+- 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: Centaur Math (/showthread.php?tid=1741)
Centaur Math - mklabgr - 08-28-2026
Centaur Math — Lance Fortnow
Published: August 19, 2026 — Computational Complexity
Lance Fortnow argues that mathematics is entering an era of “Centaur Math”, in which the strongest mathematical work increasingly comes from humans and AI working together. In the past, young researchers worried about competing with extraordinary mathematicians such as Richard Karp or Avi Wigderson, but those people had limited time and attention. AI systems such as Claude and ChatGPT potentially have much greater effective “bandwidth,” raising a harder question for new mathematicians: what role remains for humans as AI becomes increasingly capable of doing mathematics? Fortnow distinguishes this emerging situation from full Math AGI, which he defines roughly as the point where, with rare exceptions, any mathematical problem that AI cannot solve would also be beyond human ability. He does not think we have reached that point, and notes that major problems such as $P$ vs. $NP$ may still require genuinely new conceptual approaches that current AI does not reliably generate.
For now, Fortnow describes AI as something like a very capable PhD student: it possesses enormous mathematical knowledge, is excellent at programming, and can sometimes prove results humans have not yet obtained, but it still requires human direction, encouragement, interpretation, and verification. At the same time, mathematicians continue to solve problems that AI cannot. The greatest advantage therefore comes from combining the complementary strengths of both. Fortnow compares this stage with the brief “centaur chess” era, when human–computer teams could outperform either humans or computers alone. His concern is that mathematics might eventually follow chess, where computers became so dominant that a human deviating from the machine's recommendation generally weakens the play.
His advice to students is nevertheless optimistic: if you love mathematics, continue doing mathematics, but embrace AI rather than trying to compete with it directly. AI should be used to extend a mathematician's abilities rather than replace learning or become an intellectual crutch. Researchers should remain adaptable, challenge themselves, and concentrate increasingly on understanding the deeper ideas behind mathematical results. Fortnow's central claim is that mathematics is not merely the production of new theorems; it is also the creation of understanding—why results are true, what structures connect them, and what principles they reveal. He believes this deeper interpretive dimension remains fundamentally important for humans even as machines become much stronger theorem provers.
Key takeaways
- Centaur Math = mathematician + AI, with the combination currently stronger than either working independently.
- Math AGI has not yet arrived. Fortnow's benchmark is a world where essentially no human can solve a problem that the best AI cannot.
- AI today resembles a knowledgeable and fast PhD-level assistant, but still needs direction and verification.
- The future value of mathematicians may shift from simply producing proofs toward choosing problems, recognizing structure, interpreting results, and creating genuine mathematical understanding.
ARTICLE [PDF]