Deepmind veteran David Silver raises $1B seed round to build superintelligence without LLMs

2026-02-19 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Entrepreneurship & Start-ups, Emerging Technologies & Innovation · Depth: Advanced, short

Summary

David Silver, a prominent UK AI researcher and former Google DeepMind employee, is raising approximately $1 billion for his new London-based startup, Ineffable Intelligence. This seed round, led by Sequoia Capital, would value the company at around $4 billion, making it the largest seed financing ever for a European startup. Ineffable Intelligence aims to develop "superhuman intelligence" using a fundamentally different approach than current large language models (LLMs). Instead of relying on massive text datasets, the company's AI systems will learn through reinforcement learning and continuously improve via trial and error using "world models," mirroring how humans and animals adapt to their environments. Silver, a key contributor to AlphaGo, AlphaZero, and MuZero, believes LLMs are inherently limited by their reliance on human knowledge.

Key takeaway

For AI Scientists and Research Scientists exploring pathways to advanced AI beyond large language models, you should investigate reinforcement learning and world models as a core architectural approach. This strategy, championed by DeepMind veterans, suggests a shift from human-data-centric training to continuous, experience-driven learning, potentially offering a route to self-discovering, superhuman intelligence that current Transformer architectures may not achieve.

Key insights

Reinforcement learning with world models offers a path to superintelligence beyond current LLM limitations.

Principles

• Experience is the dominant medium for AI improvement.
• AI agents should continuously adapt to their environment.

Method

Ineffable Intelligence's AI systems learn through reinforcement learning, autonomously improving via trial and error using internal "world models" to predict action consequences, rather than training on static text datasets.

In practice

• Develop AI agents that learn from continuous experience.
• Implement world models for predictive action consequences.

Topics

• Reinforcement Learning
• World Models
• Superintelligence
• AI Startup Funding
• Ineffable Intelligence

Best for: AI Scientist, Research Scientist, AI Researcher, Investor, Entrepreneur

Related on AIssential

• Partnering with Ineffable Intelligence: A Superlearner for the Era of Experience — Superintelligence may emerge from AI systems learning purely through self-experience, unconstrained by human data.
• Critique of World Model — World Models must simulate actionable possibilities for purposeful reasoning, requiring multimodal, mixed-representation, generative architectures.
• AI's Secret Footprint Exposed — Reinforcement learning, focused on goal-driven interaction with the world, is foundational to true intelligence, unlike large language models.
• Why the Next Leap in AI May Depend on Learning by Trial and Error — Deep reinforcement learning offers AI a crucial mechanism for learning through action and discovery, moving beyond mere pattern recognition.
• DeepMind’s David Silver just raised $1.1B to build an AI that learns without human data — Reinforcement learning can enable AI to discover knowledge and skills independently of human data.
• Dertouzos Distinguished Lecturer: Richard Sutton — OAK proposes a domain-independent AI architecture that learns abstractions and skills from experience, scaling with computation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.