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

By Sequoia Capital

Categories: VC, Startup

Summary

Demis Hassabis claims we're 75% of the way to AGI, revealing how he strategically combined deep learning with reinforcement learning—techniques almost never mixed before 2009. His career arc (chess prodigy → game developer → neuroscientist → DeepMind founder) wasn't random but deliberately sequenced to build the exact skills needed for AGI, offering a masterclass in long-term founder positioning.

Key Takeaways

1. Strategic timing beats being first: Aim to be 5 years ahead of your time, not 50. Hassabis learned this when his ambitious 'Republic' game at Elixir Studios (simulating entire countries on Pentium processors in late 90s) failed because it was too far ahead—but when he founded DeepMind in 2009, the timing was right because deep learning had just been invented.
2. Combine adjacent disciplines for breakthrough innovation: Hassabis deliberately built expertise in neuroscience, game AI, and machine learning. When founding DeepMind, he recognized that almost no one had combined deep learning with reinforcement learning—the gap between disciplines was where the breakthrough lived.
3. Use adjacent industries as proving grounds: Hassabis used game development as a 'backdoor' to fund and test cutting-edge AI in the 1990s when AI was unfundable. Games like Theme Park (10M+ copies sold) proved people would engage with AI systems, validating the long-term bet.
4. Recruit believers by identifying overlooked technical trends: In 2009, Hassabis convinced top talent to join DeepMind by identifying that deep learning (just invented by Hinton) was a big deal that 'almost no one had really realized'—he saw signal others missed, making the sci-fi vision credible to high-caliber founders.
5. Long-term optionality planning: From age 15-16, Hassabis had a decades-long plan to build AGI. Every decision (chess, games, neuroscience) was filtered through 'will this help me eventually build DeepMind?' This radical clarity on the end goal enables strategic sequencing of experiences.

Topics

• Deep Learning + Reinforcement Learning Combination
• Founder Timing and Market Readiness
• Cross-disciplinary Innovation Strategy
• Early DeepMind Recruitment (2009)
• Using Games as AI Proving Ground

Transcript Excerpt

Dennis, thank you so much. >> Exciting to be here. Thanks everyone for coming. It's great to be here. >> We're so honored to have you at our chocolate factory. >> Yes, I just heard about that. Yeah, looking forward to the chocolate afterwards. >> Excellent. Well, Dennis, we're going to jump right in. We have one of the OGs in every way. Uh original thinkers, founders, visionaries in all things AI. Uh true believer, true scientist in in Dennis. uh we're going to spend the beginning of the convers...