08-22-2026, 02:51 PM
Generative AI has changed mathematics forever. Where to from here?
Author: Melissa Lee, Senior Lecturer, School of Mathematics, Monash University
Published: August 10, 2026
Source:The Conversation
Generative AI is moving beyond simply helping mathematicians write code, search literature, or check calculations: it is beginning to participate directly in mathematical discovery. Melissa Lee points to OpenAI’s announcement of ten advances in mathematics and computer science produced with its then-unreleased Astra model, covering areas including geometry, cryptography and coding theory. These developments force mathematicians to confront questions that were largely philosophical only a few years ago: if an AI can discover a proof, construct an example, improve an algorithm or formulate a promising research question, what remains uniquely human about mathematical creativity? The issue is therefore shifting from whether AI can contribute to serious mathematics to how its contributions should be incorporated into mathematical practice.
The mathematical community has not reached a consensus. Universities, journals, arXiv and funding agencies are increasingly having to establish policies about AI-generated work, attribution, verification and disclosure. Some mathematicians fear that extensive AI use could undermine the human culture of mathematical understanding, while initiatives such as the Leiden Declaration argue that AI should augment rather than replace mathematicians, with transparency, accountability and attribution remaining essential. Terence Tao similarly argues that mathematicians should focus on designing a culture in which AI strengthens mathematical research rather than merely accelerating theorem production.
Lee illustrates the tension through two striking examples from her own research. In one collaboration on the semiregularity problem, her co-author Saul Freedman explicitly rejected AI for ethical and environmental reasons, and the problem was solved entirely by humans. In another project, Aluna Rizzoli used an OpenAI model together with a supercomputing cluster to find, in only 43 hours, a mathematical object that Lee and collaborators had searched for for more than two years. The AI devised a substantially more sophisticated version of an existing algorithm involving the Monster group, a development Lee estimates would have required months of human work. Importantly, Rizzoli then invited the original researchers to collaborate rather than simply claiming the discovery. For Lee, this suggests that AI need not destroy mathematical culture: the crucial challenge is preserving human understanding, collaboration, transparency and intellectual integrity while exploiting AI's rapidly increasing capacity for discovery.
Key takeaways
ARTICLE
Author: Melissa Lee, Senior Lecturer, School of Mathematics, Monash University
Published: August 10, 2026
Source:The Conversation
Generative AI is moving beyond simply helping mathematicians write code, search literature, or check calculations: it is beginning to participate directly in mathematical discovery. Melissa Lee points to OpenAI’s announcement of ten advances in mathematics and computer science produced with its then-unreleased Astra model, covering areas including geometry, cryptography and coding theory. These developments force mathematicians to confront questions that were largely philosophical only a few years ago: if an AI can discover a proof, construct an example, improve an algorithm or formulate a promising research question, what remains uniquely human about mathematical creativity? The issue is therefore shifting from whether AI can contribute to serious mathematics to how its contributions should be incorporated into mathematical practice.
The mathematical community has not reached a consensus. Universities, journals, arXiv and funding agencies are increasingly having to establish policies about AI-generated work, attribution, verification and disclosure. Some mathematicians fear that extensive AI use could undermine the human culture of mathematical understanding, while initiatives such as the Leiden Declaration argue that AI should augment rather than replace mathematicians, with transparency, accountability and attribution remaining essential. Terence Tao similarly argues that mathematicians should focus on designing a culture in which AI strengthens mathematical research rather than merely accelerating theorem production.
Lee illustrates the tension through two striking examples from her own research. In one collaboration on the semiregularity problem, her co-author Saul Freedman explicitly rejected AI for ethical and environmental reasons, and the problem was solved entirely by humans. In another project, Aluna Rizzoli used an OpenAI model together with a supercomputing cluster to find, in only 43 hours, a mathematical object that Lee and collaborators had searched for for more than two years. The AI devised a substantially more sophisticated version of an existing algorithm involving the Monster group, a development Lee estimates would have required months of human work. Importantly, Rizzoli then invited the original researchers to collaborate rather than simply claiming the discovery. For Lee, this suggests that AI need not destroy mathematical culture: the crucial challenge is preserving human understanding, collaboration, transparency and intellectual integrity while exploiting AI's rapidly increasing capacity for discovery.
Key takeaways
- The important question has changed: it is no longer “Can AI do genuine mathematical research?” but “How should mathematicians work with AI?”
- AI may dramatically compress research timescales, reducing work that could take months or years to hours or days.
- Mathematical research is more than theorem production; understanding, explanation, attribution and human collaboration remain central.
- The likely future is neither purely human nor purely automated mathematics, but a form of human–AI collaborative mathematics, whose norms are only now being established.
ARTICLE
┌────────────────────────────────┐
│ KONSTANTINOS MICHAILIDIS │
└────────────────────────────────┘
│ KONSTANTINOS MICHAILIDIS │
└────────────────────────────────┘