As the era of Physical AI opens, in which artificial intelligence gains a robot 'body,' the know-how of factory workers is emerging as valuable training data for AI. Bloomberg reported on the 13th, local time, that robot companies in the United States and China are collecting production videos from Indian workers on a large scale to secure training data. Workers wear smartphones or small cameras on their heads while working, filming their hands from a first-person perspective. In shoe factories, tasks such as tying laces and cutting materials become AI training data. In logistics warehouses, procedures such as packing goods and attaching labels are also used. India has emerged as a 'robot data mine' because many tasks that are already automated in advanced economies are still performed by people there. Tesla, Figure AI, and Agility Robotics in the United States, along with Unitree Robotics, AgiBot, and UBTECH in China, are competing to develop robots, but making them work like humans in real factories and homes remains difficult. The problem is data. Large Language Models (LLMs) such as ChatGPT were able to learn from the vast amount of text and images already available on the internet. By contrast, moving robots need data that shows how people in the real world grasp, fold, stack, weld, and assemble objects. First-person work videos shot in India are traded for more than $7 per hour even without processing. If each action is annotated, the price rises to more than $15 to $30 per