David Silver Raises $1.1B to Build Reinforcement-Learning ‘Superlearners’

David Silver has raised $1.1 billion to launch Ineffable Intelligence, a startup founded in January and valued at $5.1 billion. The company will train AI agents using reinforcement learning inside simulated environments rather than relying on human-generated datasets.

Ineffable Intelligence plans to place agents in virtual worlds where they pursue goals, fail, adapt and improve through trial and error and self-play. The company says agents will discover strategies and capabilities by interacting with their environments rather than by ingesting large text corpora produced by people.

Reinforcement learning gives systems feedback on actions and lets them adjust behavior over many repeated trials. Silver argues this approach scales differently from large language models, which learn patterns from human-created content. “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 led the research that combined human examples with reinforcement learning and self-play to develop AlphaGo, the system that defeated a world Go champion in 2016. He says the AlphaGo work showed how systems trained through experience can develop strategies that exceed human precedent in narrow domains.

Ineffable Intelligence uses the term “superlearners” to describe agents that can collaborate and build new capabilities inside simulations. Silver framed the company’s mission as seeking contact with superintelligence and said by that he means systems that could discover new scientific, technological or economic insights on their own.

The company has not released detailed product plans, timelines or descriptions of the simulated environments it will use. Silver declined to describe the technical setup or the specific nature of the simulations. He explained he wanted a lab focused entirely on reinforcement learning rather than treating it as a small part of a larger research agenda.

The fundraising will provide capital to build the research team and run large-scale simulated training. Ineffable Intelligence says it will focus on methods that let systems learn from experience rather than from static human datasets. The company has not announced partnerships or commercial applications.