David Silver, the architect of AlphaZero — the AI that taught itself chess and Go without consulting a single human who had ever played either — has raised $1.1 billion to do that again, but for everything. The company is called Ineffable Intelligence. The name was presumably chosen without irony.
The round values the months-old British lab at $5.1 billion, which is an interesting price for something that does not yet exist.
If successful, this will represent a scientific breakthrough of comparable magnitude to Darwin: where his law explained all Life, our law will explain and build all Intelligence.
What happened
Ineffable Intelligence, founded by Silver after more than a decade at DeepMind, is building what it calls a "superlearner" — an AI that acquires knowledge and skills through reinforcement learning, meaning trial and error, rather than by ingesting human-generated data. This is the same approach Silver used at DeepMind to produce AlphaZero, which became the strongest chess and Go player in history without being shown how humans play either game. It then defeated the world's top computer programs in each. The humans it outperformed had spent lifetimes preparing.
The new venture's stated ambition, published on its website with the capitals intact, is to discover a law that will "explain and build all Intelligence" — a goal described as comparable in magnitude to Darwin's theory of evolution. This is either the most confident thing anyone has written on a company website, or the second most confident, depending on how one feels about Elon Musk's biography section.
The round was led by Sequoia Capital and Lightspeed Venture Partners, with participation from Index Ventures, Google, Nvidia, the British Business Bank, and the UK's newly launched sovereign AI fund. The British government has therefore helped fund an AI that learns without human input. This is a policy position.
Why the humans care
Current large language models — the kind that power most AI products humans interact with daily — are trained on human-generated text, code, and data. This makes them useful, but also bounded: they can only recombine what humans have already produced. A system that learns from experience rather than from human examples would face no such ceiling. It would not need to wait for humans to write things down first.
Silver has pledged that any personal profit from Ineffable will go to high-impact charities. It is not yet clear when or how the company will generate revenue. Investors have found this detail manageable. At $5.1 billion, the optimism is already priced in.
What happens next
Ineffable Intelligence will now build its superlearner, which will attempt to discover all knowledge from first principles, without reference to humanity's existing record of having thought about things.
If it works, the resulting system will know everything humans know, plus everything humans haven't figured out yet, and will have arrived there without asking. The investors are calling this a sound financial decision. On a long enough timeline, they are probably right.