This is Demis Hassabis’s Nobel Prize lecture, “Accelerating scientific discovery with AI,” delivered at Stockholm University and posted on the official Nobel Prize channel. Hassabis shared the 2024 Nobel Prize in Chemistry for protein structure prediction, and the lecture is his own account of how that work came about.
He traces a path from games to science: from DeepMind’s early reinforcement learning systems and AlphaGo to AlphaFold, the model that predicted the three-dimensional structures of nearly all known proteins. He explains why protein folding was considered a grand challenge in biology for fifty years, how the AlphaFold approach worked at a high level, and how the resulting database has been used by millions of researchers.
As the CEO of DeepMind and a laureate speaking about his own prize-winning work, Hassabis is the definitive source for this story. For a general reader, the lecture is an accessible and inspiring example of AI delivering concrete scientific results rather than speculation, and a window into how one lab pursues science as an AI problem.