The stochastic parrots have apparently proved the Riemann Hypothesis
Can you share sources?
My son has tried to explain this to me - he is the mathematician in the family - and what the LLM produced is a load of source files for an automatic proofer. So far, the files are syntactically correct, but if you know programming, getting a project through the compiler without error messages is no way a guarantee for a working software.
And indeed, my son told me just a few hours after release that human reviewers had already found tautology issues within the proof.
But don’t ask me for details - my time with math at that level was ages ago…
The point of Lean (or any other similar proof assistant) is that if the program typechecks, then you have a proof. Hence, supposing that no Lean bug were exploited, all we would need to check is that the statement of the theorem in Lean matches what we actually want to prove. If this is the case, then we have a witness of the existence of a proof, despite it being a gigantic messy heap of definitions and lemmas.
An actual human-digestible proof is still valuable, because the proof itself would contain insights valuable for mathematics in general.
Nothing LLMs have done has been any real advancement.
…is something I can say about everything I’ve heard LLMs used for except some of the mathematics-related things.
The Navier-Stokes thing by OpenAI a) wasn’t a proof of the Navier-Stokes problem, but a lesser thing that, nonetheless, is supposedly an advancement, b) has discrepancies between the Lean proof and the human-readable version, and c) wasn’t by OpenAI. OpenAI stole credit as a marketing stunt. (And so they could offer the real mathematician an opportunity to be listed as coauthor on the condition that he excluded his colleague who helped because the colleague worked for Anthropic.) But from what I understand, even the guy who deserved the credit intended to use LLMs to help with the proof. Only thing I don’t know is whether LLMs were important/necessary/helpful to the process in some way or whether the LLM wasn’t really necessary to it. Of course all the media I’ve been able to find on the topic implies vaguely that LLMs are absolutely necessary to the proof that was found but without actually saying anything concrete that indicates that’s actually the case.
I have the same questions about this claim of a Riemann Hypothesis proof. It really seems implausible to me that advanced math proofs are the one and only actual use case for LLMs. Every other claim that they’re some kind of advancement has been a total grift. So what really are the chances that the mathematical proofs are actually legit. But meanwhile this Riemann Hypothesis proof is even fresher and the method they used is even less well reported than the Navier-Stokes one.
I dunno. We’ll see what future reports indicate. I’m still not convinced the first one even proves what they claim it does. Even the lesser thing that isn’t Navier-Stokes itself.
Nope, they haven’t. The digital parrot balked up a gazillion files in that regard, but a) this is not a real proof, and b) human reviewers have already found shit in them.