I feel like you’re comparing apples and oranges here.
There is a vast difference between the solution to an abstract mathematical problem, and a concrete real-world problem.
The cure to cancer would have a direct benefit to humanity, cancer would be cured, less people would die.
The answer to “Is the Reimann Hypothesis true?” does not have a direct benefit, especially if the answer is unverifiable (which an opaque insanely large blob of a program and paper most certainly are). The benifit of a solution to that question lie in both the cascading effect of it verifying other mathematical proofs that assume the Reimann Hypothesis is true, and the new insights that are developed in the process of solving it.
I would argue that, yes, it isn’t a big deal.
A mathematical proof does not exist for its answer (or maybe does not solely exist for its answer, depending on the question). It exists for the logical insight and breakthroughs in knowledge that are typically required to answer these famously difficult questions.
An LLM generated 1.4gb Lean project and accompanying 250,000 page paper are completely useless in this regard due to both it’s insane size, and the insanely high ratio of meaningless gaff to actual information that LLMs always generate. So any actual insight that might exist in the proof will almost certainly never be extractable and/or useful to humanity.
Regardless of if it generated a legitimate proof for the Riemann Hypothesis or not, whatever was spat out is less than useless for everyone except maybe Anthropic Shareholders (due to the hype and their name being in the news).
Mathematical proofs are not for their answer, they are for the logical breakthroughs and novel insights that solving a problem like this elicit. The useful extensions to the total collective of human knowledge, that can be reused and applied to other fields to further human development and advancement.
The “proof” that was generated by the LLM is exactly as useful as a napkin with the words “The Riemann Hypothesis is True” written on it. An LLM generated 1.4gb Lean project, and it’s accompanying 250,000 page paper, is so useless to the field of Mathematics that it’s completely insulting to actual insightful work being done. Like the draft paper for an honest human written solution to the Generalised Riemann Hypothesis “The Music Of The Primes” written by Alessandro Maria Masullo who has been working on the paper for over a decade(!) and was less than a year to publication, but had to release it early due to the LLM publicity stunt that is Levent Alpöge/Anthropics “proof”.
3 points on How can a fancy autocomplete / stochastic parrot solve the Riemann Hypothesis? in !asklemmy@lemmy.world ·
It’s definitely possible, there are a disturbing number of times where genuine pre-production work has been leaked by an AI company having access to data they shouldn’t.
Though in this instance I don’t believe so as, before the draft was published to the internet, the “The Music Of The Primes” paper had only been shared with one other person, the author’s old Standford Professor/Mentor Brian Conrad. And from the initial surface evaluation of both I’ve heard, it seems the approach the Riemann Hypothesis with completely different methods.