Sequoia Capital• Apr 29, 2026• 26:46Interview

Demis Hassabis: We're Three Quarters of the Way to AGI

Demis Hassabis•CEO, Google DeepMind

Executive Summary

Demis Hassabis recounts the founding of DeepMind, based on the core insight of combining deep learning and reinforcement learning to pursue Artificial General Intelligence (AGI).
The company's mission is a two-step plan: first, solve intelligence by building AGI, and second, use it to solve grand scientific challenges, a vision validated by the success of AlphaGo.
DeepMind's focus on 'AI for Science' has led to breakthroughs like AlphaFold, which solved the protein folding problem, and is now being applied to drug discovery via Isomorphic Labs, with the goal of drastically reducing development timelines.
Hassabis predicts AGI will be achieved around 2030 and believes AI-driven simulations will revolutionize fields like biology, weather forecasting, and even social sciences.

12 quotes

Concerns Raised

The immense technical difficulty of building AGI and accurately simulating complex biological systems.
The historical risk of being too far ahead of the technological curve, a lesson from Hassabis's first startup.

Opportunities Identified

Revolutionizing drug discovery by reducing timelines from a decade to months or weeks.
Solving grand scientific challenges in biology, material science, and energy.
Creating new fields of science through the analysis of complex AI systems and the use of AI-driven simulators.
Achieving AGI as the ultimate tool to 'solve everything else'.

Key Themes

The Genesis of DeepMind

Demis Hassabis's diverse background in video game design, neuroscience, and computer science converged into a singular vision for DeepMind. The company was founded on the then-unconventional belief that combining deep learning with reinforcement learning was the key to unlocking significant progress towards AGI.

This highlights the power of interdisciplinary thinking and a long-term, conviction-driven approach to founding a company in a field that was not yet mainstream.

AI for Science as the Ultimate Goal

DeepMind's mission extends beyond creating AI; its ultimate purpose is to use that intelligence as a tool to accelerate scientific discovery. This is exemplified by AlphaFold's solution to the 50-year-old protein folding problem and the creation of Isomorphic Labs to revolutionize drug discovery.

This positions AI not just as a software product but as a fundamental new tool for science, capable of solving problems previously beyond human capacity and creating immense societal value.

The Power of AI-Driven Simulation

Hassabis emphasizes the potential of AI to create highly accurate simulations of complex, emergent systems that are poorly described by traditional mathematics. Examples include Google DeepMind's 'Weather Next' model, the development of a 'virtual cell' for biology, and the future possibility of simulating economies.

Accurate, learned simulators could transform R&D, policy-making, and our understanding of complex systems by enabling rapid, controlled, in-silico experimentation that is impossible in the real world.

A Principled, Long-Term Path to AGI

DeepMind was established with a 20-year roadmap to AGI, a goal considered fringe in 2010. The victory of AlphaGo was a key milestone, proving the generality and power of their approach and signaling the time to apply it to real-world science. Hassabis maintains a consistent prediction of achieving AGI around 2030.

This demonstrates the importance of setting audacious, long-range goals and using intermediate, high-profile milestones to validate progress and maintain momentum on the path to a transformative technology.

Get started free

Topics

Artificial General Intelligence (AGI)DeepMind AI for Science Drug Discovery AlphaFold Protein Folding Isomorphic Labs Reinforcement Learning Deep Learning Neuroscience Simulation Computational Biology AlphaGo Video Games Startup Strategy

Processed May 4, 2026 yt-dlp + mlx-whisper + Gemini

You're reading a preview

Get started free →