Ineffable Intelligence is a new London-based startup that is leading the largest seed funding round ever recorded in Europe. Specializing in the concept of ‘superhuman intelligence’, its foundations represent a radical departure from conventional AI development, betting that current approaches will not achieve true artificial general intelligence.
Ineffable Intelligence is raising $1 billion in an operation led by Sequoia Capital with a total valuation of $4 billion. NVIDIA, Google and Microsoft are rumored to be considering investing. When tech giants team up with elite venture capital firms for a pre-revenue startup, you know something unusual is happening. And this company promises, promises something different and more ambitious.
It is led by the illustrious, former DeepMind chief scientist David Silver, who spent a decade at the Google company helping to develop AlphaGo and AlphaStar, AI systems that outperformed the world’s best players in games like Go and StarCraft. He also contributed to Google’s family of large language models, Gemini, considered the most advanced in the sector. His credentials are not theoretical. They are proven in systems that achieved what experts thought impossible.
Superhuman Intelligence
Silver believes that to achieve superintelligence, AI systems will have to completely discard human knowledge and learn the basics, through trial, error, and self-learning, like AlphaGo learned to play Go by competing against itself millions of times. This is not a minor technical adjustment, it is a complete paradigm shift from how other companies like OpenAI, Anthropic or Google itself approach AI development.
The former scientist is not proposing incremental improvements to assistants like ChatGPT. He simply argues that the entire industry took a wrong turn. This is what sets Ineffable Intelligence apart from the crowded field of competing superhuman intelligence startups. Your article «Era of Experience» argued that AI systems trained on human data are reaching a plateauand that progress driven solely by supervised learning is evidently slowing down.
Traditional large language models learn by predicting the next word in massive text corpora. They become sophisticated pattern finders. Reinforcement learning works differently. In this branch of machine learning, AI systems are trained to make decisions by interacting with an environment and receiving rewards or penalties based on their actions, allowing agents to learn through a process of trial and error.
The solution proposed by David Silver is let AI systems learn from experience. Not from curated data sets or human feedback, but from direct interaction with environments over time in a way similar to how humans learn complex skills: We don’t memorize instruction manuals. We experiment. We failed. We adapt. We try again.
AlphaGo demonstrated this spectacularly. The result was a system that executed movements no human being had ever conceived, some of which initially seemed like mistakes, but turned out to be brilliant. Human experts could not explain these strategies because they arose from purely experiential learning, not human knowledge.
Silver believes that That path will lead to authentic superintelligence, Superhuman Intelligence.. Experience will become the dominant means of improvement and will ultimately eclipse the scale of human data used in current systems. If true, this will redefine everything we know about AI development trajectories. It remains to be seen, but the general artificial intelligence investment community is betting that it may be right and that is why it is leading the largest initial financing round in Europe.
