logoalt Hacker News

jstanley • today at 4:35 PM • 1 reply • view on HN

Your example is the difference between being aware of a question and knowing its answer.

LLMs know the answers to questions we were already aware of, and we can tell they're right, but we can't understand their working. That's what's new.


Replies

twoodfin • today at 4:49 PM

In this case, Claude has told us everything it “knows” (so far) and it’s an unintelligible (for now) Lean proof. I don’t believe it has some latent understanding at a higher conceptual level that’s remaining obscure.

If a team of 100 top-flight number theorists working together had used much less sophisticated LLMs + Lean to produce this proof, we’d be just where we are today.

➕ show 1 reply