$1.1B+ · AI + Reinforcement Learning

What they're building

Ineffable Intelligence is building what it calls a superlearner: an AI system that discovers knowledge and skills entirely from its own experience, through reinforcement learning, without relying on human-generated data as its primary input. The distinction matters. Every major AI model today, GPT-4, Gemini, Claude, is trained primarily on human text. That creates a ceiling: the model cannot surpass what humans have already expressed. Ineffable's bet is that reinforcement learning, the paradigm that produced AlphaGo and AlphaZero, can be extended beyond games and into open-ended knowledge discovery. A superlearner would interact with its environment, generate its own hypotheses, test them, fail, adjust, and iterate across domains from motor control to mathematics to scientific reasoning, without needing a human corpus as its training foundation. Silver's ambition, stated in the company's founding materials, is for the superlearner to rediscover and then transcend the greatest inventions in human history: language, science, mathematics, and technology. The company's website uses a capital T on That: "Our mission is to make first contact with superintelligence."

Why this matters

The case for RL as the path to superintelligence is not new, but it has never been better resourced or better timed. Two things have changed. First, LLM scaling is showing diminishing returns. The consensus inside frontier labs is that simply adding more parameters and more data yields smaller capability jumps than it did three years ago. Second, AlphaZero proved the core thesis in constrained domains: a system that learns from scratch, with no human data, can reach superhuman performance. AlphaZero took three days to master chess, shogi, and Go simultaneously at a level that decades of human expert programming could not match. Ineffable's bet is that this paradigm, applied at scale and generalized beyond games, is the path that avoids the human-data ceiling entirely. Silver describes this as "renewable fuel": where human data is a finite resource, self-generated experience is unlimited. The UK government's Sovereign AI Fund making this its second-ever direct equity investment is a geopolitical signal as much as a financial one: the UK is explicitly trying to anchor superintelligence research in London alongside DeepMind.