how would the impacts of AI change if the rule of "not harming" is changed to "helping humanity'?