Must I consider my AI's moral Welfare? I'd really rather not.