DeepMind's David Silver Launches $1.1 Billion AI Startup Betting on Reinforcement Learning

David Silver, the DeepMind scientist behind AlphaGo's historic 2016 win over world Go champion Lee Sedol, has raised $1.1 billion to launch a startup betting that the next era of AI won't come from today's dominant technology. Silver's company, Ineffable Intelligence, launched in January at a $5.1 billion valuation and is betting on reinforcement learning, a method where AI systems improve through trial and error. Silver argues that approach, rather than the large language models now dominating the field, offers a more credible route to superintelligence. "I think of our mission as making first contact with superintelligence," Silver told Wired. "By superintelligence, I really mean something incredible. It should discover new forms of science or technology or government or economics for itself." Popularized by philosopher Nick Bostrom in his 2014 book "Superintelligence," the term refers to AI that surpasses human intelligence across nearly all domains, while artificial general intelligence, or AGI, describes systems capable of matching human-level reasoning across a wide range of tasks. Silver argues that large language models are fundamentally limited because they learn from human-generated data, instead of building their own understanding through experience. "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." Silver has spent much of his career advancing that argument. AlphaGo, which combined human training data with reinforcement learning and self-play, developed strategies that surprised even top human players and demonstrated how AI can exceed human precedent in narrow domains. "I feel it's really important that there is an elite AI lab that actually focuses a hundred percent on this approach," he told Wired. "That it's not just a corner of another place dedicated to LLMs." Ineffable Intelligence plans to build what Silver calls "superlearners"—AI agents placed inside simulations where they can pursue goals, fail, adapt, and improve without the limits of a static human dataset. Silver declined to describe what those simulations would look like, but said the approach would allow agents to collaborate and develop capabilities autonomously. Silver argued that large language models are limited by the data they are trained on, adding that a model trained in a world where everyone believed the Earth was flat would likely keep that belief unless it could test reality for itself. A system that learns through experience, he said, could discover otherwise.