Former DeepMind researcher David Silver has raised a huge new round for a fresh AI lab with a very specific bet: the next jump in AI may come from systems that learn through experience, not from simply absorbing more human-written data.
TechCrunch reports that Ineffable Intelligence has raised $1.1 billion at a $5.1 billion valuation. The company is only a few months old, but its pitch is already attracting top-tier money because Silver helped define one of AI’s most important playbooks: reinforcement learning.
What happened
Ineffable Intelligence is a British AI lab founded by Silver, the former DeepMind researcher associated with systems such as AlphaZero. Its aim is to build what it calls a “superlearner” — an AI system that discovers knowledge and skills from its own experience rather than relying mainly on human-generated examples.
According to TechCrunch, 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 Sovereign AI fund. That is a serious syndicate for a company that has barely started operating.
The ambition is not subtle. The company says it wants to create a system that can learn broadly through trial and error, echoing the way DeepMind’s earlier game-playing systems improved by playing against themselves rather than copying human strategies.
Why it matters
Most of the recent AI boom has been built on scaling: more data, more compute, larger models, and better post-training. That approach is powerful, but it has obvious constraints. High-quality human data is expensive, messy, legally contested, and finite.
A successful “learn by doing” model would change the economics. Instead of needing a bigger internet-sized corpus, a lab could create environments where AI systems generate experience, test strategies, and improve from feedback.
That is why reinforcement learning keeps coming back into the conversation. It already helped produce major breakthroughs in games and robotics-style simulations. The unresolved question is whether the same approach can generalise to messy, open-ended real-world tasks.
What this means for the AI industry
The funding shows investors are willing to back alternatives to the large language model scaling race. Ineffable is not alone. Other researcher-led labs are also raising enormous early rounds on the promise that a different training recipe could unlock more capable systems.
It also strengthens London’s position as a serious AI hub. DeepMind’s presence has created a dense alumni network, and the UK government’s involvement through Sovereign AI suggests national strategy is now tangled with AI startup funding.
For companies building with AI, the practical takeaway is simple: the next wave may not just be better chatbots. If reinforcement-learning-first systems work, they could be better at planning, testing, optimisation, and autonomous decision-making.
Our take
The valuation is extraordinary for such a young company, but the thesis is credible enough to watch closely. AI labs are running into limits around data quality, licensing, and model reliability. A system that can learn robustly from experience would attack all three problems at once.
That does not mean Ineffable will pull it off. Learning in controlled games is very different from learning safely in the real world. The risk is that “without human data” becomes a great fundraising story before it becomes a dependable product.
Still, this is one of the more interesting signals in AI right now. The industry is searching for the path beyond brute-force scaling, and Silver’s new lab is making a very expensive bet that the answer is experience.
