Burned by OpenAI, mathematicians are now making their own AI models from scratch

What happened
After OpenAI released solutions to hundreds of difficult math problems in October, Caltech professor Sergei Gukov and his team began developing their own AI models from scratch. Gukov's group at Caltech is focused on building open-source AI systems specifically designed for solving mathematical problems. This move comes amid criticism from mathematicians who are dissatisfied with how major AI companies, especially OpenAI, handle mathematical research and data. The effort is part of a broader push within the math community to regain control over the tools and methods used in their field.
Key details
- OpenAI published solutions to hundreds of longstanding math problems in October.
- Sergei Gukov leads a lab at Caltech developing open-source AI models for mathematics.
- Most of Gukov's current models are specialized for single math problems and are not general-purpose.
- The Open Math OS platform, demonstrated by Terence Tao at Caltech, gives mathematicians control over their data and access to open-source AI tools.
- Some mathematicians have accused OpenAI of using private chat logs to solve specific math problems.
Why it matters
Mathematicians and academic researchers now have the option to build and use AI models that they fully control, instead of relying on proprietary systems from large companies. This change gives them greater oversight over how their data is used and how solutions are generated, addressing concerns about transparency and ownership. For the math community, this development means more autonomy and the ability to tailor AI tools to their specific research needs.
