General Intuition, the New York-based startup building a foundation model that trains AI agents how to move through space and time, is in talks to raise around $300 million, sources familiar with the matter told TechCrunch. Because apparently, teaching machines to navigate the physical world is the next logical step after teaching them to write terrible poetry.
The raise comes eight months after General Intuition spun out of Medal, a platform for uploading and sharing video game clips, with a $134 million seed round. The fresh funds would bring the startup’s valuation up to just over $2 billion, sources say. That’s a lot of money for a company that basically wants to turn your Call of Duty clips into robot brains.
Sources tell TechCrunch General Intuition has secured funds from backers, including Jeff Bezos and Eric Schmidt, as well as existing investors Khosla Ventures and General Catalyst. Because nothing says “we believe in this” like two of the richest men in the world throwing money at a company that trains AI to not bump into walls.
Pim de Witte, who co-founded Medal, founded and leads General Intuition alongside co-founders Eloi Alonso, Adam Jelley, and Vincent Micheli - researchers who bring expertise in world modeling and simulation. The team is essentially saying, “We’ve seen you fail at jumping over lava pits in Super Mario for years, and we’re going to use that data to teach robots how to walk.”
The startup trains embodied AI and world models using Medal’s dataset of 2 billion videos per year from 10 million monthly active users. The startup’s pitch is that such a dataset - unique because it allows AI to learn from interactive, first-person gameplay - is the perfect base to teach machines deep spatial-temporal reasoning, allowing them to perceive, anticipate, and interact in real time in simulation. So basically, your gaming failures are now the foundation for the robot apocalypse.
That dataset has reportedly attracted the attention of OpenAI, which previously attempted to acquire Medal. And sources say OpenAI hasn’t been the only big AI lab to come knocking. Because everyone wants a piece of the action when you have billions of hours of humans failing to jump, shoot, and drive.
The world model space that General Intuition is playing in is heating up. Startups like Runway, Decart, and World Labs have all recently released world models, and Google’s Genie 3 recently began integrating Google Maps data for more real-world simulation capabilities. It’s a race to see who can make the most realistic sandbox for AI to learn in, with the added bonus of potentially building Skynet.
All of these companies see gaming and robotics training as near-term commercial use cases, but General Intuition takes a different approach: It builds world models to train agents, not to sell them. The agents are the product, and the startup’s unique dataset gives it a path to viability. So they’re not selling the hammer; they’re selling the hand that swings it.
General Intuition will use the funds to scale up its compute capacity so it can release a new product by the end of summer or early fall, according to a source familiar with the matter. Just in time for AI agents to start complaining about the heat.