A well thought article, as usual, by Stefaan Verhulst, PhD. Personally, I don't think decentralisation is a goal per se. It is context-dependent. In the current European context, I applaud the ambition of the idea of CERN for AI, insofar it means striving for excellence, investing in R&D and not spreading the resources too thin. As Mario Draghi makes clear, Europe lags in top research centres. However, if you look at the most prominent case of AI in science disruption, Alphafold, it was a cross-institutional collaboration between top AI scientists (from Google Deepmind) and domain experts (biologists at EMBL, Cambridge, UK) who hosted the data, and knew them in depth. The same pattern can be seen in other cases: AI in science flourishes through the collaboration of AI experts and domain experts. In this sense, centralizing AI talent is unlikely to be effective. I would rather call for greater investment in excellent research across the board, from life sciences to new materials to climate sciences, and make sure that these top research centres have strong in-house AI capacity. A second question would be about the interaction between supercomputers and research centres. Because if a CERN of AI had to exist, it would normally be associated with a top supercomputer: LUMI in Finland, Mare Nostrum in Barcelona, or Leonardo in Italy. And this is actually happening, in Barcelona, where top scientists such as Francisco Doblas Reyes (climate) and Alfonso Valencia (computational biology) are already using AI to push the frontiers of their research. So in this sense, some concentration is already happening around supercomputers.
🌍 A CERN for AI ? 🤖 👉 A few months ago, the European Group of Chief Scientific Advisors proposed a bold idea: creating a European research organization for artificial intelligence, modeled on the renowned CERN 🏛️. The European Commission quickly embraced the concept, and earlier this month, the Draghi Report reinforced the call for decisive EU action on AI, warning that "the risk for Europe is to be totally dependent on AI models designed and developed abroad." 🚨 🤔 In an earlier reflection, I explored whether a CERN for AI truly makes sense, given the multi-dimensional nature of AI's development and deployment needs. 👉 While CERN has undoubtedly been a global success in advancing nuclear research, it is, in many ways, a product of another era 🌐. Founded in 1954, CERN represented a singular institutional model—one that centralized cutting-edge research in a single location and under one governance structure. However, today's challenges in AI—and more broadly in the era of global digital governance—are far more complex. In our interconnected world, no single institution, no matter how well-funded or groundbreaking, can effectively manage the scale and diversity of AI development alone. 🤔 What’s the alternative? ➡️ Drawing inspiration from Nobel Laureate Elinor Ostrom's theory of "polycentric governance" 🏛️🔗, I suggest we must look beyond a singular European AI institution. ➡️ Instead, we should consider polycentric governance—a model where multiple autonomous, but overlapping, centers collaborate to manage shared resources and advance scientific progress. ➡️ This model encourages distributed collaboration, giving access to diverse actors across geographies and sectors. Such a model could ensure that the best minds, data, and resources from around Europe and the world are contributing to AI innovation in ways that are sustainable and responsible for all 🌱. ✍️ Eager to hear your and other POVs 💻 Toward a Polycentric or Distributed Approach to Artificial Intelligence & Science: https://lnkd.in/ezXxaX_Z 💻 Draghi Report - The future of European competitiveness (Part 2): https://lnkd.in/eg7Tzw3a (Jean-Claude Burgelman, Francesca Bria, David Osimo, Andrea Renda) #AI #ArtificialIntelligence #Europe #CERNforAI #PolycentricGovernance #ElinorOstrom #Innovation #DigitalGovernance #Collaboration #AIResearch #EU #Draghireport