OpenAI's Navier-Stokes claim sparks backlash from mathematicians over credit
OpenAI's September 8 announcement that its AI agents solved the famous Navier-Stokes problem has triggered anger among mathematicians over insufficient credit for human work and concerns their research data was used without permission.
On September 8, OpenAI announced that its artificial intelligence agents had cracked the Navier-Stokes problem, one of mathematics' most famous and difficult open challenges. Had a human mathematician achieved this, they would have earned prize money and recognition from peers. Instead, the announcement has provoked what some mathematicians describe as an "existential crisis" within their field, along with accusations that OpenAI improperly relied on other people's work.
The complaints raised by mathematicians echo concerns already voiced by artists and other professionals worried about AI's impact on their fields. Like any large language model, OpenAI's output depends on absorbing work previously produced by human mathematicians. Although the company's paper does cite sources, many believe it fails to adequately credit several mathematicians who were reportedly close to solving the problem themselves.
Mathematician Tristan Buckmaster has raised concerns that his own work on the Navier-Stokes problem, conducted using OpenAI's Codex model, may have been seen by OpenAI's team. The company has denied directly accessing this material but has not ruled out that data generated through Buckmaster's use of its products "helped improve" its model.
This absence of clear attribution and compensation for the human labor underpinning AI systems has turned many researchers against the companies building them. Mathematicians point out that AI firms are eager to showcase their models' mathematical abilities as a selling point, yet still depend on human mathematicians to check work that is often messy and difficult to interpret, and to determine whether it has any real value.
Despite these tensions, mathematicians remain notably open to using AI, even as many sign open letters criticizing tech companies. Most acknowledge the technology's genuine power, particularly in mathematics, and discussions continue about how the field might evolve to incorporate AI while humans guide and interpret its output. Fields Medal winner Bill Thurston observed back in 2010 that the purpose of mathematics is clarity and understanding, not theorems themselves — qualities that remain distinctly human even in the AI era.
