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

General Intuition has raised $320 million in new funding, bringing the company's total capital raised to $2.3 billion, to train AI agents using millions of hours of video game footage. The company believes that the complex decision-making and problem-solving required in games—especially fast-paced action games—can help AI develop more human-like intuition. This is a different approach from traditional AI training, which often relies on curated datasets or text-based sources.

General Intuition argues that existing AI models, including large language models, lack true 'grounding' in physical and dynamic environments. By training on action-rich game data, AI can learn to react to unexpected situations, plan under pressure, and act in real time—skills that could transfer to real-world scenarios such as driving, warehouse logistics, or robotics.

Key backers in this latest round include a16z and previous investor Nvidia. The company was co-founded by former DeepMind researchers. Notably, this funding round arrives after General Intuition launched a widely publicized 'MineRL' Minecraft challenge that asked teams to train AI agents to solve open-ended tasks inside the game.