Mardi 8 Décembre 2026
- Imprimer
- Partager
- Partager sur Facebook
- Partager sur LinkedIn
Toward the mechanization of science: how AI will dramatically change how scientific research is done
The talk will be followed by a debate on the subject beyond the classical 10mn questions to the speaker so will likely last more than the traditional 1h.
Abstract
Mathematics may be the first discipline to face a disruption that almost no one anticipated. In September, OpenAI announced what it describes as a Lean-formalized proof of the Navier-Stokes Millennium Problem. In October, it released 722 research manuscripts from an unreleased model, reporting progress on hundreds of open problems, some of which several leading mathematicians have described as worthy of a Fields Medal if achieved by a human. Community verification has begun, and if this momentum continues, the pace and volume of new results could soon exceed what humans alone can check. This would be a turning point for how mathematical knowledge is established. Other sciences are likely to face a similar transformation, possibly sooner than many expect. The US Genesis Mission is a nationwide effort to integrate advanced AI models and agents into every step of the scientific workflow. In this talk, I will introduce the Genesis Mission, review recent AI results in mathematics and their impact on the mathematical community, and report on an experiment in AI-based hypothesis generation. Applied to actual research problems, AI-based hypothesis generation makes it possible to explore the hypothesis space, and we observe, perhaps unsurprisingly, that there are many distinct pathways to solving a scientific problem. This challenges practices that science has built around the limited capacity to explore and verify alternatives, from the publication of a single path to discovery to the standards of reproducibility. Babbage mechanized computation. Hamming predicted the mechanization of science. We may be witnessing it.
Biography
Franck Cappello is an IEEE and ACM Fellow and an Argonne Distinguished Fellow. He is leading the Best Practices for Scientific Workflows activity in the DOE Genesis Mission project.
Date et lieu
Mardi 8 Décembre à 14:00
Bâtiment IMAG, Salle séminaire 1 (RDC)
Organisé par
Bruno Raffin
Responsable équipe DATAMOVE
- Imprimer
- Partager
- Partager sur Facebook
- Partager sur LinkedIn