How video games teach AI to act in the real world
New York-based startup General Intuition has raised $320 million at a valuation of $2.3 billion. The company is betting on an unconventional idea: millions of hours of video games could serve as the foundation for training artificial intelligence capable of navigating and making decisions in the real world.
General Intuition was founded in October 2025 by 31-year-old entrepreneur Pim de Wit, the creator of the Medal platform, where gamers publish gameplay clips. The new funding round increased the company's total raised investment to $454 million. The round was led by Khosla Ventures, with investors including General Catalyst, Jeff Bezos, and former Google CEO Eric Schmidt.
Why video games became a source of data for AI
General Intuition's main advantage is access to Medal's data. The platform stores not only game recordings but also information about player actions, including key presses. Unlike many developers who have to guess user intent from video footage, the company receives pre-labeled data on how people make decisions and interact with the virtual world.
Based on this, the startup trains so-called world models—simulations that help AI understand space, objects, and causal relationships. Essentially, the company is trying to teach artificial intelligence to intuitively perceive the environment, using game worlds as a training ground.
From Fortnite to robots
During a demonstration at the New York office, de Wit showed an AI agent that had been playing a game resembling Fortnite for over 100 hours continuously, developing spatial reasoning.
Then, a quadruped robot entered the room. According to de Wit, the same AI that controls the agent in the game also controls the robot. The device navigated using a single camera and independently explored the office space.
It took only eight minutes of real-world data to adapt the model to a specific robot. According to the company, this demonstrates that skills acquired in a virtual environment can be effectively transferred to the physical world.
How the company will spend the $320 million
General Intuition does not plan to build autonomous vehicles or robots itself. The startup's goal is to provide infrastructure and tools that will allow other companies to develop autonomous systems faster.
The majority of the new funding will be directed toward expanding computing power and training the next generation of models. Part of the investment will also go toward launching a more affordable API for clients in the gaming, simulation, and robotics sectors.
Despite interest in robotics and autonomous systems, General Intuition states that it will not participate in the development of technologies intended to harm people. The company allows the use of its models in search and rescue operations but excludes participation in projects related to lethal autonomy. Against the backdrop of growing Silicon Valley interest in the military use of AI, this stance noticeably distinguishes the startup from its competitors.
The Tech on Telegram Follow The Tech on Telegram
Keep up with the latest technology and startup news.