A months ago headlines read that ChatGPT “solved” an 80 year old Erdős problem.

I found Cal Newport’s take on this to be much more balanced. From what I remember, ChatGPT didn’t “solve” anything. It dumped out a bunch of text, that humans reviewed, and it gave them an idea for how to disprove a thing Erdős thought was true, but couldn’t prove.

https://youtu.be/fhZRWZ6J4k4

In cases like this, it seemed like the LLM got a lot more credit than it deserved. So I’m guessing it would go something like that.

LLMs are increasingly solving math problems on their own, producing entire correct proofs

I like Cal but his takes consistently miss the slope of improvement of the technology