AI-generated digital twins for science

Project results are expected to contribute to all of the following expected outcomes:

• Innovative use of artificial intelligence (AI), in line with the GenAI4EU initiative^[1], and other key EU initiatives, like the Apply AI strategy, is demonstrated to setup and manage highly complex modelling and simulation activities across different research infrastructures (RIs).
• New best practices and guidelines for model generation and setup, including simulation and simulation-observation data fusion, are established.
• Emergence of new applications and use cases, in line with the GenAI4EU initiative and other key EU initiatives, like the Apply AI strategy.

The aim is to use generative artificial intelligence (AI), in line with the GenAI4EU initiative and other key EU initiatives, like the Apply AI strategy, to deliver digital twins of complex real-world systems that advance the state-of-art of European research infrastructures and show transformative potential in their operations. The solutions should pave the way for new methods to conduct research by the RIs through AI generated and powered digital twins. They should propose suitable setups for creation of new digital twins or enhancing existing ones, already available within the scientific communities and relevant research infrastructures and for which an existing baseline on generative Artificial Intelligence models exists, the required data fusions, visualisations and execution of the workflows. The focus is on key impact sectors, such as healthcare (including mental health), transportation, agriculture, environment or manufacturing, or other fields related to the Clean Industrial Deal, in line with the European objectives seeking to leverage AI to address societal challenges, improve public services and drive inclusive economic growth. The proposed solutions, underpinning the provision of improved and advanced future services, should support RIs in new areas of research and/or a wider community of users, including clearly identified and relevant industrial, scientific or policy users and enhance the potential of the RIs in addressing EU’s policy objectives and socio-economic challenges. The proposed solutions should, at least in part, be demonstrated in a relevant environment that can lead to sustainable production and include a cost estimate for the operations.

The proposed work is expected to benefit from recent advances in various AI areas, like Foundation Models (e.g. Large Language Models (LLMs)), Federated Learning and link to relevant European AI initiatives, such as the European AI factories^[2], to demonstrate European capabilities to generate large on-demand impact simulations.

Proposals focusing on digital twins for decision-making in the area of the Earth systems and related socio-economic impacts will need to adhere to the standards and best practices set by the Destination Earth^[3] initiative of the European Commission to allow coupling of new digital twins with the existing Destination Earth system.

The proposals should cover all of the following aspects:

• Demonstrate the ability to setup complex modelling and digital twin setups using generative AI in line with the GenAI4EU initiative and other key EU initiatives, like the Apply AI strategy.
• Ensure a data driven approach in which new simulations and Digital Twins are set up and demonstrate the chosen approach to work in at least three different application scenarios.
• Enable non-technical decision makers to express scenarios for challenges to be translated into impactful simulations, making use of existing digital twins, models and data sources.
• Demonstrate trustworthiness/reliability in the use of AI and practice clear and open communication of AI use and impact to users.

Additionally, the proposals should also include work on some of the following points:

• Best practices to use AI to generate digital twins and their setup.
• Contribute to the European AI Factories and their abilities to facilitate large scale AI setups.
• Demonstrate that proposed solutions work for at least two different scientific domains to highlight common principles and best practices in this approach.

Proposals should take advantage of the opportunities and developments offered by existing Horizon Europe research and innovation actions (RIA) developing new simulation and observation capabilities, the Horizon Europe CSA HORIZON-CL4-2025-03-HUMAN-18: GenAI4EU central Hub, and new and emerging ICT infrastructures (e.g. EuroHPC and the AI-on-demand platform). Successful proposals in the energy-intensive industrial transition field will be encouraged to closely cooperate with the European Commission’s Joint Research Centre (JRC) to foster coordination with on-going related activities.

[1] This call falls under the ‘GenAI4EU' initiative as in the Communication from the Commission to the European Parliament, the Council, the European Economic And Social Committee and the Committee of the Regions on boosting startups and innovation in trustworthy artificial intelligence ((COM(2024) 28 final of 24.1.2024).

[2] https://digital-strategy.ec.europa.eu/en/policies/ai-factories

[3] https://digital-strategy.ec.europa.eu/en/policies/destination-earth