Summary
NVIDIA has announced a partnership with British AI startup Ineffable Intelligence to develop next-generation AI systems built on reinforcement learning (RL). Ineffable Intelligence was co-founded by David Silver, the Google DeepMind architect who led the development of AlphaGo and AlphaZero — two of the most significant reinforcement learning breakthroughs in AI history.
The startup focuses on AI models that learn through trial and error from repeated experiences rather than relying primarily on human-generated training data. NVIDIA will provide its Grace Blackwell chips to power the collaborative research, bringing together Ineffable’s RL expertise with NVIDIA’s hardware dominance.
The partnership signals a renewed industry focus on reinforcement learning as a path to more capable AI systems, at a time when many researchers believe the scaling of large language models trained primarily on static data may be approaching diminishing returns.
Sources
Commentary
David Silver is arguably the most important figure in modern reinforcement learning. His work on AlphaGo didn’t just beat the world champion at Go — it demonstrated that RL agents could discover superhuman strategies that no human had conceived. Having him leave DeepMind to start a focused RL company, and immediately landing NVIDIA as a hardware partner, is a strong indicator of where the next wave of AI capability gains may come from.
The timing is telling. The LLM scaling paradigm — more data, more parameters, more compute — is showing signs of plateau. Reinforcement learning offers a fundamentally different approach: instead of learning from static human data, agents learn by interacting with environments and optimizing for outcomes. If Ineffable Intelligence can apply the AlphaZero philosophy to broader domains beyond games, the results could be transformative. NVIDIA backing this with Grace Blackwell silicon suggests they see RL as a major compute growth driver.