David Silver built the AI that stunned the world by mastering Go. Now he’s betting billions that the path to superintelligence lies not in vast troves of human text, but in machines that learn through raw experience.
Silver, who led Google DeepMind’s reinforcement learning efforts behind AlphaGo and AlphaZero, has launched Ineffable Intelligence in London. The startup just closed a $1.1 billion seed round at a $5.1 billion valuation—Europe’s largest ever. Backers include Sequoia Capital, Lightspeed Venture Partners, Nvidia, Google, and the UK’s Sovereign AI Fund, as reported by Wired and Bloomberg.
Three months old. No product. No revenue. Yet investors poured in. Why? Silver’s track record. And his conviction that large language models like ChatGPT hit a wall.
LLMs gobble human data. They remix it brilliantly. But they can’t escape its bounds. Imagine an LLM raised in a flat-Earth simulation. It would parrot the illusion forever, blind to spherical truth without real-world trials. Reinforcement learning flips that script. AI agents interact, fail, adjust—discovering principles anew.
“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.”
From Go Board to Superlearners
Flash back to 2016. AlphaGo faced Lee Sedol, 18-time world champion. It won 4-1. Not by memorizing games. By self-play: millions of simulated matches, refining strategies through rewards and penalties. Move 37 in game two? A human-like feint no pro had seen. Pure invention.
AlphaZero took it further. Zero human games. It learned chess, shogi, Go from rules alone. In hours, it outplayed Stockfish, the top engine. These weren’t parrots. They were originators.
Silver’s mentors, Rich Sutton and Andrew Barto, earned the 2025 Turing Award for RL foundations. Sutton, Silver’s University of Alberta advisor, tweeted excitement over Ineffable: “David Silver’s new $4bn company… will fulfil the promise of the Era of Experience.”
But RL scaled in games. Real worlds? Messier. Simulations become key at Ineffable. Agents pursue goals inside virtual realms, collaborating or competing. Watch how they treat lesser AIs. Benign? Scale up. Malignant? Intervene early. Safety baked in from observation.
“I think of our mission as making first contact with superintelligence,” Silver said. “It should discover new forms of science or technology or government or economics for itself.” (Wired)
And the money? Silver pledges all equity proceeds to high-impact charities via Founders Pledge. Billions potentially. Humility amid ambition.
Why Investors Swarm—and Skeptics Lurk
News broke in January 2026. Silver’s sabbatical from DeepMind stretched into departure. Ineffable incorporated November 2025; he became director January 16, per UK filings (Fortune). Funding talks heated fast. By February, reports pegged $1 billion at $4-5 billion valuation, led by Sequoia (Financial Times).
Sequoia’s Sonya Huang and Alfred Lin co-led, with Lightspeed’s Ravi Mhatre. Nvidia chips the compute. Google invests in its ex-star. X buzzed: “The final nail in the coffin for ‘just scaling transformers,'” posted @GenAI_is_real, echoing Silver’s April 2025 DeepMind podcast where he pushed the “era of experience.”
Yet challenges loom. RL demands massive compute for trials. Sample inefficiency plagues it—far more steps than LLMs’ data slurps. Real-world deployment? Robots stumble. Simulations must mirror physics perfectly, or agents learn wrong lessons.
Silver knows. At DeepMind, RL powered RLHF for chatbots, boosted Gemini math skills via AlphaProof. But pure RL labs are rare. OpenAI flirts via agents; DeepMind dabbles. Ineffable goes all-in. No LLM crutches.
Critics question scalability. One X user dismissed: “reinforcement learning has been researched HEAVILY… nobody has EVER figured out how to get an AI to improve itself once it enters the wild.” Fair point. AlphaGo stayed in Go’s box.
So what’s next? Ineffable recruits DeepMind alumni. Builds agent swarms in sims. Targets discoveries: novel math proofs, protein folds, maybe economics models no human dreamed.
Silver’s pitch resonates because LLMs plateau. Human data exhausts. Hallucinations persist. RL promises autonomy. Endless fuel.
London heats up. Silver’s UCL professor gig anchors talent. UK funds back it. But success? Simulations must evolve. Rewards must align. Or it joins the hype graveyard.
One thing clear. The man who birthed AlphaGo doesn’t chase hype. He chases what works. Trial by error. The ultimate test.
WebProNews is an iEntry Publication