Rob Bensinger comments on Naive Hypotheses on AI Alignment