A.I. Could Solve Some of Humanity’s Hardest Problems. It Already Has.

DeepMind founder Demis Hassabis discusses the development of artificial intelligence, focusing on DeepMind's journey from creating game-playing AIs to solving complex scientific problems like protein folding with AlphaFold. Hassabis explains the shift from logic-based AI systems to machine learning, particularly deep reinforcement learning, which allows AI to learn from data and experience. The conversation explores AlphaFold's creation, its training on existing protein structures, and its ability to predict new protein structures with atomic accuracy. Hassabis also addresses the potential risks of AI, including its use in synthetic biology, and advocates for international cooperation in AI safety research.