Training AI through games: General Intuition valued at $2.3 billion

New York-based startup General Intuition aims to open a new chapter in the world of AI and robotics. The company is working on unique technology that teaches robots to navigate the real world using video games. This innovative approach has attracted investor attention, resulting in the startup closing a major funding round with a valuation of $2.3 billion. This is reported by Techcrunch.com reports .

According to TechCrunch, General Intuition raised $320 million in its latest investment round. This brings the company's total funding to $454 million. The startup's founder, 31-year-old Pim de Witte, spinning off from his Medal project, is using millions of hours of video game footage uploaded by gamers to train AI.

From games to the real world: How does the technology work?

The uniqueness of the General Intuition approach lies in the fact that it does not rely solely on video images. While most competitors force AI to learn just by watching video, de Witte's team uses data on button presses and action labels performed during gameplay. This helps the AI model better understand spatial-temporal reasoning.

In a demonstration at the company's office, an AI agent was shown that has been playing Fortnite continuously for 100 hours. Most surprisingly, the same "brain" also controls a four-legged robot moving around the office. The robot explores its surroundings through its single camera and learns to avoid obstacles. Experts note that just eight minutes of real-world data are enough to adapt the robot to a new environment.

Future prospects and robotics

Currently, robots often move like young children, sometimes bumping into chair legs or trash cans. However, the agentic model developed by General Intuition has the ability to quickly bridge the gap between simulation, gaming, and real life. This is expected to enable robots to move freely in complex urban settings or factories without human intervention in the future.

For technologically developing countries, such developments could lay the foundation for creating autonomous robots serving in logistics, agriculture, and security sectors. The use of the video game industry for such serious scientific purposes significantly reduces the costs of training AI.

General Intuition's success shows that experience in the virtual world will soon be reflected in the physical world. The company now plans to further refine its models and integrate them into various types of robotic devices.