Demis Hassabis on AI, Game Theory, Multimodality, and the Nature of Creativity | Possible
3Digest
This episode of Pivot features a conversation with Demis Hassabis, co-founder and CEO of Google DeepMind, about the future of AI. The conversation begins with an introduction to Demis Hassabis and his journey from a childhood chess prodigy to a leading figure in AI. He discusses the importance of global participation in AI development and the need to avoid limiting it to a small geographical area like Silicon Valley. The conversation then shifts to DeepMind's groundbreaking achievement with AlphaFold, a model that solved the protein folding problem, a long-standing challenge in biology. Demis Hassabis explains the development of DeepMind's approach to AI, which combines deep learning and reinforcement learning, drawing inspiration from the human brain. He discusses the potential of AI to accelerate scientific discovery, highlighting the importance of "root node problems" that can unlock entire branches of knowledge. The conversation then explores the concept of embodied intelligence and whether AI needs real-world experience to truly understand the world. Demis Hassabis discusses the potential of multimodal AI for robotics, arguing that general models that can understand the physics of the world can be applied to robotics with minimal additional effort. He also discusses the future of coding and the potential for "vibe coding," where natural language is used to program computers. The conversation concludes with Demis Hassabis sharing his sources of optimism for the future and expressing his desire for more discussion about fundamental questions about reality. He expresses his hope that AI will lead to breakthroughs in medicine, potentially solving all diseases within the next 15 years.
Outlines
Introduction and AI's Potential to Solve Global Challenges
This segment introduces Demis Hassabis, co-founder and CEO of Google DeepMind, and the topic of the episode: the future of AI. It highlights the importance of global participation in AI development and the need to avoid limiting it to a small geographical area like Silicon Valley.
DeepMind's Breakthroughs and the Future of Scientific Discovery
This segment discusses DeepMind's groundbreaking achievement with AlphaFold, a model that solved the protein folding problem, a long-standing challenge in biology. It also explores the potential of AI to accelerate scientific discovery, highlighting the importance of "root node problems" that can unlock entire branches of knowledge.
Embodied Intelligence and the Future of Robotics
This segment explores the concept of embodied intelligence and whether AI needs real-world experience to truly understand the world. It discusses the potential of multimodal AI for robotics, arguing that general models that can understand the physics of the world can be applied to robotics with minimal additional effort.
The Future of Coding and the Rise of Vibe Coding
This segment discusses the future of coding and the potential for "vibe coding," where natural language is used to program computers. It also explores the benefits and challenges of real-time multimodal AI, which can process multiple types of input, such as voice, text, and vision, simultaneously.
AI's Impact on Medicine and the Potential to Solve Disease
This segment focuses on the potential of AI to revolutionize medicine, particularly in the area of drug discovery and disease treatment. It highlights the importance of having multiple centers of AI innovation around the world to foster diversity of thought and approach.
Optimism for the Future and the Importance of Fundamental Questions
This segment concludes the episode with Demis Hassabis sharing his sources of optimism for the future and expressing his desire for more discussion about fundamental questions about reality. He expresses his hope that AI will lead to breakthroughs in medicine, potentially solving all diseases within the next 15 years.
Keywords
AlphaFold
A deep learning model developed by DeepMind that can predict the 3D structure of proteins, solving a long-standing challenge in biology.
Deep Learning
A type of machine learning that uses artificial neural networks to learn from data, enabling AI systems to perform complex tasks like image recognition and natural language processing.
Reinforcement Learning
A type of machine learning where an AI agent learns by interacting with its environment and receiving rewards for desired actions.
Multimodal AI
AI systems that can process and understand multiple types of input, such as text, images, audio, and video, simultaneously.
Embodied Intelligence
AI systems that have a physical body and can interact with the real world, such as robots.
Vibe Coding
A future approach to coding where natural language is used to program computers, making coding more accessible to a wider range of people.
Root Node Problems
Fundamental scientific problems that, if solved, can unlock entire branches of knowledge and lead to significant advancements.
Q&A
What was the moment that Demis Hassabis went from thinking about chess to thinking about thinking?
Demis Hassabis had an epiphany around the age of 11, realizing that chess, while enjoyable, was not the best use of his mind. He became fascinated by the fact that someone had programmed a computer to play chess and wanted to understand how it was done.
What is the significance of AlphaGo's Move 37 in the context of AI creativity?
Move 37 was a truly creative move in the game of Go, never seen before in thousands of years of human play. It demonstrated that AI systems can go beyond what human programmers know and generate original solutions.
What are the potential benefits of real-time multimodal AI for robotics and other applications?
Multimodal AI can enable robots to understand the world in a more comprehensive way, processing information from multiple sources like vision, audio, and language. This could lead to more sophisticated and adaptable robots that can perform a wider range of tasks.
Why is it important to have multiple centers of AI innovation around the world?
Having multiple centers of AI innovation fosters diversity of thought and approach, ensuring that AI development is not limited to a single geographical area or set of values. This is crucial for ensuring that AI is developed in a way that benefits all of humanity.
What is Demis Hassabis's vision for the future of AI in medicine?
Demis Hassabis believes that AI has the potential to revolutionize medicine, leading to breakthroughs in drug discovery and disease treatment. He hopes that within the next 15 years, AI will help us solve all diseases, making today's medicine seem like medieval practices.
Show Notes
How can AI help us understand and master deeply complex systems—from the game Go, which has 10 to the power 170 possible positions a player could pursue, or proteins, which, on average, can fold in 10 to the power 300 possible ways? This week, Reid and Aria are joined by Demis Hassabis. Demis is a British artificial intelligence researcher, co-founder, and CEO of the AI company, DeepMind. Under his leadership, DeepMind developed Alpha Go, the first AI to defeat a human world champion in Go and later created AlphaFold, which solved the 50-year-old protein folding problem. He's considered one of the most influential figures in AI. Demis, Reid, and Aria discuss game theory, medicine, multimodality, and the nature of innovation and creativity.
For more info on the podcast and transcripts of all the episodes, visit https://www.possible.fm/podcast/
Listen to more from Possible here.
Learn more about your ad choices. Visit podcastchoices.com/adchoices
Table of contents
United States
