+- MKLab (https://mklab.gr)
+-- Forum: [INDEX] (https://mklab.gr/forumdisplay.php?fid=1)
+--- Forum: MATHEMATICS (https://mklab.gr/forumdisplay.php?fid=3)
+---- Forum: ARTICLES (https://mklab.gr/forumdisplay.php?fid=13)
+----- Forum: AI AND TECHNOLOGY (https://mklab.gr/forumdisplay.php?fid=158)
+----- Thread: Mathematics in the age of AI [Tao] (/showthread.php?tid=1693)
Mathematics in the age of AI [Tao] - mklabgr - 08-20-2026
Mathematics in the Age of AI
Author: Terence Tao
Published: August 17, 2026
Terence Tao argues that the most important question raised by AI in mathematics is not simply whether AI will become capable of research-level mathematics, but what mathematics itself is ultimately trying to achieve. Rather than debating predictions about AI capability, Tao assumes as a working hypothesis that reasonably powerful AI systems will soon be able to perform a substantial fraction of research-level mathematical tasks. Under that assumption, mathematicians will be forced to make explicit many values that have traditionally remained implicit: mathematics is not merely about producing correct proofs, but also about developing theories, increasing human understanding, training mathematicians, building communities, connecting ideas, and creating knowledge that becomes part of the lasting mathematical canon.
Tao illustrates this through mathematical problem solving. At first one might define success as simply “solving as many unsolved problems as possible.” But this quickly expands into a much richer pipeline:
problem solved → proof verified → proof clearly explained → result understood and accepted by the community → result incorporated into the canonical theory of the field.
AI may become extremely good at the first two stages—generating proofs and formally verifying them using systems such as Lean—but the later stages depend heavily on human understanding, exposition, refereeing, interpretation and judgment. A formally correct 100-page AI-generated proof that no mathematician understands may therefore have surprisingly little mathematical value. Tao even proposes a practical principle: if the human authors cannot give a convincing expert-level explanation of their result and properly attribute its intellectual origins, the result should not yet be considered ready for publication.
The deeper change Tao anticipates is a transition from “proof scarcity” to “proof abundance.” Today, mathematical institutions—journals, prizes, hiring, priority conventions and research programs—largely assume that obtaining a new proof is difficult and rare. If AI can generate thousands of legitimate results, the bottleneck shifts from producing proofs to filtering, understanding, explaining, refereeing and integrating them into mathematics. Tao therefore argues that mathematicians should place less emphasis on simply being the first to prove something and more emphasis on what he calls proof digestion: exposition, verification, peer review and canonicalization. He also stresses transparency about AI use and keeping human authors responsible for mathematical claims.
Key takeaways
- AI may force mathematics to redefine what counts as mathematical progress. A correct theorem is only one component of useful mathematics.
- The main future bottleneck may be human attention rather than proof generation. AI could produce results faster than mathematicians can verify, understand or integrate them.
- Understanding remains central. Tao strongly resists the idea that a formally verified but essentially incomprehensible AI proof should automatically count as a completed mathematical contribution.
- Mathematical culture may need to move from a “theorem economy” toward an “understanding economy,” rewarding exposition, refereeing, synthesis and theory building much more than it currently does. (a
The central idea in one sentence
Quote:Tao’s message is essentially that if AI makes proving theorems cheap, the real value of mathematicians will increasingly lie in deciding what is important, understanding why it is true, explaining it, connecting it to other mathematics, and turning isolated results into coherent theory.
ARTICLE [PDF]