Vlad Mikulik comments on Risks from Learned Optimization: Introduction