OpenAI is at the center of a dispute over an AI-generated solution to a Millennium Problem, with researcher Tristan Buckmaster accusing the company of using his drafts and pressuring co-author Levent Alpöge.
The incident has sparked broader concerns about the reliability of AI labs in research. The Navier-Stokes problem, one of the Clay Millennium Problems, carries a $1 million prize and has drawn public statements from all parties involved.
Buckmaster claims OpenAI trained its models on drafts he and Alpöge uploaded to Codex, calling it 'absolute academic malpractice.' OpenAI employees, including Boaz Barak, argue that the model did not require external help, noting it initially proved a stronger claim than the researchers' own.
The company acknowledged a gap in certainty, stating it cannot rule out that de-identified data from their products helped improve its models.
OpenAI CEO Sam Altman defended the company, stating the effort began due to rumors about Anthropic's models solving a Millennium Problem. He claimed the team acted with integrity and that the project was driven by curiosity. However, Alpöge contradicted Altman, saying he would have liked to collaborate and that authorship did not matter to him.
The case raises critical questions about how AI labs engage with the research community. Rumors alone were enough for OpenAI to deploy significant resources toward solving the problem, potentially undermining original research. Mathematician Terence Tao warned that such rumors could trigger massive AI-powered efforts to outpace traditional research, threatening the traditions of open science.
The incident highlights the risks of feeding research data into AI systems, as labs have not earned the benefit of the doubt on training and data practices. Whether OpenAI trained on the submitted solutions remains unclear, and the company has not provided definitive answers. The dispute underscores the need for transparency and trust in AI-driven research.
Source: thedecoder
