Matthew Barnett comments on New paper: Corrigibility with Utility Preservation