Another argument against utility-centric alignment paradigms — LessWrong