Post

the Dinitz-Garg-Goemans conjecture, a graph theory problem that’s gone unsolved for 30 years, was solved with this prompt

"unsolved for 30 years" makes it sound like hundreds of mathematicians were grinding on this problem constantly for the past three decades like all the others that "AI" has "solved" so far, this one was pretty much sitting on a shelf

I'd also like an actual mathematician to confirm that it's been solved

What, you mean we shouldn't just take the co-founder of an AI startup's word for it? Yes, we need human peer reviewers to confirm that the Lean that was output by the model actually matches the problem.

So it's closer to an item in the jira backlog that has been sitting there forever

brb looking up "jira backlog"

Thank you geeksforgeeks www.geeksforgeeks.org/software-tes...

What I was imagining at first (having read nothing whatsoever about the problem) was more along the lines of Fermat's last theorem or the twin prime conjecture - easy or even trivial to prove over an arbitrarily large set of inputs but impossible to generalize to infinity without a novel approach.

Also, on X The Everything App with a bunch of crypto bros retweeting it for clout should really not be the way that new mathematical proofs are announced.

Agreed, it shouldn't be the way anything's announced really I also find the sheer amount of "yes this has been solved" agreement with no one actually discussing the mathematical process—only the prompts—to be a bit odd

They're boasting about how short the prompts were because it implies that the model is smarter if it was able to reach the answer with less context. But the file upload and the web search results are also part of the prompt.