Rohin Shah comments on Risks from Learned Optimization: Introduction