RUUZEN
AI

After Getting Burned by OpenAI, Mathematicians Now Develop Their Own AI Models from Scratch

Caltech Professor Sergei Gukov and his students are creating open-source mathematical AI models; according to him, they have solved some problems before OpenAI.

20 Nov, 10:28Read in Russian
businessinsider.com

In Gukov’s laboratory, models are built from scratch and tailored for mathematical tasks. Most can solve only one specific problem, while a more versatile system is still under development.

In October, OpenAI solved hundreds of complex mathematical problems that had stumped specialists for years. Many mathematicians believe such answers are difficult to use without a week of human verification and decoding. Gukov argues that in-house AI models will help researchers control solutions and data usage.

The community is divided: some prefer to do mathematics without AI, while others are actively adopting the technology. Mathematician Tasmin Chu described the systems created by mathematicians as a way to regain control over their tools of production.

Mathematician Terence Tao at Caltech presented Open Math OS—a platform from his nonprofit organization. It provides access to open-source models, cloud computing, and collaborative tools, while also allowing control over data usage. The issue intensified after two mathematicians accused OpenAI of using their correspondence in solving the Navier-Stokes problem.