• •TechCrunch reported that XDOF is providing specialized training data for physical robotics companies.
• •The data collection process involves labor-intensive tasks such as recording physical interactions in real-world environments.
• •It remains unclear how many AI labs have contracted with XDOF or how the firm ensures long-term data consistency across diverse robot hardware.

Data collection firm XDOF has begun supplying training data to AI labs aiming to improve physical robotic movement. While Large Language Models have benefited from vast repositories of internet text, robotics developers face a scarcity of high-quality, real-world physical training data. This work is reportedly manual and labor-intensive, requiring humans to physically interact with environments to generate usable logs. Whether this outsourced data model will solve the 'physicality gap' in AI remains to be seen as companies scale their training operations.