IAS-9 will lead one of four pilot projects from the Helmholtz Foundation Model Initiative

Together with partner institutions of the Helmholtz Association, Forschungszentrum Jülich wants to create a new generation of AI foundation models for science. The projects are pioneering the development of foundation models to take the application of AI in science to a new level. The work is part of the newly launched Helmholtz Foundation Model Initiative (HFMI), which is being funded by the Helmholtz Association with a total of around 23 million euros.

Two of the four pilot projects are being coordinated by Forschungszentrum Jülich. Among other things, the researchers have set themselves the goal of developing the first basic AI model for climate research, which will form the basis for one of the most precise climate and weather models in the world. The use of AI should also help to bring together the enormous wealth of new data and findings in materials research and thus accelerate the implementation of innovative solar cell concepts. In another pilot project with Jülich participation, scientists want to use a new basic AI model to improve our understanding of the global carbon cycle.

Foundation models are AI applications that are able to solve a range of complex problems based on a very broad knowledge base. The well-known chatbot ChatGPT from OpenAI is also based on such a model. Foundation models are significantly more powerful and flexible than conventional AI models and are also suitable for scientific applications. Through targeted training with large amounts of data and the use of generative AI, they are able to understand complex relationships on the basis of learned patterns, generate new relationships and make predictions.

Starting from 01.05.2024, Prof. Stefan Sandfeld will lead the project:

SOL-AI: A symbiotic modular foundation model for accelerating solar energy materials development

Photovoltaics is a key technology for the energy transition. Innovative solar cell concepts must be implemented more quickly to achieve the necessary increase in the global use of cost-effective solar power. Research and development activities in this area are increasing rapidly, resulting in a wealth of scientific publications. However, the sheer volume of data is limiting the implementation of the latest findings. SOL-AI is a foundation model designed to fundamentally reform material informatics in this field. It is able to bring together the diversity of experimental data and results in the field of photovoltaic materials research in order to drive innovation in various areas: from accelerated component development and optimization to the discovery of new solar materials. SOL-AI aims to develop solutions that will have practical relevance for research and industry.

Participating Helmholtz Centers are: Forschungszentrum Jülich (Stefan Sandfeld, Stefan Kesselheim, Christoph Brabec), Karlsruhe Institute of Technology (Pascal Friedrich, Sebastian Krummscheid, Ulrich Paetzold), Helmholtz-Zentrum Berlin für Materialien und Energie (Eva Unger, Kevin Jablonka) and Helmholtz-Zentrum Hereon (Francesca Toma)

[Link to the press release of the Helmhotz Association]

[Link to the press release of the Forschungszentrum Jülich (German only)]