Chinese Academy of Sciences academician and Beidian Shuzhi AI experts discuss the 'multiplier effect' of AI and mathematics

On July 24, 2026, the 'Sixteen Provinces (Regions, Cities) Mathematics Annual Meeting and Applied Mathematics and Artificial Intelligence Academic Forum' was held in Yuxi City, Yunnan Province, gathering more than 500 experts and mathematics teachers. During the forum, Beidian Shuzhi Chief Scientist Dou Dejing and Zhou Xiangyu, an academician of the Chinese Academy of Sciences, engaged in a deep dialogue about AI's capabilities in mathematics, the changing role of mathematicians, and the future of AI-assisted mathematical research. They presented data showing that mainstream large models improved from failing college entrance math exams in 2024 to scoring above 135 out of 140 in 2026, and can now achieve perfect scores in math olympiads without prior knowledge of the problems. However, Zhou emphasized that AI's core strength lies in pattern retrieval and mechanical deduction, not genuine mathematical innovation, and the explainability of AI's mathematical derivations remains an unresolved bottleneck. They discussed the Jacobian conjecture, noting that AI's ability to conduct large-scale heuristic searches and computations could help find counterexamples, yet the 'black box' problem persists. Zhou asserted that mathematicians will not be replaced because they do what machines cannot: proposing new questions, building new frameworks, and discovering new laws. He also raised academic integrity concerns regarding AI-generated results and citation ethics. Dou presented Beidian Shuzhi's industrial approach, including building computing centers across 30+ cities, a full-stack 'data-computing-model-application' system, and practical deployments in healthcare and manufacturing, such as a skin disease model achieving 90% accuracy on intermediate physician exams and reducing diagnosis time by 20%. The article also mentions that Anthropic's Claude Fable 5 allegedly disproved the Jacobian conjecture over a weekend, and cites Fields Medalist Efim Zelmanov's view that AI is like a 'highly talented young mathematician' but humans must bear the responsibility of truly understanding proofs and extracting insights.