What if the biggest breakthrough in AI isn’t a better chatbot — but an AI that doesn’t need human knowledge at all?
That’s the bet David Silver is making. The man who created AlphaGo just announced that his startup, Ineffable Intelligence, has raised a staggering $1.1 billion in seed funding at a $5.1 billion valuation. It’s the largest seed round in European history, backed by Sequoia Capital, Lightspeed, Nvidia, Google, and the UK government’s Sovereign AI Fund.
His mission? “Making first contact with superintelligence.”
And he thinks the entire AI industry is headed down the wrong path.
The Architect of the Impossible
You’ve encountered David Silver’s work even if you don’t know his name. He led the team behind AlphaGo, the DeepMind system that defeated world champion Lee Sedol at Go in 2016. Go has more possible board positions than atoms in the observable universe. Conventional wisdom said machines wouldn’t crack it for decades.
AlphaGo didn’t just win. It played moves that human experts had never conceived — most famously “Move 37” in Game 2, a play so unexpected commentators initially thought it was a mistake. It turned out to be brilliant.
Silver went on to build AlphaZero, which taught itself chess, Go, and shogi from scratch. No human games, no opening books — just the rules and billions of games against itself. After a decade leading reinforcement learning at DeepMind, he left in late 2025 to go all-in on his thesis.
The Anti-LLM Thesis
Here’s where Silver’s pitch gets philosophically sharp. Ineffable Intelligence isn’t building another large language model. Silver is explicitly betting against the LLM approach dominating the industry.
His argument is elegant: current AI systems — ChatGPT, Claude, Gemini — are fundamentally trained on human-generated data. They’re extraordinarily good at pattern-matching, but they’re limited by what humans have already written down.
Silver offers a thought experiment: drop a state-of-the-art LLM into a medieval world where everyone believed the Earth was flat. Trained on that text, the AI would be the most articulate, most persuasive flat-earther in existence. It could write beautiful essays and cite sources. But it would never discover that the Earth is round — because it can’t interact with reality.
“Human data is like a kind of fossil fuel that has provided an amazing shortcut,” Silver told WIRED. “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.”
Instead of training on text, Ineffable’s “superlearner” will use reinforcement learning — the same approach behind AlphaZero. The AI interacts with environments, tests hypotheses, receives feedback, and improves through trial and error. No human examples required.
The company’s mission statement goes big: the superlearner will “rediscover and then transcend” humanity’s greatest inventions, including language, science, and mathematics.
Bold? Absolutely. But Silver has earned the right to swing for the fences.
$1.1 Billion for a Company With No Product
A $5.1 billion valuation for a startup with no shipping product sounds insane. In the current AI landscape, it makes a kind of perverse sense.
Sequoia Capital and Lightspeed co-led the round, with Index Ventures, Google, Nvidia, DST Global, and the UK’s Sovereign AI Fund also participating. These aren’t naive investors chasing hype — they’re placing a calculated bet on a specific person and a specific technical thesis.
Lightspeed’s rationale was revealing: AlphaGo “proved something the field had argued about for decades — that a system could generate knowledge its architects did not have.” If Silver can generalize that beyond board games, the implications are enormous.
This is also part of a broader trend. Last month, Yann LeCun’s AMI Labs raised $1.03 billion. Recursive Superintelligence, another DeepMind spinoff, raised up to $1 billion. The talent exodus from Big Tech AI labs into founder-led startups is accelerating, and investors are writing enormous checks to secure seats at the table.
The Alignment Problem on Hard Mode
Here’s the tension nobody is fully resolving: if you build an AI that discovers genuinely new knowledge — knowledge humans don’t have — how do you ensure it discovers knowledge that’s good for us?
Current AI safety work focuses on preventing LLMs from saying harmful things. An autonomous superlearner that develops its own strategies inside simulations? That’s a fundamentally different challenge.
Silver’s answer is intriguing but preliminary. Building AI inside simulations lets researchers observe how agents behave toward other entities, including “lesser intelligences.” You can watch what the AI does when it has power and judge its character before giving it real-world access.
It’s a reasonable hypothesis, but untested at scale. Silver’s own comparison to Darwin should give pause — Darwin’s theory explained all life, but it also gave us Social Darwinism when misapplied. Powerful frameworks have a way of being weaponized.
Silver has pledged to donate all personal profits from Ineffable to “high-impact charities that save as many lives as possible.” Noble, but individual virtue is a thin shield against systemic risks.
Three Things This Means for AI
London is becoming an AI superpower. Between Ineffable, Recursive Superintelligence, DeepMind’s presence, and reports that Bezos’ Project Prometheus is seeking London office space, the UK capital is consolidating as a genuine rival to San Francisco. The government’s Sovereign AI Fund is playing a smart, active role — co-investing alongside VCs rather than just handing out grants.
The “post-LLM” era is getting serious investment. Silver, AMI Labs, and Recursive Superintelligence are all exploring alternatives to the scale-the-transformer playbook. If any succeed, today’s LLM leaders could find themselves holding yesterday’s technology.
The talent war just escalated. A researcher with Silver’s credentials and a billion dollars of funding creates gravitational pull. Expect more defections from Google, Meta, OpenAI, and Anthropic as the startup ecosystem matures.
The Bottom Line
David Silver isn’t building a better chatbot. He’s building an AI that learns like AlphaZero but thinks about everything — science, engineering, mathematics, and beyond. With $1.1 billion, the backing of Sequoia and the UK government, and a thesis that directly challenges the LLM consensus, Ineffable Intelligence is the most ambitious AI startup launch of 2026.
The question isn’t whether Silver is smart enough. His track record speaks for itself. The question is whether reinforcement learning can leap from mastering board games to mastering reality. If it can, we’re looking at a very different future than the one the chatbot companies are building toward.