Tensions between OpenAI and the math community have once again escalated. Fields Medal winner Tao Zhexuan recently reprinted a statement issued by the Human Mathematics Association on her personal blog, criticizing OpenAI for ignoring mathematical research standards and calling on mathematicians to stop collaborating with it. The controversy stemmed from a number of mathematical research manuscripts published by OpenAI. Its internal model tried to solve about 4,000 mathematical problems, finally sorted out research materials involving 372 groups of results, and disclosed part of Lean's formal proof code. However, the scale of the results did not receive unanimous recognition from the academic community. Mathematicians are concerned not only about whether the proof is correct, but also whether the deduction can be understood, whether the work of previous people is fully quoted, and whether the new method can be integrated into the existing knowledge system through peer discussion. Formal verification can check logical deductions, but it cannot automatically replace manual review of the meaning of a proposition, research contributions, and academic background. “Publishing more than 700 documents at once shows not academic research, but power.” The statement said.

Tensions between OpenAI and the math community have once again escalated. Fields Medal winner Tao Zhexuan recently reprinted a statement issued by the Human Mathematics Association on her personal blog, criticizing OpenAI for ignoring mathematical research standards and calling on mathematicians to stop collaborating with it. The controversy stemmed from a number of mathematical research manuscripts published by OpenAI. Its internal model tried to solve about 4,000 mathematical problems, finally sorted out research materials involving 372 groups of results, and disclosed part of Lean's formal proof code. However, the scale of the results did not receive unanimous recognition from the academic community. Mathematicians are concerned not only about whether the proof is correct, but also whether the deduction can be understood, whether the work of previous people is fully quoted, and whether the new method can be integrated into the existing knowledge system through peer discussion. Formal verification can check logical deductions, but it cannot automatically replace manual review of the meaning of a proposition, research contributions, and academic background. “Publishing more than 700 documents at once shows not academic research, but power.” The statement said.