April 22, 2024 — Together with partner institutions within the Helmholtz Society, Julich Athletic Center We aim to build a new generation of AI-based models for science. A related pilot project is pioneering the development of fundamental models to take the application of AI in science to a new level.
These projects are part of the newly established Helmholtz Foundation Model Initiative (HFMI), which has received approximately 23 million euros of funding from the Helmholtz Association.
Two of the four selected pilot projects are coordinated by Forschungszentrum Jülich. For example, researchers aim to develop one of the first AI-based models for climate research that will form the basis of the world's most accurate weather and climate models.
The use of AI can also help pool vast amounts of new data and discoveries in materials research, thereby accelerating the implementation of innovative solar cell concepts. In another pilot project involving Forschungszentrum Jülich, scientists aim to use a new AI-based model to better understand the global carbon cycle.
Foundational models are AI applications that are built on a very broad knowledge base and can solve a wide variety of complex problems. OpenAI's famous chatbot ChatGPT is also based on such a model. The underlying model is significantly more powerful and flexible than traditional AI models, making it suitable for scientific applications. Through targeted training and use of generative AI with extensive datasets, you can understand complex relationships, generate new connections, and make predictions based on learned patterns.
HClimRep: New climate model captures interactions between atmosphere, ocean, and sea ice
What if we could predict future climate more accurately, faster, and more efficiently? Could we better combat the causes of climate change and reduce its resulting effects? Can the effects of global warming be made clearly visible to everyone? The HClimRep project aims to answer precisely these questions.
Researchers are developing one of the world's most accurate weather and climate models by building one of the first AI-based models for climate research that combines atmospheric, ocean, and sea ice data. Masu. Thanks to extensive training on Europe's first exascale computer, this deep learning model with billions of parameters is now capable of performing complex “what-if” experiments and other modeling tasks about the ocean and atmosphere. Masu.
Participating Helmholtz Centers: Forschungszentrum Jülich, Alfred Wegener Institute, Helmholtz Center for Polar and Marine Research, Karlsruher Institute for Technology, and Helmholtz-Zentrum Hereon
Source: Jülich concentration camp