Pim de Witte, CEO of General Intuition, believes that robotics is on the verge of a transformation akin to the ChatGPT moment in AI. In a recent interview with TechCrunch, de Witte highlighted how the industry should shift focus from building specialized robot models to developing foundation models that can generalize across various environments. This shift could revolutionize how robots are trained and deployed.
Foundation Models in Robotics
Traditionally, companies have relied on extensive real-world datasets to create task-specific robot models. However, de Witte argues that a better approach is to concentrate on high-quality datasets that enable the creation of versatile foundation models. “A lot of companies right now are doing lots of specialized work focused on individual embodiments, individual environments, and individual robots,” he stated.
He believes that as general models emerge, much of the specialized work will become redundant. General Intuition aims to develop models that possess a base level of reasoning about space and time, significantly reducing the need for extensive real-world data collection.
Investment and Development of General Intuition
General Intuition recently raised $320 million at a $2.3 billion valuation, bolstered by its unique approach to training robots. The company has developed a foundation model by utilizing millions of hours of video game data, which includes information on human interactions with controllers.
