Lex Fridman Podcast

#475 – Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games

2372 snips
Jul 23, 2025
Demis Hassabis, Co-founder and CEO of DeepMind, is a trailblazer in AI and a Nobel Prize winner for his protein structure predictions. In their conversation, he dives into the future of AI and its ability to simulate complex natural processes, revolutionizing fields like biology. They explore the evolving gaming landscape influenced by AI, discussing how it can enhance interactivity and creativity. Hassabis also touches on the potential of evolutionary algorithms in AI, the quest for artificial general intelligence, and the ethical implications of these technological advancements.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
INSIGHT

Learnable Structure in Nature

  • Natural systems have evolved structure subject to evolutionary pressures, making them efficiently modelable by classical algorithms.
  • This suggests classical computers can solve complex natural problems without brute force, unlike random or man-made systems.
INSIGHT

AI's Intuitive Physics From Observation

  • VO3 shows advanced physical realism, like fluids and lighting, from passive video observation without interaction.
  • This challenges the assumption that embodied AI or robotics is necessary for understanding intuitive physics.
ANECDOTE

Hassabis's Gaming Roots

  • Demis Hassabis began his career designing game AI, igniting a lifelong passion for open world and simulation games.
  • He envisions future games dynamically generated by AI, evolving uniquely with each player interaction.
Get the Snipd Podcast app to discover more snips from this episode
Get the app