Login
You're viewing the functional.cafe public feed.
  • Nov 21, 2026, 4:46 AM

    Some fun facts about the Millinneum Prize Problems:

    1. A proof must be published in a qualifying journal and independently verified in order to qualify. No AI-generated solution yet meets those criteria.

    2. The only solution so far accepted was produced without AI in 2003.

    3. Each prize is $1M, a substantial sum for most mathematicians, but likely less than what AI companies spend on the compute required to produce their solutions.

    4. GenAI proofs tend to be extremely verbose and inefficient (the Reimann hypothesis proof is reportedly ~250,000 pages long), and will take years, if not decades, to verify or disprove. Mathematicians could plausibly waste entire careers wading through inconclusive slop.

    5. The problems aren't particularly urgent, eg. people don't regularly die of Hodge conjecture, but LLMs are better at generating plausible-looking proofs than they are at, say, solving climate change, and AI companies need splashy stories to sustain the narrative that they can solve Big Problems, so they've converged on the Prize Problems as a more attainable way to attract favorable coverage.

    💬 6🔄 45⭐ 51

Replies

  • Nov 21, 2026, 4:51 AM

    If I had to make a prediction, I would say that the full solutions LLMs propose to the Prize Problems will never be independently verified. The sheer amount of work involved is too vast, and most mathematicians will prefer to spend their time working out their own proofs. Human-produced proofs will likely beat most AI proofs to the prizes simply because they'll be easier for their colleagues to verify.

    💬 1🔄 1⭐ 14
  • Nov 21, 2026, 1:43 PM

    In effect, these corporate sponsored AI-generated proofs are the mathematical equivalent of flooding FOSS projects with vibecoded pull requests. Taking them seriously means committing huge amounts of unpaid time and labor to checking their work, with a disproportionately high risk that they'll turn out to be useless.

    💬 0🔄 3⭐ 0
  • Nov 21, 2026, 5:08 AM

    @lrhodes The Anthropic proof might already have been disproven. (As I understand it) there's already a paper discussing how it relies on an older, known-buggy version of Lean, and even if the proof compiles correctly in updated versions - hard to say, given how fucking massive it is! - we still wouldn't be any closer to a correct proof of the Riemann hypothesis.
    philpapers.org/archive/NIEIOT-

    💬 2🔄 2⭐ 4
  • Nov 21, 2026, 10:19 AM

    Wow, that was quick. The arguments seem stronger than that the version is buggy (although I am no Lean expert). It seems to be that some of the things Alpöge defined that are supposed to be tests are definitionally true and so test nothing.

    Like I said I am no Lean expert and don't know what he defined vs what was provided by Lean.

    @Video_Game_King @lrhodes

    💬 1🔄 0⭐ 1
  • Nov 21, 2026, 1:54 PM

    @dogfox @Video_Game_King @lrhodes I would not place excess confidence in that source (if you look around the site you'll understand better why I say that). Quite a few people have run independent compilations of the proof with different checkers. The statement that was proved was the one in mathlib, which is generally agreed to be a correct statement of RH. That doesn't mean 100% as there have been cross-checker bugs in the past, but it's not very likely.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Nov 21, 2026, 5:11 AM

    @lrhodes I have a really hard time believing that a system which can’t do a basic word problem or hold onto a mathematical thought for more than half a sentence is going to actually prove Reimann. Unless it’s an infinite monkey theorem kind of thing.

    💬 0🔄 1⭐ 6
  • 💬 0🔄 0⭐ 2
  • Nov 21, 2026, 9:52 AM

    @lrhodes
    250,000 pages? Lol, get fucked. As if these machines can write 250,000 pages of Lean code thats both on topic and bug free... LOL, LMAO even.

    💬 0🔄 1⭐ 4
  • Nov 21, 2026, 10:29 AM

    @lrhodes Also, the point of the Millennium prizes, like Hilbert's problems before then, is that a proper proof would have to involve new and interesting mathematics that would open up whole fields of research.

    As you point out, whether the Hodge conjecture is true or false really isn't of any value, it's the work in pricing it so that is the value.

    LLM proofs have none of that value.

    (*slight caveat that a black box proof that P=NP would probably have some value, the other problems much less so)

    💬 1🔄 0⭐ 0
  • Nov 21, 2026, 12:10 PM

    @screw_dog @lrhodes Lik your caveat, an incomprehensible but valid proof that RH is *false* would be huge because it would alter other efforts all over mathematics just like P=NP would alter lots of assumptions about what is safely difficult in computing. But the expected results alter nothing except through the methods in the proof being illuminating.

    💬 0🔄 0⭐ 0