General Intuition's $2.3B bet that video games can train AI agents for the real world

June 26, 2026 - 04:27

A startup called General Intuition has secured $320 million in fresh funding to expand its unusual approach to artificial intelligence: training AI agents on millions of hours of video game footage. The company argues that the complex decision-making, spatial awareness, and rapid reactions required in games can help AI develop something closer to human intuition, rather than simple pattern matching.

The total bet on this idea now stands at $2.3 billion, according to company disclosures. General Intuition believes that action data from games -- including fighting, racing, and strategy titles -- provides a richer training ground than static text or curated image sets. The logic is that a virtual environment forces an AI to deal with cause and effect, timing, and incomplete information, much like the real world does.

The new capital will go toward building larger simulation engines and hiring engineers who specialize in reinforcement learning. The company plans to license its trained agents to robotics firms, autonomous vehicle developers, and industrial automation companies. The goal is to give machines a baseline of common sense that does not require explicit programming for every possible scenario.

Critics question whether skills learned in a game transfer cleanly to physical reality, where physics, noise, and unpredictability are far messier. But General Intuition's backers are betting that the sheer volume of gameplay data -- combined with advances in neural network architecture -- will bridge that gap. If they are right, the line between playing a game and navigating the real world may start to blur.