@PSL2Z@mathstodon.xyz

· 7mo ago

Kevin Barreto, one of the first people to get a publicly available LLM (AI) to solve a relatively known open mathematical problem, gave a great talk yesterday, which I organised. The slides are available, with the detailed description of his workflow:
https://drive.google.com/file/d/1VQk1_Se74ffYrTs1u5dUXq9gnxuzkMFv/view

drive.google.com

8

0

3

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 2mo ago

Replying to

@johncarlosbaez@mathstodon.xyz A qualified PhD student can gain access to frontier models without paying this money themselves, and should.

1

2

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Replying to

@highergeometer@mathstodon.xyz Borcherds https://epochai.substack.com/p/ai-math-chat-1-thinking-smarter-not Hairer participates in the First Proof challenge, and his problem wasn't solved in February. A lot of people write about negative results: Litt's takes on this topic are always very sober; benchmarks like FrontierMath are not yet saturated. "Work outside the topics that Tao, a well-known polymath, works on" is an interesting requirement.

epochai.substack.com

AI + Math Chat #1 - Thinking smarter not harder

“How close are today’s large-language models to genuine mathematical creativity?” with mathematicians Richard Borcherds, Sergei Gukov, Daniel Litt, and Ken Ono (originally uploaded Aug 18, 2025).

3

14

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Replying to

@highergeometer@mathstodon.xyz People have little incentive to do this kind of work that is going to become obsolete in a few months if not weeks, likely before the paper (if it is submitted for publication) goes through the publication pipeline.

2

5

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Today at an interview: "In one of the Three Musketeer's sequels by Alexander Dumas, Aramis is interviewed for the position of the head of the Jesuit Order. Each candidate must reveal a secret. So, now I will reveal a secret for you." And then I told them about an open conjecture that we proved last week.

1

0

1

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Replying to

@highergeometer@mathstodon.xyz I don't know how few, I randomly searched four and found two who are involved in the discussion (in my first reply).

1

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 7mo ago

Thomas Fernique's page is so enjoyable. Footballs of weird shape, which can actually be kicked! Wooden plane partition / arctic circles! Cooling of Penrose tilings!
https://lipn.univ-paris13.fr/~fernique/gallery.html

lipn.univ-paris13.fr

Thomas Fernique - Gallery

2

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 9mo ago

Replying to

@tao This is what they call "jagged frontier" of AI capabilities [with the expectation of increased jaggedness]. https://helentoner.substack.com/p/taking-jaggedness-seriously

helentoner.substack.com

Taking Jaggedness Seriously

Why we should expect AI capabilities to keep being extremely uneven, and why that matters

1

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Replying to

@highergeometer@mathstodon.xyz Like, here is a paper in this genre https://arxiv.org/abs/2506.17114 but it is just irrelevant today. The paper studies ChatGPT o3 and was posted in June, and in August ChatGPT 5.0 was released.

arXiv.org

Mathematical Proof as a Litmus Test: Revealing Failure Modes of Advanced Large Reasoning Models

Large reasoning models (e.g., R1, o3) have demonstrated remarkable mathematical problem-solving abilities. However, the high reported accuracy of these advanced models on popular datasets, reliance on purely numerical evaluation and potential benchmark leakage, often masks their true reasoning shortcomings. To address this, we propose leveraging the inherent rigor and methodological complexity of mathematical proofs as a diagnostic tool to expose these hidden failures. Specifically, we introduce

0

4

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 5mo ago

People in my Mathstodon feed react to AI in math in one of three ways: hate, pragmatism, or curiosity. I also encountered willful ignorance. I don't think these attitudes are equally good. I am pragmatic, and I like curiosity.

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Replying to

@highergeometer@mathstodon.xyz Riemannian geometry https://terrytao.wordpress.com/2019/02/19/on-the-universality-of-the-incompressible-euler-equation-on-compact-manifolds-ii-non-rigidity-of-euler-flows/

What's new

On the universality of the incompressible Euler equation on compact manifolds, II. Non-rigidity of Euler flows

I have just uploaded to the arXiv my paper “On the universality of the incompressible Euler equation on compact manifolds, II. Non-rigidity of Euler flows”, submitted to Pure and Applie…

0

1

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Replying to

@highergeometer@mathstodon.xyz What do you want from a proper write-up?

0

9

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 4mo ago

Replying to

@highergeometer@mathstodon.xyz Fair, "publishable papers tend to contain positive results" is an observation that is hard to argue with

0

7

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 9mo ago

Replying to

@johncarlosbaez@mathstodon.xyz Here is the version without the "instanton" label: https://commons.wikimedia.org/wiki/File:Masses_and_sizes_of_objects_in_our_Universe.png

commons.wikimedia.org

0

1

0

PSL₂(Zabolotskii) @PSL2Z@mathstodon.xyz

· 2mo ago

Replying to

@johncarlosbaez@mathstodon.xyz Re. costs predictions: perhaps. My point is just that today, as of July 2026, Buzzard is right (except the students can and should work around the costs). Re. your last point: of course, no doubt. Looks like we'll have to be good ourselves and raise good students.

0

2

0