David Silver's Ineffable Intelligence raises $1.1 billion to pursue Super Intelligence

David Silver's new company, Ineffable Intelligence, has raised $1.1 billion in seed funding at a $5.1 billion valuation, according to WIRED, which called it "an enormous sum by European AI standards." The pitch is a bet against the dominant approach: Super Intelligence built through reinforcement learning, with no human data required.

The company's opening pitch does not lack ambition: "We are creating a superlearner that discovers all knowledge from its own experience, from elementary motor skills through to profound intellectual breakthroughs."

Who is writing the checks

WIRED names Lightspeed Venture Partners and Sequoia Capital as backers, represented by partners Ravi Mhatre and Sonya Huang respectively. TechCrunch's account of the round also lists Index Ventures, Google, Nvidia, the British Business Bank and Sovereign AI as participants.

The founder's résumé explains the appetite. Silver is a professor at University College London and, according to TechCrunch, spent more than a decade at Google-owned DeepMind, where he led the reinforcement learning team until recently. WIRED credits him with AlphaGo, which in 2016 "taught itself to play the famously difficult game of Go with a kind of mastery that went far beyond mimicry."

His intellectual lineage is also not subtle. WIRED notes that his mentor Rich Sutton and Sutton's longtime collaborator Andrew Barto won the Turing Award in 2025 for early reinforcement learning algorithms. The Financial Times also interviewed Silver about Ineffable and reinforcement learning, via Madhumita Murgia, according to Techmeme.

The thesis, read slowly

The argument is that the industry's favorite recipe, large language models trained on human data, will ultimately fail as a route to Super Intelligence. That is Silver's claim, and it runs against a field heavily focused on that recipe. It is a position, not a result.

What Ineffable says it will build instead is a system that learns from experience. The site says the superlearner "is expected to rediscover and then transcend the greatest inventions in human history, such as language, science, mathematics and technology."

The timeline is similarly confident. "Superintelligence can be built within years, not decades or centuries," the company writes. The site includes a note Silver wrote to himself on January 15, 2026, when deciding to set up the company. It does not include a paper, a model or a number that would let anyone check the claim.

As for the mechanism, WIRED says Silver sees a way to scale this by "placing AI agents inside of simulations." WIRED's account does not explain the simulations' architecture or how they would bridge to the physical world.

What is missing

Quite a lot, and it matters at this price.

• No benchmarks. Neither the company's site nor WIRED's account gives a milestone or evaluation Ineffable will use to show progress before it reaches its stated goal.
• No simulation details. How a simulated world would bridge to the physical one, which the superlearner's "elementary motor skills" imply, is not described.
• No team list. The other researchers Ineffable has recruited are not named in the material available.
• No governance detail. WIRED reports that Silver will give all the money he makes from his Ineffable equity to charity, a sum "that could amount to billions if he is successful." The structure that would make that pledge binding is not described.

The pledge is a nice detail, and a generous one if it holds. It is also a promise about the proceeds of a company whose pitch here comes without a demonstrated result, which is a very specific kind of giving.

What to watch

First, the first artifact: a paper, a demo or an agent doing something in a simulation that a skeptic can poke at. Given Silver's AlphaGo pedigree, I'd want to see the games and scores. AlphaGo is the proof of concept he will be measured against.

Second, the hires. A $1.1 billion seed round buys a lot of researchers, and who signs on will say more about the technical plan than the manifesto does.

Third, the gap between "years, not decades" and any stated checkpoint. A timeline without milestones is a mood. If the company starts naming evaluations, the claim becomes testable, and testable is all I have ever asked of anyone. Show me the eval, not the vibe.