AI Solves Longstanding Math Problems, Mathematicians Remain Essential

Recent advances in artificial intelligence are reshaping mathematical research, moving large models and specialized agents from elementary arithmetic errors toward contributions on long-standing theoretical problems, Northeastern University mathematicians said. Once unreliable even for simple calculations, AI systems improved markedly in mathematical reasoning. Google DeepMind won a silver medal in an elite high-school competition in 2024, while OpenAI took gold a year later. More recently, reported results include a May 2026 advance on a 1946 problem posed by Paul Erdős about the maximum number of unit-distance pairs among points in the plane, overturning a long-standing expert prediction. In September, OpenAI announced that its agents reported a solution to the Navier-Stokes problem, a central challenge in fluid mechanics, after a process said to take about 88 hours and draw on roughly 10,000 bots scanning decades of literature. Ken Duffy, chair of mathematics at Northeastern University in Boston, said such problems are useful benchmarks because they have clear success criteria. He also cautioned that research mathematics is not only about correct answers. Its value lies in asking good questions, constructing explainable proofs and opening new methods for exploration. AI currently excels at brute-force search, data-heavy tasks and rapid pattern linking, but often fails to provide human-readable reasoning. Some machine-generated proofs have been checked in formal systems such as Lean, and some have been withdrawn when errors emerged. The technology is already changing adjacent scientific fields. Duffy’s team used AI to identify and map millions of bone-marrow cells, sparing graduate students laborious image-by-image analysis. In biology, AI has also helped predict protein structures at massive scale. James Boyd of the Wolfram Institute argued the field should broaden toward experimental mathematics, using computation to reveal patterns while retaining human oversight of the investigative process. Despite automation concerns, Duffy said demand for math graduates remains strong in finance, engineering and quantitative roles, because AI still hallucinates and cannot replace critical judgment. The near-term implication for researchers and students is that AI is becoming a powerful reasoning assistant, but humans remain essential for framing problems, validating arguments and deciding what counts as meaningful progress.