Can Go AIs be adversarially robust?