AI Robots Excel at Folding Towels, It’s the Operators Who Need Improvement

The competition in the field of artificial intelligence is intensifying, particularly with the emergence of robotic assistants. These pioneering systems are being meticulously trained using algorithms based on the tasks performed by real individuals in everyday jobs.

Nilesh Christopher, a reporter for the Los Angeles Times, recently explored the phenomenon of AI “arm farms”in a compelling investigation, shedding light on the inner workings of these innovative robots.

In a notable development last month, robotics firm FigureAI unveiled its latest humanoid robot, known as the ‘Figure 03′.The company claims this model can efficiently handle various household tasks, including dishwashing, laundry operations, and clothes folding.

Following this, 1X showcased their own humanoid robot, named NEO, designed not only for chores but also as a social companion. However, one of its features — the ‘Expert’ setting — raised significant concerns. This option allows a 1X staff member to take remote control of the robot, challenging the common perception of AI autonomy and privacy in home settings.

The Figure 03 puts dishes into a dishwashing machine. — Figure AI

Inside the AI “Arm Farms”Driving Robotics Innovation

Christopher’s investigation highlights the growing trend of AI “arm farms, ”where employees engage in mundane tasks while recording their actions via GoPro cameras. These recordings serve as crucial data for training robotic AI systems to perform similar chores effectively.

For instance, workers at these facilities might repetitively fold towels, generating the necessary data for optimal clothing folding techniques. One such start-up, ObjectWays, has a stringent policy where employees must restart their task if they take longer than a minute or make an error.

Interestingly, many of the workers involved are engineers, unaccustomed to such domestic tasks, and they rotate responsibilities during their shifts.“We often have to delete up to 200 videos due to trivial mistakes in how items are handled, ” an employee from ObjectWays explained.

This practice of human data collection extends beyond basic tasks. Another start-up, Micro1, employs workers equipped with smart glasses to document a variety of daily actions. Meanwhile, FigureAI collaborated with a real estate company to record movements inside 100, 000 homes, aiming to equip robots with the understanding needed to navigate these environments seamlessly.

My latest for @latimes: I travelled to southern India to document the rise of the AI “arm farms”— where young engineers strap GoPros to their foreheads and fold laundry or pack boxes to teach humanoid robots how to do chores.https://t.co/o6dgEtbuEN pic.twitter.com/qyBsBwzJo6

— Nilesh Christopher (@NilChristopher) November 3, 2025

According to Mohammad Musa, founder of Deepen AI, while current robotic operations often rely on a blend of remote control and human-derived data from outside Western regions, this reliance is expected to diminish over time as automation technology improves. Musa noted, “A mix of genuine and synthetic data is presently utilized, stemming from human demonstrations and controlled environments, with the automation gradually changing this dynamic.”

This breakthrough comes on the heels of other remarkable advancements in AI, including a patient equipped with Elon Musk’s Neuralink brain implant who was able to feed himself using a robotic arm controlled by his thoughts. Additionally, an AI-generated music artist recently garnered attention by landing a lucrative record deal, with AI-produced songs rapidly climbing music charts, a clear testament to the ongoing surge in artificial intelligence applications.

Source&Images