As the race toward artificial general intelligence continues, some leaders in AI are looking beyond text-heavy web data. Pim de Witte, CEO of General Intuition, argues that video games may offer a richer training ground because they capture motion, timing, and cause-and-effect in ways language models alone cannot.
General Intuition, a New York-based startup backed by Bezos and now valued at $2.3 billion, has raised $320 million in fresh funding. The round drew support from investors including Coatue, Eric Schmidt, and researchers connected to MIT and Google DeepMind.
The company's approach centers on world models trained with gaming data, aiming to help AI systems better understand how objects and agents behave in physical environments. That capability is increasingly seen as important for the next generation of physical AI, from robotics to real-world decision systems.
General Intuition emerged from Medal TV, a gaming platform that helped shape its data strategy. The company is now positioning itself at the intersection of entertainment, machine learning, and spatial intelligence, where interactive environments can become a blueprint for smarter AI.
De Witte also points to the need for clear ethical boundaries as such systems evolve, especially if the technology is later adapted for sensitive uses. For now, the bigger story is how game worlds may help machines learn the logic of the real one. In the years ahead, this shift could redefine how AI is trained to understand movement, context, and action.