David Silver, creator of AlphaGo, launches Ineffable Intelligence to build superintelligence through reinforcement learning rather than large language models

Silver founded Ineffable Intelligence after leaving Google DeepMind, raising $1.1 billion in seed funding at a valuation of $5.1 billion. He plans to give all money made from equity in the company to charity.

The company focuses on reinforcement learning—where AI models learn through trial and error—rather than the large language model approach most AI companies use. Silver argues LLMs are limited because they learn from human text (like a 'fossil fuel') instead of learning independently from the world.

Silver proposes placing AI agents inside simulations to scale from confined worlds like Go to real-world complexity, allowing agents to learn goals and collaborate. He believes this method could help ensure resulting AI behaves in ways aligned with human values.