Steven Byrnes comments on What are the high-level approaches to AI alignment?