General Intuition raises $320M to use video game data to train robots

General Intuition is testing world models that will act as training environments for agentic models. | Source: General Intuition

General Intuition US Inc. this week raised $320 million in Series A funding. The company said it plans to use the financing to build AI models that can perceive, predict, and act in virtual and physical environments.

While physical AI has become a dominating topic in robotics, General Intuition claimed that it is taking a unique approach. Instead of gathering hundreds or thousands of hours of real-world data or generating simulated data, the company uses billions of gameplay clips uploaded to Medal, a platform that allows users to post gaming moments. Pim de Witte, founder and CEO of General Intuition, also co-founded Medal.

The gaming clips capture humans perceiving an environment and deciding how to move through it. This is what makes it valuable, said the company. Text-based models only provide descriptions of reality, which aren’t enough for training physical AI, it said.

In addition, the videos come with embedded action labels. These record exactly what button a player presses and when, giving General Intuition more information about how players make decisions.

The New York-based company said its Series A brings its valuation to $2.3 billion. The round also brings its total funding to $454 million, which comes after the $134 million it raised in October. General Catalyst led the round, which also included participation from Amazon founder Jeff Bezos and former Google CEO Eric Schmidt.

General Intuition quickly gains momentum

Many of the most powerful foundation models are trained on written words. However, General Intuition said human intelligence far exceeds language. It emerged over millennia of interaction and exploration, through the endless cycle of intent, action, and consequences across diverse environments, it asserted.

Truly intelligent machines must move from words to worlds, and acquire the capacity to perceive, anticipate, and improvise. They need to obtain a general intuition of reality, said General Intuition.

The company explained that its models learn from unique, action-labeled video datasets across countless environments. It said this diversity leads to uniquely capable agentic systems.

Since its founding in 2015, General Intuition said it has worked to develop action models that decide what action to take, and world models that predict the outcome of actions.

It plans to use the funding to scale its compute capacity and focus on pretraining the next version of its model. General Intuition also hopes to make its API more broadly available this summer, according to TechCrunch.