CERN and Mistral AI to develop Flare for collision physics — TimesLIVE

The European Organization for Nuclear Research (CERN) and French company Mistral AI have entered into a long-term strategic partnership to create Flare, a foundational artificial intelligence model for scientific research. As TimesLIVE reports, the model will be trained on billions of simulated particle collisions.

A model for collider physics

CERN said Flare is intended to become the first foundational model for collider physics. It will use a principle similar to that used by generative AI large language models, but it will be trained on particle-collision data rather than texts.

The model is planned to be used instead of some of the specialized algorithms currently employed for particle reconstruction and identification. A CERN representative told Reuters that the center has been using AI for about ten years to convert raw experiment data into event images, but several tools are used for this at different stages.

More current news is available on the UA.News Telegram channel Telegram.

Large Hadron Collider data

CERN stores approximately one exabyte of experimental data from the Large Hadron Collider, equivalent to about 1 million terabytes. Following the collider's current upgrade, expected around 2030, the volume of collisions is set to increase tenfold.

CERN Chief Information Officer Enrica Porcari said the partnership is intended to strengthen AI capabilities for the entire physics community and make them available for shared use. The parties did not disclose the project timeline or budget, but noted that such models could also be useful for other sciences in the future.

Mistral AI CEO Arthur Mensch, in an article for Le Monde, described advanced models as necessary for scientific progress. The Paris-based company also unveiled a new open-weights model this month, which it plans to release on October 27.