Vikram Saraph @vikramsaraph.com · Jan 7 This is also true of Math StackExchange: data.stackexchange.com/math/query/e... I remember when I first asked ChatGPT to prove a theorem and it did poorly. LLMs have gotten much better at writing correct proofs since then. I have to wonder how math professors deal with this as they assign homework
3 likes 2 replies
Replies
If you believe LLMs compete on quality/capability, then it's easy to nod along when someone says "today's LLMs are the *worst* they will ever be, from now on," but it's very hard to really act as if you believe it's true.
Activity on MathOverflow, which is generally restricted to discussion of research-tier questions, has been more resilient to the overall trend on StackExchange networks, but you can still see a decline in questions asked. Interesting and kind of sad.