CERN for AI

A proposed intergovernmental, publicly funded institution for AI research and development, modeled on CERN, the European Organization for Nuclear Research, in which member states pool funding, computing infrastructure and researchers. Versions differ in purpose. Research-infrastructure proposals, common in Europe, seek to give universities and smaller states access to frontier-scale computecompute, n.: The raw computing power used to train and run AI, usually measured by how many specialized chips are used and for how long. With data and algorithms, it is a main driver of AI progress. and talent, reduce dependence on a few private and foreign companies, and pursue open science and open modelsopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private.. Safety-oriented proposals envision an international laboratory that would develop the most capable systems, or study their risks, under shared governance and strict security, sometimes combined with limits on private frontier development, as in the 2023 proposal for a Multinational AGI Consortium (MAGIC). Critics note that the analogy is imperfect: particle physics has little commercial or military value, whereas frontier AI is led by well-funded companies and is dual-usedual-use, adj.: Describes technology or knowledge that can be used both to help and to harm: the same AI that finds software bugs for defenders can find them for attackers., raising hard questions about membership, control of model weights and concentration of power.

The ministers endorsed a CERN for AI but could not agree whether it should build frontier models or only provide compute.