Korean Home-Service Platform Miso Launches 'Miso Motion' to Collect Robot Training Videos from Skilled Workers
Miso leverages its vast worker network to enter the physical AI infrastructure space by collecting first-person perspective video data.
South Korean home-service platform Miso announced on November 3, 2026, the launch of 'Miso Motion,' a physical AI data business. The initiative aims to collect video data of field workers' operations to address the shortage of large-scale egocentric (first-person) data required for training physical AI.
Physical AI refers to intelligent systems capable of controlling robots to grasp, move, and handle objects in real-world spaces. Unlike text or images, training such AI requires extensive video records of hand movements, tool angles, and pauses. Since this data cannot be scraped from the internet, it must be collected through direct filming.
Participants in 'Miso Motion' will wear hat-mounted cameras provided free by Miso to record their work as usual. Target fields include cooking, bartending, beauty, carpentry, repair, and horse care—industries relying on manual skills. Both individuals and businesses can participate and receive compensation.
Miso's decision stems from its existing resources. The company has operated cleaning, moving, and air-conditioning services for nearly a decade, with over 100,000 partners and more than 10 million cumulative orders. The company believes its experience in recruiting nationwide workers and managing service quality translates directly into an advantage for high-quality data collection.
Similar data collection models have emerged overseas, with startups buying such operational videos as humanoid robot development accelerates. However, this has sparked controversy regarding low compensation and the risk that workers might 'teach the robots that replace them.' Miso stated that participants can confirm filming conditions and compensation standards beforehand to respect workers' rights.