alignment Forum