Charlie Steiner comments on Which counterfactuals should an AI follow?