Ah, yes, just what I like to see from an AI safety researcher.