Twitter thread on AI safety evals

By richard_ngo @ 2024-07-31T00:29 (+38)

This is a linkpost to https://x.com/RichardMCNgo/status/1814049093393723609

This is a crosspost, probably from LessWrong. Try viewing it there.