Three years ago, AI systems could not reliably add two numbers; a year ago, internal models at OpenAI and DeepMind achieved the equivalent of a gold-medal score on the International Mathematical Olympiad. Today, these systems autonomously resolve major open questions. This trajectory demands a radical rethinking of the mathematics profession, as the production of mathematical text is becoming increasingly disconnected from mathematical understanding.
Litt notes that the community lacks consensus on the goals of mathematics—whether problem‑solving, poetic play, or transmitting understanding—but stresses that current operationalizations, such as valuing theorem‑proving, are incomplete when AI can generate correct proofs without human comprehension. He warns that if academic institutions do not adapt, the equilibrium of mathematical progress could stall despite AI's capabilities.
To address this, Litt proposes reconceptualizing the mathematics PhD: the degree should be awarded primarily on the basis of a rigorous defense in which the student explains a deep topic to examiners until they are satisfied, rather than on the written thesis alone. This shift would ensure that expertise is demonstrated through understanding and communication, irrespective of whether the underlying results were produced with AI assistance. He also urges extending oral evaluation to graduate admissions, mirroring existing faculty‑hire interviews, and cultivating a robust seminar culture where speakers must convey their topic to the audience's satisfaction.
These reforms aim to preserve the human aspects of mathematics—understanding, explanation, and community dialogue—while accepting that AI will continue to generate vast amounts of mathematical text. Litt emphasizes that the value of mathematics lies not merely in true statements but in the ability to engage with and build upon them, a goal that requires deliberate institutional change.