Moral realism and AI alignment — LessWrong