Large language models like ChatGPT and Claude have mastered text, but they struggle with a fundamental part of intelligence: understanding how objects move through space and time. That limitation is a major roadblock for artificial general intelligence, or AGI. Now, a New York-based startup backed by Jeff Bezos thinks it has found a better teacher for AI—video games.

General Intuition, valued at $2.3 billion, just closed a $320 million funding round from Coatue, former Google CEO Eric Schmidt, and researchers at MIT and Google DeepMind. CEO Pim de Witte argues that gaming data provides richer, more structured examples of physical interactions than the messy, unlabeled content scraped from the open internet. His company trains what it calls "world models"—AI systems that learn how physics, cause and effect, and spatial relationships actually work by watching gameplay.

The startup spun out of Medal TV, a platform for capturing and sharing gaming clips, giving it access to millions of hours of gameplay data. De Witte appeared on TechCrunch’s Equity podcast to explain why this approach could be the next leap for physical AI—machines that can navigate and manipulate the real world. He also addressed the ethical questions that come with building models powerful enough to be used in defense applications.

Subscribe to Equity on YouTube, Apple Podcasts, Overcast, Spotify, and all major podcast platforms. Follow @EquityPod on X and Threads.

Source: TechCrunch