Rohin Shah comments on (notes on) Policy Desiderata for Superintelligent AI: A Vector Field Approach