/file/attachments/orphans/BM-Sidley-Farewell-Nobel_578290.jpg "How AI Is Transforming Scientific Discovery and Intellectual Property")
Last week, OpenAI made headlines by claiming a solution to the Navier-Stokes problem, a significant mathematical challenge regarding fluid dynamics. This problem has puzzled scientists for decades and is one of the seven Millennium Prize Problems, which have stumped experts for almost 90 years. OpenAI announced that one of its unreleased models had solved the problem in just 88 hours. While this development seems remarkable, it stirred controversy as academic allegations emerged regarding the source of the AI’s solution.
Tristan Buckmaster, a mathematics professor at New York University, raised concerns that OpenAI may have accessed his unpublished work. Two Spanish mathematicians, Luis Martínez-Zoroa and Diego Córdoba, had been attempting to tackle the Navier-Stokes problem, and Buckmaster had collaborated with Levent Alpöge from Anthropic, using AI tools for their research. They relied on multiple AI models, including those from OpenAI, to enhance their understanding.
As OpenAI geared up to announce its breakthrough, Buckmaster discovered that the approach used by their AI was similar to their ongoing research. During discussions with OpenAI, he sought clarification on whether their work had influenced the AI’s training but received vague responses. The situation escalated when Buckmaster was offered two choices: publish a partial result alongside OpenAI’s announcement or publish his findings on his own and claim the associated prize but with restrictions regarding Alpöge’s involvement.
Buckmaster opted for transparency and indicated he would go public with his concerns. In response to accusations, OpenAI’s Sébastien Bubeck denied any wrongdoing and stated that academic norms were followed. However, OpenAI admitted it could not completely dismiss the possibility that anonymised data from Buckmaster and Alpöge’s work had influenced its models.
This incident underscores the complexities of attribution in the age of AI, raising questions about academic credit for future discoveries in science and mathematics. As AI continues to evolve, it challenges the traditional frameworks of recognition in the academic world and leaves us pondering the future significance of established accolades like Nobel prizes.
