i'd be concerned that a process like RLHF is trying to index a relatively infinitesimal part of the network. That could either be done by pruning abilities, or learning cognition which is successful…

Posters: anthrupad. anthrupad

2 likes recorded by the source corpus.

Topics: AI & LLMs, AI safety

View Post: https://x.com/anthrupad/status/1722470215064191148