For those who are out of the loop, here is a TLDR of the situation:

Joint statement by the Lean and Mathlib communities, endorsed by the Lean FRO and the Mathlib Initiative.

On November 20, 2026, Levent Alpöge released a proof of the Riemann Hypothesis as a Lean library comprising 17 million lines of code.

We have full confidence that Alpöge’s proof of the Riemann Hypothesis is correct, and we congratulate him on this achievement.

Source: https://leanprover-community.github.io/blog/posts/riemann-hypothesis/

  • gnutrino@programming.dev
    link
    fedilink
    English
    arrow-up
    34
    arrow-down
    2
    ·
    edit-2
    9 hours ago

    You run your stochastic parrot in a loop generating Lean and have a (proper, deterministic, non-AI) Lean validator run on the output each time. In this case, you also have a human iteratively give the stochastic parrot starting points and general directions to try. And then you burn millions of dollars of compute effectively monkey-typewritering it until the validation passes.

    I think people really underestimate the sheer scale of resources allocated to that last part - this is way beyond something a normal mathematician on an academic budget could do and, ironically enough, I don’t think even most mathematicians would argue that this is a sensible use of those resources.

    • NoLemurs@lemmy.world
      link
      fedilink
      arrow-up
      15
      arrow-down
      1
      ·
      9 hours ago

      This captures the essence of what the LLM is doing, but is ignoring the fact that the whole thing was being driven by a very good mathematician.

      The headlines keep saying things like “Claude proves the Riemann Hypothesis”, but an accurate statement would be “Levent Alpöge proves the Riemann Hypothesis using AI assistance.” The AI didn’t have to search the whole space of possible solutions - Alpöge was there to direct and prune the search space based on his own ideas about what was likely to work. Alpöge couldn’t have done this on his own, but the LLM couldn’t have either.

      • gnutrino@programming.dev
        link
        fedilink
        English
        arrow-up
        6
        ·
        9 hours ago

        but is ignoring the fact that the whole thing was being driven by a very good mathematician.

        Is it?

        In this case, you also have a human iteratively give the stochastic parrot starting points and general directions to try.

        • NoLemurs@lemmy.world
          link
          fedilink
          arrow-up
          5
          arrow-down
          1
          ·
          9 hours ago

          Ugh. I swear I read your post three times and someone missed that. My brain is clearly not working this morning. My apologies.

    • Photonic@sopuli.xyz
      link
      fedilink
      arrow-up
      6
      ·
      12 hours ago

      I don’t think even most mathematicians would argue that this is a sensible use of those resources.

      Is it… because they did the math?