Mathematician Accuses OpenAI of Using His Research
A professor at New York University, Tristan Bakmaister, accused OpenAI of potentially using his research and that of his colleague Levant Alpoge to solve a major mathematical problem related to the Navier-Stokes equations, after OpenAI claimed to have proven singularity formation in finite time. Bakmaister reported that OpenAI representatives pressured him to exclude his coauthor from Anthropic and threatened his career if he went public.
Consensus
- Tristan Bakmaister is a professor at New York University.
- Bakmaister and Levant Alpoge worked on problems related to the Navier-Stokes equations and other fluid dynamics systems.
- They used large language models, including Codex from OpenAI and Claude from Anthropic, during their research.
- Bakmaister asked OpenAI whether its internal model had access to or was trained on their Codex sessions.
- OpenAI claimed it did not use specific user data from Bakmaister and Alpoge but acknowledged that anonymized data from their use of OpenAI products might have contributed to model improvement.
- OpenAI announced it had proven singularity formation in finite time for a variant of the Navier-Stokes equations with a smooth external force.
- Bakmaister reported that OpenAI offered him two options: publish his result and have OpenAI publish theirs the next day, or write a paper on OpenAI’s result while excluding Alpoge as a coauthor.
- Bakmaister rejected both options.
- The problem is related to one of the seven Millennium Prize Problems, with a $1 million prize offered by the Clay Institute.
Points of divergence
- OpenAI representative Sebastien Bubeck threatened Bakmaister with career damage if he went public about the situation. — tg_theinsider
- Bakmaister initially contacted OpenAI to clarify rumors about Anthropic solving a key problem, not to inquire about OpenAI's results. — vm
- OpenAI did not respond directly to Bakmaister's question about when the first prompt was sent, and only later agreed it occurred within the last few days after information about their work reached OpenAI. — vm
- OpenAI did not answer Bakmaister's question about whether its model was trained on their Codex sessions. — vm
- Bakmaister and Alpoge worked on the problem for about a year. — tg_theinsider
- The work was done over a shorter period, with Bakmaister contacting OpenAI only recently. — vm
- The article mentions Mrintank Sharma, head of AI safety at Anthropic, who resigned and warned of existential risk from AI. — vm
Coverage (2 sources)
- «Why should you destroy your career?» Mathematician claims OpenAI could use his work to solve a millennium-old problem — The Insider
- Mathematics professor suspects OpenAI of using his work — Вечерняя Москва
Key entities
- OpenAI
- Tristan Bakmaster
- New York University
- Anthropic
- Claude
- Codex
- Professor of Mathematics
- Levent Alpge
- Sébastien Boubek