China building AI task force with 1,000+ scientists to beat OpenAI and Anthropic at maths

Chinese mathematicians have launched a Beijing alliance led by Shing-Tung Yau to bring AI into advanced mathematical research. The move comes days after OpenAI solved 377 math problems with AI.

In Short

• The new alliance links eight institutions and targets a 1,000-scientist network
• Researchers want AI to frame conjectures, formalise proofs and explain reasoning
• Yau said recent advances showed AI tackling problems mathematicians had struggled with

Math may be the next frontier in the AI race. After OpenAI announced that its AI model had cracked the Navier-Stokes Millennium Challenge, China wants to join in on the action with an alliance that may have over 1,000 mathematicians that will use AI in their research.

Leading Chinese mathematicians launched an alliance in Beijing last Sunday to integrate AI into mathematical research, the South China Morning Post reported. This initiative is being led by Fields medallist Shing-Tung Yau. The alliance, named “The Yau Laboratory Alliance for Mathematics and Artificial Intelligence” will bring together eight leading institutions and aims to build a network of more than 1,000 scientists.

AI is like a tsunami in math

Shing-Tung Yau is pitching the alliance as an effort to narrow what he described as a wide gap between China and the US in using AI for frontier science. “A year ago, I did not think AI was this powerful,” he said. “The development of AI in mathematics is like a tsunami.”

Yau argued that Chinese mathematicians should organise around the technology rather than avoid it. The alliance will focus on developing specialised AI models for mathematics, using the technology to generate conjectures and formalise proofs, and building systems whose reasoning can be more easily interpreted.

Shing-Tung Yau believes that China needs to close the gap to the US in Yau frontier scientific research, especially in AI systems capable of advancing mathematics and physics. He said Chinese models had performed well on tasks for ordinary users, but remained far behind leading US systems in cutting-edge research. “But if you ask them to do the work OpenAI has done over the past month, the gap is very large,” Yau said.

His comments come days after OpenAI released over 700 papers with findings on 377 math problems. At the same time, the Association for Human Mathematics has urged mathematicians to boycott OpenAI, stating that mathematicians “did not ask for this work to be done.”

OpenAI continues to face scrutiny over its math solutions. After releasing 722 mathematical manuscripts, the company withdrew three manuscripts the following day after identifying an error that invalidated an argument and affected two related papers, while revising 14 others. Some mathematicians criticised the speed and scale of the release, saying mathematical results required careful verification, clear explanations and proper acknowledgement of previous work. Meta too has previously released papers on solving six major math problems.

- Ends