General Intuition has secured $320 million in funding at a $2.3 billion valuation to develop physical AI models. The startup aims to revolutionize robotics by using massive amounts of video game data to train models capable of spatial-temporal reasoning. This approach allows robots to learn complex movements with minimal real-world data, potentially accelerating the development of autonomous systems across various industries.
How does General Intuition train its models?
General Intuition trains its physical AI models using millions of hours of video game data, which provides the necessary information for spatial-temporal reasoning. By leveraging this data, the company creates models that understand movement and interaction. This allows robots to perform tasks after being fine-tuned on just eight minutes of real-world data, significantly reducing the need for extensive data collection typically required in robotics development.
The future of embodied AI
The company aims to become the base layer for physical AI, similar to how large language models serve as the foundation for text-based applications. CEO Pim de Witte believes that by providing a generalized model, they can make it significantly easier for other companies to build robotics solutions. This shift could move the industry away from specialized, data-heavy training toward more efficient, scalable embodied AI systems.
