At the Helmholtz Association, we aim to train Foundation Models and make them available to the scientific community.
AI experts from various Helmholtz Centers have come together to design a concept for the three-year Helmholtz Foundation Model Initiative (HFMI). The initiative‘s goal is to develop fully functional models while considering ethical and legal standards. Seven pilot projects have been selected, involving scientists from thirteen Helmholtz Centers.
A key unique feature of the HFMI is the collaboration between AI experts and domain scientists within these projects. Helmholtz ensures that not only do the models make a meaningful contribution to research, but also that the results withstand empirical scrutiny. The funded projects are expected to provide clear added value to science. Moreover, they adhere to the principle of Open Science, ensuring their final results are accessible to society—including the code, training data, and trained models.
