In recent months, non-commercial AI models have made significant advances in the field, and their creators have claimed to have solved hundreds of problems that were not, raising many questions within the global scientific community.
In September, OpenAI had already quietly shaken up mathematics by claiming to have solved, thanks to one of its AI models, the so-called Navier-Stokes equation, in less than four days.
This progress, which is still difficult to measure, raises existential questions within the mathematical community.
Scientists have sometimes been working on these subjects for years, even decades.
Generative AI has progressed very quickly in this area.
Its competitors are not to be outdone: the Chinese RedNote model obtained a perfect score at the International Mathematics Olympiad in Shanghai in July, and that of Anthropic, which produces Claude, verified the proof of Pierre de Fermat's last theorem, in just eleven days, when it had taken 360 years for a human to achieve this.