His research in machine learning covers risk management, safety methods and evaluations, LLM honesty, health applications, and data selection for large-scale deep learning. Across these areas, his publications as lead author have been covered by TV and newspapers like The Guardian, Time, etc, while others have been cited by ministers or incorporated into national legislation.
Previously, Sören completed his PhD in machine learning at the University of Oxford under Yarin Gal funded by Google DeepMind, and worked on learning human preferences and game-theoretical machine learning with David Duvenaud and Roger Grosse at Toronto’s Vector Institute and UC Berkeley, and with the Centre for the Governance of AI at Oxford. He has completed additional degrees in machine learning (UCL), mathematics (Amsterdam) and Future Planet Studies (Amsterdam).