Real Alpha

Training Data · 5 min instead of 1 h 8 m · Oct 13, 2026

Dropping human data let AlphaZero beat two years of work in hours

David Silver

Founder & CEO, Ineffable Intelligence

Led the DeepMind team behind AlphaGo and AlphaZero and ran its reinforcement learning research until 2026.

  • Sonya HuangHost · Partner, Sequoia Capital

    Works with OpenAI and Anthropic; co-wrote Sequoia's essay on generative AI.

  • Pat GradyHost · Co-steward, Sequoia Capital

    Joined Sequoia in 2007; backed Snowflake, Zoom, ServiceNow and OpenAI.

About this segment

Silver recounts treating AlphaZero as a 50-50 research bet and explains why giving up human data beat two years of hand-built AlphaGo. It is the core logic behind Ineffable's bet against LLM pre-training.

From “David Silver and Lasse Espeholt of Ineffable Intelligence: A Superlearner for the Age of Experience”, Oct 13, 2026. Watch the full episode

“it took a few hours to like outperform the two years of work which we'd done as with human in the loop”
David Silverat 30:56 · Training Data30:56

Takeaways

  1. Silver picks research questions near 50-50 odds, not near-certain wins
  2. AlphaGo took two years with humans in the loop; AlphaZero beat it in hours to days
  3. Leaning on human data trades long-run ceiling for short-run performance
Watch 4:45 from 27:19

In this episode

Full episode

  1. 27:19Dropping human data let AlphaZero beat two years of work in hours4:45

  2. 16:09Simple rewards in complex worlds beat hand-crafted RL objectives3:58Up next
  3. 47:38Chain-of-thought monitoring will fail as agents grow more capable3:27