“Maxim Bar Kogan believes that as AI models become super-intelligent, they will be able to help humans solve the problem of mechanistic interpretability more effectively.”