1416 points•573 comments•0mo ago•by MaxMussio•seangoedecke.com ↗

LLMs reward expertise

The article argues that domain expertise remains the key factor in using large language models effectively, even as LLMs enable non-experts to generate outputs. It uses Terence Tao’s conversation with ChatGPT on a recently discovered counterexample to the Jacobian Conjecture to illustrate how expert prompting steers the model, yields concise results, and pushes for the right direction. It also notes that human understanding of a problem or codebase often limits the model’s performance, and that AI-safety discussions are part of the broader context.

Discussion (573)

No comments yet

Top
Trending
New
For You
Bookmarks
Dossier