Collecting robot training data is dirty, unglamorous work. Some AI labs are already paying XDOF to do it.

Physical AI faces significant data challenges that hinder its ability to achieve milestones similar to those of large language models. Effective mechanisms for data collection and analysis are essential for these systems to learn and adapt. Without addressing these challenges, the advancement of physical AI may lag behind that of large language models.