Mathematicians Burned by OpenAI Are Now Building Their Own AI Models from Scratch

Deep News40 minutes ago

The conflict between the mathematics academic community and major AI companies is escalating. After OpenAI announced it had used AI to crack multiple mathematical problems, it did not receive broad recognition from the academic world. Instead, it triggered fierce debates over the attribution of research results, data usage rights, and the value of AI-generated mathematical solutions. Some mathematicians believe that commercial companies are turning mathematical research results into products and marketing material without adequately respecting the contributions and rights of the academic community.

Against this backdrop, a growing number of mathematicians are attempting to bypass commercial AI companies and independently develop open-source models and research tools. According to reports, Caltech professor Sergei Gukov is leading a team to build an AI model specifically designed for mathematical research, while Fields Medalist and UCLA professor Terence Tao is promoting the establishment of an AI collaboration platform where mathematicians control their own data.

The focus of this debate has extended beyond whether AI can solve mathematical problems itself, reaching further into questions of who controls research data, how to evaluate AI-generated mathematical results, and whether the academic community can retain its autonomy amid the rapid expansion of commercial AI.

Academic Community Questions OpenAI: Disputes Over Solutions and Data Usage

Mathematicians' dissatisfaction with OpenAI has a long history, and this controversy centers on two aspects: whether AI-generated mathematical solutions have genuine research value, and whether commercial companies have the right to use academic data and research results.

In October of this year, OpenAI announced that its model had cracked multiple mathematical problems that had remained unsolved for years, but this news did not generate widespread praise in the mathematics community. Some researchers believe that the solutions published by OpenAI lack sufficient academic value, and that it even takes human mathematicians weeks to determine whether its derivations are rigorous and whether the problems were truly solved.

Tasmin Chu, a mathematics doctoral student at Caltech and a prominently anti-AI Substack blogger, called these solutions "junk." She believes that OpenAI's move is more about leveraging research results in the field of mathematics to generate publicity for itself, rather than genuinely advancing mathematical research.

The issue of data usage has further intensified the conflict between the two sides. Previously, two mathematicians accused OpenAI of using their chat records without authorization to help solve the famous Navier-Stokes problem. This incident made mathematicians even more concerned that their interactions with AI systems could be used for commercial model training or research without permission.

Chu views mathematicians' efforts to independently develop AI systems as part of a struggle for "ownership of the means of production." However, she also acknowledges that how this trend will ultimately develop remains to be seen.

Bypassing Commercial Companies: Mathematicians Begin Building Specialized Models

Faced with the dominance of commercial AI companies, Gukov has chosen to start with model development. In his laboratory at Caltech, he leads a small group of students in designing and training open-source AI models aimed at mathematical research. According to him, these models have already solved some of the mathematical problems that OpenAI tackled.

However, self-built academic models still face obvious capability limitations. Gukov compares them to a Ferrari: excellent performance on specific tasks, but difficult to meet the needs of a broad range of users.

Currently, most of these models are optimized for a particular type of mathematical problem, and once they are asked to reason across domains, their capabilities become noticeably limited. By contrast, models from commercial AI companies such as Anthropic possess greater generality and can handle mathematical problems across different disciplines. Gukov said the team is developing models with stronger general capabilities, but they are not yet ready.

Nevertheless, he remains positive about AI's participation in mathematical research, believing that this field is full of room for exploration. He describes current AI research as being like Columbus exploring a new continent — exciting.

Mathematicians Have Not Reached Consensus: Terence Tao Pushes for an Autonomous Platform

The initiative to build independent models does not mean the mathematics community has formed a unified stance. There are still clear differences among scholars over whether AI should enter mathematical research.

Anti-AI scholars represented by Chu hope to maintain traditional research methods as much as possible. She stated that she still hopes to conduct mathematical research entirely without relying on AI, and believes this path remains viable. Another group of younger mathematicians is more positive about AI, believing that new technology could bring breakthroughs to mathematical research, and some see it as an opportunity to change existing research models.

Terence Tao is trying to find a balance between the two sides. He recently presented Open Math OS at Caltech, launched by the nonprofit organization he founded. This platform integrates open-source frontier AI models, cloud computing resources, and collaboration tools, with the key focus being to allow mathematicians to decide for themselves how their research data is used, rather than handing full control of data to commercial AI companies.

From specialized models to autonomous collaboration platforms, mathematicians are attempting to build a technical system different from that of commercial AI companies. Their core demand is not only to enable AI to better solve mathematical problems, but also to retain control over research data, knowledge outputs, and how they are used.

As AI's involvement in mathematical research continues to increase, the contradiction between the technological advantages of commercial models and the academic community's data autonomy will persist. Whether mathematicians can build an autonomous research ecosystem capable of rivaling large AI companies remains the key question in this debate.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap
Leave a comment