This is Demis Hassabis’s Crick Insight Lecture, delivered at the Francis Crick Institute in London and posted on the official Google DeepMind YouTube channel in 2022. Hassabis, the co-founder and chief executive of DeepMind, traces the company’s arc from game-playing systems to scientific tools.
He recounts how AlphaGo learned to beat the world’s best Go players through self-play and reinforcement learning, and then how that approach informed AlphaFold, the system that predicts protein three-dimensional structures from their amino-acid sequences. He explains why protein folding was a decades-old grand challenge in biology and how solving it opens the door to faster drug discovery and other research. The talk predates the 2024 Nobel Prize in Chemistry that Hassabis later shared for AlphaFold, so it captures his thinking at an earlier moment.
For a general audience, this is an accessible, firsthand account of how a research lab moved from games to real science. It pairs naturally with the AlphaFold and AlphaGo milestone entries and with his later Nobel lecture, showing the through-line in his own words.