International Collaboration on AI Factories and HPC-AI

Expected Outcome:• Strengthening the current and future capacities of the EU HPC-AI ecosystem, and in particular of AI factories, by supporting the active EU participation in the international initiatives for scientific and engineering massive GenAI HPC-based solutions.

• Delivery of a high-quality plan addressing the development of massive GenAI models for science in Europe.
• Ensuring that the EU’s vision, priorities and ethical standards are well reflected in the discussions, roadmaps, and other technical activities and in the governance of the Trillion Parameter Consortium (TPC) [1].
• Contribution to the development of a competitive European converged HPC-AI ecosystem.
• Aligning the EU and national initiatives and bridging the gaps between EU and international efforts in these domains, and interaction and collaboration with other similar international efforts.
• Improving the sharing of information, best practice and expertise at European and world-level to address critical scientific challenges in these domains and ensuring that this knowledge is appropriately disseminated to key EU initiatives, in particular AI Factories.

[1] Trillion Parameter Consortium (TPC)

Scope:Proposals are invited for a Coordination and Support Action to actively participate in the organisational and technical activities of the Trillion Parameter Consortium (TPC) [1] to guide and prepare European HPC for the convergence of supercomputing and AI in massive Generative Artificial Intelligence (GenAI) models for science. Proposals should demonstrate a clear link with the TPC, aiming at creating extreme - scale state-of-the-art trustworthy and reliable generative AI models and to address and discuss the related key challenges to support the advancing of AI for science using HPC.

Main activities:

• Coordinate and establish a EU-level representation in the governance of the TPC, ensuring the EU’s views in strategic decisions and contributing to EU’s sovereignty.
• Support the organisation and active participation of the EU stakeholders in technical activities such as roadmaps, working groups, dedicated workgroups, etc. of the initiative.
• Analyse the relevant research and operational challenges and produce and maintain high-quality research roadmaps with recommendations for research actions at the European level related to the TPC. Engage with and disseminate the results to the relevant European stakeholders and communities related to the TPC, to EuroHPC actors such as the RIAG, INFRAG, ETP4HPC, BDVA, and to other relevant projects and initiatives such as AI Factories, DARE FPA on RISC-V hardware, EuroHPC Hosting Entities, HPC Centres of Excellence (CoEs), etc.

The action should consist of a core consortium of key European players in the related domains, and should support the participation of individuals from other scientific and industrial players and organisations in Europe that are considered necessary for the success of the goals of the proposal and related to the main activities of the TPC.

Background:

The rapid advances in GenAI, in particular in Large Language Models (LLMs), and the increasing challenges of effectively using exa and post-exascale HPC architectures to meet the demands of novel AI based applications are changing the whole HPC-AI ecosystem. The magnitude of such challenges is fostering an indispensable collaboration with the key stakeholders at world level that are currently gathering their efforts in major activities to tackle those challenges and prepare the future, in particular the TPC [1]. European initiatives, and in particular the AI Factories, must benefit from the active involvement of EU stakeholders to maintain their current and future competences in this fast-moving environment. It is therefore critical that Europe sends a clear signal of coordinated involvement with a support action so not to be just followers in this major initiative.

[1] Trillion Parameter Consortium (TPC)