SummaryBot comments on Giving AIs safe motivations