What would we do if alignment were futile? — LessWrong