CERN 和法国人工智能公司 Mistral 于 10 月 15 日签署了一份意向声明，旨在建设一个“大型物理模型”，CERN 称其为碰撞物理学的首个基础模型。该模型将基于模拟的大型强子对撞机数据进行训练，以帮助重建和识别粒子。目前尚未披露预算或时间表。
CERN and French AI company Mistral signed a statement of intent on 15 October to build a “Large Physics Model“, which CERN calls the first foundation model for collider physics. It will be trained on simulated Large Hadron Collider data to help reconstruct and identify particles. No budget or timeline was disclosed.
CERN 和 Mistral 计划为粒子物理学构建一个 AI 模型，欧洲实验室在周四表示。双方签署了意向声明，以达成关于科学领域人工智能的长期协议。
CERN and Mistral plan to build an AI model for particle physics, the European lab said on Thursday. The two signed a statement of intent for a long-term deal on AI for science.
该计划的核心是一个“大型物理模型”，CERN 称其为碰撞物理学的首个基础模型。此类模型会在一个巨大的数据集上进行一次训练，随后可用于多种任务。
At its centre is a “Large Physics Model”, which CERN calls the first foundation model for collider physics. Such a model is trained once on a huge dataset and then used for many jobs.
该模型将从大型强子对撞机（LHC）中数十亿次模拟质子碰撞中学习，LHC 是世界上最大的粒子加速器。CERN 表示，该模型应能帮助重建碰撞、识别粒子、标记异常信号，并测试新型探测器设计。
It will learn from billions of simulated proton collisions in the Large Hadron Collider (LHC), the world’s largest particle accelerator. CERN said it should help rebuild collisions, spot particles, flag odd signals, and test new detector designs.
目前，CERN 使用一系列独立工具将原始探测器信号转换为物理学家可用的数据。实验室表示，一个模型可能学习整个链条，从而取代这些工具。
Today, CERN uses a chain of separate tools to turn raw detector signals into data that physicists can use. One model could learn the whole chain instead, the lab said.
“高能物理学提供了一个独特的环境，在此环境中，人工智能可以以极高的严谨性进行开发和测试，”CERN 首席信息官 Enrica Porcari 表示。
“High-energy physics provides a unique environment in which artificial intelligence can be developed and tested with exceptional rigour,” said Enrica Porcari, CERN’s chief information officer.
截至目前，CERN 已存储约一艾字节（即一百万太字节）的 LHC 数据。随着计划于 2030 年左右启动的 LHC 升级，碰撞次数将增加十倍。
CERN has stored about one exabyte, or one million terabytes, of LHC data so far. The number of collisions will rise tenfold with an LHC upgrade due to start around 2030.
对 Mistral 而言，此次合作有助于其在科学与工程领域的布局。这家总部位于巴黎的公司早在今年早些时候就收购了物理 AI 初创公司 Emmi AI，随后推出了一款工程工程产品。
For Mistral, the deal adds to a push into science and engineering. The Paris-based firm bought Emmi AI, a physics AI startup, earlier this year and later launched an industrial engineering product.
“我们共同拥有推动高能物理学前沿的机会，”Mistral 人工智能工程副总裁、Emmi AI 联合创始人 Johannes Brandstetter 表示。
“Together, we have the opportunity to push the frontiers of high-energy physics,” said Johannes Brandstetter, Mistral’s vice president of AI for engineering and a co-founder of Emmi AI.
这项工作将从粒子物理学开始，但双方表示，这些模型以后也可能帮助其他领域。双方均未透露预算或时间表，路透社报道。
The work will start with particle physics, but both sides said the models could later help other fields. Neither gave a budget or a timeline, Reuters reported.