Does being smart and innovative require that AIs can never truly, deeply conform to social rules? [In development still]