$2.3 Billion Bet Turns Video Games Into Real-World AI Robot Trainers

General Intuition just secured a massive funding round that highlights a bold new approach to building smarter artificial intelligence.

The company raised 320 million dollars at a 2.3 billion dollar valuation by using video game data to train AI agents.

Game Footage Builds Smarter World Models

This method stands out because it draws on millions of hours of labeled gameplay from the Medal platform to teach spatial reasoning.

Unlike many competitors that rely only on video, General Intuition captures exact player actions like button presses to create better understanding of cause and effect.

The same AI brain that navigates Fortnite-style environments can also control a real quadruped robot after just minutes of fine-tuning.

Such transfer from games to physical robots could speed up development of helpful machines in factories or disaster zones.

Investors including Khosla Ventures and Jeff Bezos see this as a scalable shortcut that avoids the high costs of collecting real-world data alone.

Ethical Choices Shape AI's Human Impact

Founder Pim de Witte brings a unique background from humanitarian work that leads the company to ban uses that could harm people.

This stance contrasts with some Silicon Valley trends and aims to keep AI focused on positive applications like search and rescue.

The company also launched the Nerve platform to let gamers earn money by labeling data or operating robots remotely.

This creates new opportunities for a generation facing job changes from advancing AI technology.

Overall the work points toward a future where AI agents handle complex real tasks more naturally and responsibly.