Nobel Prize lecture: Demis Hassabis, Nobel Prize in Chemistry 2024

Key ideas

In his Nobel lecture, Hassabis explains how games led DeepMind to AlphaFold and sets out three tests for spotting other problems that suit the same AI methods.

From a machine transcript of the audio. Quotes link to the moment they are said.

1. 1

   Use games as a proving ground

   Hassabis says DeepMind started with games because systems can play themselves to generate data, and because a score or a win gives a clear metric. That made it easy to see whether an algorithmic idea was making progress.

2. 2

   Test problems against three criteria

   Hassabis looks for a massively combinatorial search space, a clear objective to optimize, and plenty of data or an accurate simulator. He says many real-world problems fit this profile or can be made to fit it.

3. 3

   Expect many innovations, not one

   Hassabis says AlphaFold 2 needed over a dozen innovations, combining learned components with evolutionary and physical constraints built into the architecture. He says it was key that the team mixed biologists and chemists with machine learning and engineering people.

4. 4

   Ask how far a result can spread

   Hassabis says AlphaFold's speed made it practical to fold roughly 200 million known proteins in about a year. The team open-sourced it and built a free database with EMBL-EBI, after consulting over 30 biosecurity and bioethics experts. He reports that over 2 million researchers have used it.

5. 5

   A hedged view of what comes next

   Hassabis suggests biology may be too complex for a few equations and that AI could be its description language, an era he calls digital biology. He loosely conjectures that patterns found in nature can be modeled efficiently by classical learning algorithms. He also calls AI dual-purpose and says those leading the field should engage government, academia and civil society.

In Demis’s words

What’s available here

• Original source

  Linked. Opens on YouTube.

• Player on this page

  Available. YouTube’s own player, loaded only when you press play.

• Full text or transcript

  Not reviewed. No license to show a text version here.

• Our summary

  Draft. Written by us with AI from a machine transcript of the video’s audio (Whisper) and checked against it. Machine transcripts can mishear; each quote links to the moment so you can hear it. Not yet rights-reviewed.

• Ask questions with AI

  Not reviewed. Off until there is licensed text to work from.