CERN and Mistral plan collider physics AI model

CERN and Mistral have signed a statement of intent to develop a Large Physics Model for collider research, a proposed artificial intelligence system intended to help analyze particle collisions at the Large Hadron Collider. The agreement was signed October 15 as the first initiative in a longer collaboration on scientific AI. The model would be trained on billions of simulated proton collisions and is intended to support collision reconstruction, particle identification and searches for unusual signals. CERN and Mistral have described it as the first foundation model for collider physics. Its proposed capabilities have not yet been demonstrated in reported performance results. The partners say the system could bring several parts of particle analysis together in a more general-purpose model. Current research uses algorithms to interpret detector recordings and infer the particles produced in collisions; machine-learning tools already contribute to some parts of that work. The planned training inputs are simulated events, distinct from experimental records collected by CERN. CERN holds about one exabyte of data from the LHC, equivalent to roughly one million terabytes. The...