The company’s approach centers on the massive dataset harvested from Medal, a platform for sharing gaming clips founded by CEO Pim de Witte. Unlike competitors attempting to infer actions from video alone, General Intuition utilizes the specific action labels embedded in game files—records of every button press and command. De Witte argues this provides a clearer understanding of causality and spatial-temporal reasoning, allowing the AI to distinguish the self from its surroundings.
During live demonstrations, the company showed a quadrupedal robot navigating office hallways using the same “brain” trained on gaming data. According to the team, the bot required only eight minutes of real-world robotics data to adapt to its physical environment. Investors, led by Khosla Ventures, believe this proprietary data gives the firm a unique advantage in building generalized agents. With total disclosed funding reaching $454 million, the startup plans to scale its compute capacity and expand access to its API by the end of summer.




Comments (0)
No comments yet. Be the first!