Google DeepMind CEO Demis Hassabis on AI, Creativity, and a Golden Age of Science

⬅️ Back to Podcasts

  • Podcast: All-In Podcast
  • Hosts: Chamath Palihapitiya, Jason Calacanis, David Sacks, David Friedberg
  • Guest: Demis Hassabis — CEO of Google DeepMind, Nobel Prize winner
  • Duration: ~31 minutes
  • Listen: Apple Podcasts | YouTube

Demis Hassabis, fresh off his Nobel Prize for AlphaFold, on the state of AI and where it is heading.

DeepMind as Google’s Engine Room

Hassabis describes DeepMind as “the engine room” of Alphabet, with roughly 5,000 people building models like Gemini that are integrated across Google’s products. The mission: build cutting-edge AI and deploy it to billions of users immediately.

The Genie World Model

Genie creates interactive 3D environments from text prompts — no traditional rendering engine. It “reverse engineers intuitive physics” by learning from millions of videos. Users can control fully generated worlds in real time with the arrow keys. Consistent physics. Persistent objects. All pixels generated on the fly.

AI Cannot Yet Be Truly Creative

Current AI cannot come up with a new conjecture or hypothesis. That is the key missing ingredient for AGI. Hassabis pushes back on claims that today’s models have “PhD-level intelligence.” They have some PhD-level capabilities but are not generally capable. Consistency across domains, reasoning, and continual learning are still missing.

AGI in 5-10 Years

Not the 2-3 years some predict. The gap between narrow AI and general intelligence is larger than it appears.

AlphaFold to Isomorphic Labs

AlphaFold solved a 50-year grand challenge in biology. Isomorphic Labs is extending that to drug discovery — aiming to reduce timelines from years to weeks or days. The approach combines machine learning with known chemistry and physics.

The Scientific Renaissance

Hassabis’s vision: achieving AGI will usher in a “new golden era of science” that transforms energy, health, and every other domain. That is why he has spent his entire career on AI.

Co-Creation, Not Replacement

The future of creative industries will be co-creation between human visionaries and AI tools. Top creators will produce higher quality work using the same tools as everyone else. Creatives will be supercharged, not replaced.

Efficiency Gains

Model efficiencies are improving 10-100x over two years. AI will ultimately contribute more to solving energy and climate problems than it consumes.

Crepi il lupo! 🐺