TurnTrout comments on Towards deconfusing wireheading and reward maximization