AlphaGo Creator David Silver Launches Ineffable Intelligence to Build Superintelligence

A New Bet on Reinforcement Learning

David Silver, the researcher whose AlphaGo program stunned the world in 2016 by mastering the game of Go, has founded a new company called Ineffable Intelligence. The startup aims to build superintelligence by focusing on reinforcement learning, in which AI models acquire new capabilities through trial and error. The goal is to create "superlearners" that surpass human intelligence across many domains.

The strategy sets Ineffable Intelligence apart from most AI companies, which are racing toward superintelligence by scaling up large language models. Silver believes that path will ultimately fail. "As amazing as LLMs are, they learn from human intelligence rather than building their own," he said in an interview with WIRED from his London office. "I think of our mission as making first contact with superintelligence."

Human Data as Fossil Fuel

Silver argues that training systems on human generated data has limits. "Human data is like a kind of fossil fuel that has provided an amazing shortcut," he said. "You can think of systems that learn for themselves as a renewable fuel, something that can just learn and learn and learn forever, without limit."

To illustrate the weakness of the LLM approach, Silver offers a thought experiment: release a large language model into a world that believes the Earth is flat. Without interacting with the real world, the system would remain an avid flat earther even as it improved its own code. An AI that learns about the world by itself, he says, could make genuine scientific discoveries. He compares today's AI to biology before Darwin, when no unified theory explained what life was.

Simulations, Safety, and Funding

The central technical challenge is moving from confined worlds like Go to the complexity of the real world. Silver says the answer lies in placing AI agents inside simulations, where they can learn to achieve goals and collaborate with one another. He declined to detail what these simulations will look like.

The approach may also address alignment concerns. Because agents develop inside simulations, researchers can observe how they behave toward others, including lesser intelligences. Ravi Mhatre, a partner at Lightspeed Ventures, which backs the company, says he pressed Silver on safety and came away convinced the method offers a better route to aligned AI because it depends less on learning from human behavior.

Ineffable Intelligence has raised $1.1 billion in seed funding at a valuation of $5.1 billion, an enormous sum by European AI standards. Silver has recruited top researchers from Google DeepMind and other frontier labs. He also pledged to donate all money he makes from his equity, a sum that could reach billions, to high impact charities.

A Singular Reputation

Silver's track record gives the venture unusual credibility. His mentor Rich Sutton, together with Andrew Barto, won the Turing Award in 2025 for developing early reinforcement learning algorithms. While the field has shifted toward LLMs trained on vast quantities of human text, reinforcement learning remains essential, having shaped chatbot outputs through human feedback and enabled recent advances in math and programming.

Sonya Huang of Sequoia Capital, an investor in the startup, says fewer than a handful of people have done truly foundational AI work and Silver is one of them. Andrew Dai, who worked with Silver at Google DeepMind, praises both his ideas and his management style, noting that he respects other people's opinions and gives researchers freedom. For Silver, the draw is the science itself. "In terms of pure science, I see this as the most important scientific mission that we could possibly go on," he said.

Sources: Wired · Financial Times · Techmeme