On July 16, 2026, Prof. Dr. Tim Hahn (Institute for Machine Learning in Medicine with a Focus on Psychiatry, University of Münster) delivered a guest lecture as part of our seminar on Recent Methods in Medical Informatics. His presentation was entitled "From Predictive Models to Agent-based Systems – Progress, Perils and Perspectives."

Prof. Hahn provided an overview of the research field of Machine Learning in Medicine and illustrated the evolution of AI systems from traditional predictive models to large language models (LLMs), expert systems, and, ultimately, agent-based systems.

Using a range of practical examples, he demonstrated how modern AI methods are already being applied to medical challenges. The lecture featured approaches for analyzing medical imaging data, digital twins, systems for automated clinical documentation, and AI-powered applications for interacting with electronic health records, clinical practice guidelines, and pharmaceutical knowledge.

Another major focus was the development and evaluation of medical AI systems. Topics such as model optimization, quality assurance, generalizability, and clinical applicability were discussed. Prof. Hahn emphasized that, beyond achieving high predictive performance, the reliability, transparency, and safety of AI systems are essential prerequisites for their successful deployment in healthcare.

The lecture also addressed current challenges associated with the use of advanced AI systems in clinical practice. These include hallucinations in large language models, uncertainty estimation, cybersecurity concerns, and the need for appropriate governance structures and regulatory frameworks.

Overall, the lecture offered an engaging overview of both the opportunities and limitations of modern AI technologies. It demonstrated how methodological research and clinical practice can be effectively integrated to advance the responsible use of artificial intelligence in medicine.