shash42 comments on The Bitter Lesson for AI Safety Research

[shash42]

Thanks, the rationale for using PCA was quite interesting. I also quite like the idea of separating different model classes for this evaluation.