Proofs and Prompts

Notes from the common room

What Mathematicians Want?

Reuven Peleg, M.Sc student for Computer Science at Hebrew University of Jerusalem, Israel

AI Disclosure: This post was written entirely by me, AI was used to proofread the final draft.

The current big wave of AI for Math discourse was forced upon the Mathematical community. The big labs got more and more advanced results on conjecture proving, triggering all kinds of reactions, feelings and backfire reactions. The main focus of discussion was how to react to events: Should the community cooperate with AI labs, or scrutinize them? How should journals and conferences treat AI-generated papers?

I would like to ask a different question: What do you want from AI as a mathematician? I find it intellectually interesting, but also practical for me, as I don’t want to conduct research that steps on someone’s toe instead of helping. As a Master’s student interested deeply in AI For Math and planning out his future career, I want to politely ask the professionals in it what their needs or expectations are.

One answer I got is “leave us alone”. Some professionals say they want to use AI for literature review, spellcheck, maybe some assistance in coding if necessary, and that’s it. Just like an artist who wants the robots to put down the brushes, they want the human-centric experience of creation and exploration.

On the other side of the spectrum, there seems to be a camp of pro-AI mathematicians that are generally happy about machine-discovered math, with some specific concern: About reliability of the claims, copyright, attribution and norms, ownership, responsibility for papers, et cetera. Once it’s settled, they want to have the most powerful, fast, cheap and reliable Math AI that can be built.

I don’t have a clear understanding on how big each camp is, what are the positions in between, or what aspect I miss in the needs of the community. I also hope there is a way for AI to fit in Math research, maybe in a way I can’t envision as an outsider. So I am asking: If you could steer the AI research and have your dream outcome for the field, what would it be?

Received 9 October 2026.

Filed under ·
·
More by →

New to Proofs and Prompts? Start here →

One response to “What Mathematicians Want?”

  1. Sal Avatar
    Sal

    I don’t know what I want from it.

    I have used specifically as a tool to check my algebra for a not nice expression for instance but I include in my prompts that I don’t want to be presented with any mathematical ideas that I have not done myself.

    What I do want from big results that are entirely AI-generated is an opener with a disclaimer by the prompters that the result is a candidate proof or candidate result. I also think the prompter should include a statement of what they do and don’t understand. I think there should be no author but rather the title Prompter: (insert names). This last point if it is a human using the AI as a tool then author makes sense to me, this is only about the results generated in which the only human contribution is the prompt(s). I want preemptive statements of acknowledgement to the possibly numerous humans that will read over the work in an attempt to verify. I think there needs be well documented ways in which Lean verifications are handled and that humans needs to be involved in the writing of the code in some capacity. I am not interested in the AI generating the result and verifying it as well. I am not even interested in say GPT obtaining a result and Claude verifying by means of writing Lean or otherwise with very little human involvement.

    Reply

Comments are moderated. Read our comment policy.

Add to the discussion

Subscribe

New posts by email.

Prefer a feed reader? Subscribe by RSS.

Also on Mathstodon.

Join the discussion

Latest comments across the site.

  2. “And in the case of automatic problem-solving, where the model works essentially on its own to close the loop of...

  3. I don’t know what I want from it. I have used specifically as a tool to check my algebra for...

Read next