CEO: VIDEO GAMES BEAT INTERNET FOR AI TRAINING

Large language models excel at processing text but struggle with understanding how objects move through physical space and time. This fundamental gap limits their path toward artificial general intelligence. General Intuition's thesis centers on a simple observation: video games are inherently designed to simulate realistic physics, motion, and cause-and-effect relationships. Unlike unstructured internet data, game environments provide clean, logical systems where actions produce predictable consequences. The startup believes training AI on gaming data could teach models the spatial reasoning and intuitive physics that current systems lack. Video games offer controlled environments with clear rules, physics engines, and visual feedback—elements that mirror how humans learn about the world. This approach represents a departure from the text-heavy training methods that powered recent AI breakthroughs. Whether gaming data can deliver the missing ingredient for AGI remains uncertain, but the company's bet reflects growing recognition that internet-scale text alone may have limitations for developing more generalized artificial intelligence.