OpenAI assures that it has solved one of the great mathematical problems of the millennium but they accuse it of plagiarism

OpenAI, the company that created ChatGPT, stated this Tuesday that one of its artificial intelligence models has found the solution to the Navier-Stokes problem, one of the six millennium mathematical problems, which has remained unsolved for 90 years. These enigmas were established in the year 2000 by the Clay Mathematics Institute, which offers a prize of one million dollars for each solution found. "This milestone represents substantial work by mathematicians and AI researchers," the company explained in a report in which it published the answer.

According to the company, 10,000 AI agents worked on it in a coordinated manner for 88 hours, through a model that has not yet been presented to the public and that is "significantly more capable" than its latest ones. In this sense, the company insists that this milestone is irrefutable evidence that we are already "in the next period of AI progress."

Although OpenAI claims to have solved the problem, it should be noted that it will still take months to verify if, indeed, the company has solved it correctly.

Accused of plagiarism

However, the alleged historic milestone has been overshadowed by accusations of plagiarism. According to The Wall Street Journal and the CEO of OpenAI himself, Sam Altman, the company began testing its new model on September 1st with the so-called Millennium Problems just after rumors reached them that two of these problems were about to be solved.

The rumors pointed to the research by New York University mathematician Tristan Buckmaster and Anthropic researcher -a competing company- Levent Alpöge, who were working together personally and free of any official agreement. "It is true that we tried it because last week there were rumors on the internet that Anthropic's models had solved a Millennium problem and we were curious to know if ours could do it too," Altman noted. In practice, however, Buckmaster and Alpöge had made significant progress in solving the problem, but they had not found the solution.

Finally, the night before OpenAI published the solution, the university professor published a PDF with important details on the subject. He asserted, on one hand, that he had not yet solved the problem and that he had recently been in contact with the company. Specifically, they discussed the fact that one of OpenAI's models had supposedly found an answer to the enigma, and that it had likely done so through one of the methods that Buckmaster and Alpöge were using. "Almost no one else I know was working on [this method]. It is not the direction one arrives at in a few days by giving a model the statement of the problem," the professor explains in the PDF.

Thus, the company offered the professor to co-sign the solution, with the sole condition that Alpöge not be included due to his connection with Anthropic. They insisted that they wanted the researcher "to be removed from the authorship".

The professor refused both requests, stressing that he had felt "pressured" by OpenAI and that he had not been able to review the way in which the models had solved the enigma. One day later, OpenAI published the solution.

For its part, Altman's company has denied the accusations of plagiarism, but acknowledges that it cannot be ruled out that the conversations that the researchers have had with the AI may have contributed to solving the problem in some way.