General Intuition: Transforming Robotics with Video Game-Trained AI Models

Pim de Witte, CEO of General Intuition, is pioneering a shift in the robotics industry by advocating for the use of foundation models instead of building specialized natural language processing models from scratch.

• General Intuition proposes using better quality datasets instead of vast amounts of real-world data, helping transfer intuition about movement and interaction across various environments.
• General Intuition, underpinned by video game data, has developed a foundation model exhibiting human-like intuition in spatial-temporal reasoning.
• This shift reduces the need for extensive real-world data collection, suggesting a mere few minutes is sufficient for model training.
• De Witte’s vision, supported by key investor Vinod Khosla, is demonstrated by their robotic model that only required eight minutes of fine-tuning to function effectively.
• General Intuition raised $320 million last month, emphasizing their belief in leveraging action data from video games to train AI agents, reflecting a valuation of $2.3 billion.
• The company aims not to create physical robots but to provide foundational AI models that other firms can customize, significantly easing the path for future technological developments.