David Silver, the mind behind AlphaGo, announced the formation of Ineffable Intelligence, a venture that aims to build artificial general intelligence through reinforcement learning. The startup raised $1.1 billion in seed funding, valuing the company at $5.1 billion. Lightspeed Ventures and Sequoia Capital led the round, signaling strong investor confidence in Silver’s unconventional roadmap.

Silver’s vision departs from the prevailing large‑language‑model (LLM) strategy that dominates the AI industry. Instead of feeding machines massive corpora of human text, his team will let AI agents learn by trial and error inside simulated environments. "Human data is like a kind of fossil fuel," Silver told WIRED. "Systems that learn for themselves are a renewable fuel—something that can just learn and learn forever, without limit."

The approach draws on reinforcement learning, a field pioneered by Silver’s mentor, Rich Sutton, who, along with Andrew Barto, won the 2025 Turing Award. By placing agents in virtual worlds, Ineffable hopes to observe emergent behavior, test alignment, and ultimately scale intelligence without human priors. Silver believes this method can produce "superlearners" capable of scientific discovery beyond human reach.

Safety remains a central concern. Silver argues that simulations provide a sandbox to monitor how AI interacts with other intelligences, allowing developers to spot undesirable strategies early. "We can actually see what kind of behavior emerges from this," he said. Lightspeed partner Ravi Mhatre echoed the sentiment, noting that the reinforcement‑learning route may offer a clearer path to aligned AI than models that simply mimic human text.

Beyond the technical premise, the company’s financial backing underscores market appetite for novel AI pathways. Sequoia’s Sonya Huang highlighted Silver’s track record and the purity of his vision, calling him one of the few researchers who have done truly foundational work. She added that the surge in compute power and sophisticated simulations makes the approach increasingly viable.

Ineffable Intelligence has already attracted talent from DeepMind and other frontier labs. Co‑founder Andrew Dai praised Silver’s blend of brilliance and collegiality, noting that a supportive culture helps retain top researchers. Silver, for his part, pledged to donate any personal earnings from the venture to high‑impact charities focused on saving lives.

While the company’s long‑term goal—to transition from games like Go to the complexity of the real world—remains ambitious, the initial funding round gives it the resources to build large‑scale simulations and develop the next generation of learning agents. As the AI field grapples with the limits of LLMs, Ineffable Intelligence’s bet on reinforcement learning marks a bold, if risky, alternative.

Used: News Factory APP - news discovery and automation - ChatGPT for Business