Richard_Ngo comments on Reward is not the optimization target