General Intuition raises $320M for gamer-data bet on real-world AI agents (runtimewire.com)

General Intuition has successfully raised $320 million, led by Khosla Ventures, to advance its ambitious goal of training AI agents capable of navigating both virtual and real-world environments using a proprietary archive of gamer behavior from its predecessor, Medal. With a valuation of $2.3 billion, this funding aims to enhance AI model training through action-labeled gameplay data that records player inputs along with video, offering a richer context for understanding human behavior. This could address a significant gap in robotics, where conventional models often rely solely on video observations. The company's unique approach, which contrasts with other AI models that infer actions from visual data alone, seeks to leverage billions of gameplay clips uploaded annually to generate models that learn not only from scenes but also from intentional actions and their consequences. With commercial APIs already in use in game and robotics sectors, General Intuition is focused on creating agents rather than just training environments—a strategy that positions it competitively in the AI landscape. However, the real challenge lies in transferring learning from game simulations to practical applications in robotics and autonomous systems, an outcome yet to be conclusively demonstrated at scale.