Ok, I’ll bite. Why should CFAR exist? Rationality training is not so obviously useful that an entire org needs to exist to support it; especially now that you’ve iterated so heavily on the curriculum, why not dissolve CFAR and merge back into (e.g.) MIRI and just reuse the work to train new MIRI staff?
even more true if CFAR is effective recruitment for MIRI, but merging back in would allow you to separately optimize for that.
Ok, I’ll bite. Why should CFAR exist? Rationality training is not so obviously useful that an entire org needs to exist to support it; especially now that you’ve iterated so heavily on the curriculum, why not dissolve CFAR and merge back into (e.g.) MIRI and just reuse the work to train new MIRI staff?
even more true if CFAR is effective recruitment for MIRI, but merging back in would allow you to separately optimize for that.
It’s at least as important for CFAR to train people who end up at OpenAI, Deepmind, FHI, etc.
I’m sure the methods of CFAR have wider application than to Machine Learning...