Live·Open questions in longevity research
All news
Scientific ComputingEcosystemIndustry & policy

A proposal would let mathematicians share ideas through AI with credit to their authors and have their usefulness count toward hiring and grant decisions

14 November 2026· 261114010

A proposal would let mathematicians share ideas through AI with credit to their authors and have their usefulness count toward hiring and grant decisions

On November 14, mathematician Michael Freedman, a Fields Medal recipient (one of the highest honors in mathematics), proposed that major AI companies and researchers create a shared space for exchanging ideas. Participants would allow all their messages and questions to be used to train models.

Freedman used to advise graduate students to tackle a central problem in their field, so that any progress would be useful to colleagues. He now suggests using AI to learn established approaches and connections to other fields, then finding a question that the models cannot resolve. Faced with such a difficult problem, a researcher could look for a special case or modify its conditions to obtain a result that would help other researchers.

A major problem can occupy a researcher for years, while an academic career requires results that can be documented and presented. Freedman proposes: “We would only need to contribute a workable idea, without having to write a full paper.”

A mathematician could describe an approach in a conversation with AI while continuing to seek a complete solution. These ideas would form a shared library. In a conversation with another researcher, the model would suggest a relevant approach from the library and credit its author. Permission to train on participants’ messages would help models improve at solving problems and explaining mathematics.

Sharing an idea this early carries a risk of losing priority: a colleague may reach a complete result sooner. Freedman accepts this risk: “We accept that we may be scooped.” Under his proposal, AI would track how ideas help other researchers’ work around the world and report this to the author or their department chair. A researcher could demonstrate their contribution even if someone else reached the complete solution first.

If universities and funders agree to consider this kind of help to colleagues, sharing ideas early could earn their authors professional recognition. Freedman proposes quantifying the usefulness of ideas when needed. Such a measure could support a job or grant application and inform a tenure decision.

Originally published on Telegram by Ukhvat NewsView on Telegram
Sources
#mathematics#ai-training#idea-sharing#author-credit#research-evaluation#academic-careers