Mathematicians are grappling with what they describe as an existential crisis after OpenAI published solutions to longstanding problems in the field, according to a Decoder interview with The Verge's London-based AI reporter Robert Hart.
What Happened
OpenAI recently published a set of solutions to longstanding math problems that sparked intense debate within the mathematics community. According to the report, AI systems have undergone a significant capability transition over roughly six months to a year, going from being "very terrible" at mathematical reasoning to seemingly genuinely quite good at professional-level work in a compressed timeframe. Hart noted that as recently as 2024, conventional wisdom held that AI models struggled particularly with math, citing the well-known example of these systems' inability to count the number of R's in the word "strawberry." The new solutions published by OpenAI went "off like a bombshell" among mathematicians.
Why It Matters
The development raises fundamental questions about the future role of academic mathematicians and mathematics as a discipline. Mathematicians Hart spoke with are questioning what happens to university programs, grants, and the training of new generations of researchers if frontier AI models can simply answer outstanding mathematical questions. The interview highlighted a paradox: while AI remains "truly terrible" at basic arithmetic tasks such as counting, determining days of the week, or telling time, it has reached a level where it excels at forging connections between different areas and applying old methods in new ways—skills central to advanced academic math work that often involves abstract reasoning rather than numerical calculation. Mathematicians have identified specific subfields like topology where AI capabilities appear more limited.
The Bottom Line
The mathematics community is navigating uncharted territory as AI systems demonstrate professional-level capability in certain mathematical domains while remaining unreliable for elementary tasks. Whether these advances represent a genuine shift in the discipline's future or primarily serve as marketing for frontier AI labs remains a subject of active debate among researchers.