A lecture on AI and risk models. The Delphi-2M model and the Future of Predictive Medicine with use of AI Beyond the Risk Score.

The University of Iceland Centre of Artificial Intelligence and deCODE genetics have started a joint colloquium in statistics and AI. The Colloquium will be held weekly on Tuesdays, from 9:00 to 10:00 in Tjarnarsalur, Sturlugata 8, with the first lecture in six days, November 11th. Our first speaker will be Thor Aspelund, professor of Biostatistics, Centre of Public Health Sciences, Faculty of Medicine. The topic will be: A lecture on AI and risk models. The Delphi-2M model and the Future of Predictive Medicine with use of AI Beyond the Risk Score. This lecture explores the Delphi-2M, an AI model that uses transformer-based architectures to predict the onset of over 1,000 diseases—sometimes decades in advance. Unlike traditional risk scores focused on single outcomes, Delphi-2M models the entire trajectory of a patient’s health using longitudinal data and generative techniques. We will examine its design, the AI approach, statistical methods, validation, and performance, and discuss its implications for clinical practice, synthetic data generation, and the future of personalized medicine. The Nature paper describing the Delphi-2M model: Learning the natural history of human disease with generative transformers | Nature All are welcome.