Noa Nabeshima comments on Clarifying some key hypotheses in AI alignment