TED Tech DeepMind's Demis Hassabis on the future of AI | The TED Interview
Mar 15, 2024
Demis Hassabis, founder of DeepMind, discusses the future and controversy of AI. Topics include AlphaGo beating world champions, gaming in tech, and the revolutionary impact of AlphaFold on protein research.
AI Snips
Chapters
Transcript
Episode notes
Early Games Shaped His Path
- Demis Hassabis started programming games as a teenager and worked at Bullfrog, helping create Theme Park and Black & White.
- Those early games used AI for core gameplay and shaped his fascination with intelligence and simulation.
Why Games Are Ideal Testbeds
- DeepMind uses games and simulations as a practical training ground because they allow millions of runs without hardware constraints.
- Simulations speed research by offering rich, controllable environments and benchmarkable rewards.
Pair Representation With Reward Learning
- Combine deep learning for representation with reinforcement learning for decision-making to build agents from raw inputs.
- Use the learned model to imagine futures and narrow the search for effective actions.