TurnTrout comments on Intuitive examples of reward function learning?