Startup turns gamers' inputs into training data for real-world AI

A UK-based startup is collecting player inputs from video games and converting them into training data for AI systems designed to operate in the physical world. The approach treats gameplay actions, like movement and decision-making within simulated environments, as a proxy for real-world navigation and interaction data.

The company believes gaming data offers a scalable, low-cost way to train embodied AI models, such as those used in robotics or autonomous systems, without needing expensive real-world data collection. This fits into a broader trend of startups seeking creative data sources to train next-generation AI models as traditional internet text data becomes scarcer.