Post by Alexander Löser

Europe is striving for sovereignty in AI, particularly in the areas of intelligent agents and language models. I believe we have a strong opportunity to catch up with leading open-source efforts by focusing on three key levels: foundational models and data, adapting models to specific domains, and building deep value chains at the application level. There is still a lot of work ahead, but Europe has what it takes, including talent, funding, GPUs, and access to valuable data. Promising technologies for modularity such as small, specialized reasoning models, ensembles, or mixtures of experts can also help address challenges like siloed thinking and the limited infrastructure or training available for developing large base models. Thank you to acatech – Deutsche Akademie der Technikwissenschaften and Birgit Obermeier for this interview. Joachim Koehler Nicolas Flores Herr Daniel Porta Simon Ostermann Joerg Bienert Andreas Weiss Hanke Michael Moritz Augustin Patrick Bunk Dr. Dagmar Dirzus Volker Tresp Andreas Hotho Kristian Kersting Stefan Dietzel Martin Genzel Hans Uszkoreit Prof. Dr. Feiyu Xu Mehdi Ali, PhD Wolfgang Nejdl Dr. Jan Philipp Harries Johannes Bubenzer Thomas Niessen Maria Ernst Berlin University of Applied Sciences Berlin (BHT)