How Human Data Is Teaching AI-Powered Robots Everyday Tasks | Tech It Out

[00:00:00]Now, we hear a lot about humanoids these days. You know, folding clothes, cooking meals, and what have you. But, to do that, they have to learn how to, and they can only do that by watching humans.

[00:00:10]>> Well, and that's created a new kind of job. People are being paid to wear cameras and record their daily routine so AI systems can study [music] their movements. And many of these workers are in India, where AI companies are collecting thousands of hours of videos to [music] train the next generation of humanoids.

[00:00:28]>> Tech tip of the day. If somebody asks you to record a video to train a robot to do your job, say no.

[00:00:39]>> Everyday tasks that humans barely think about, like making a bed or chopping food, are becoming valuable training material [music] for artificial intelligence. Tech companies are now paying people to record these routine activities [music] in the hope of teaching future robots how to replicate them.

[00:00:57]In India's Tamil Nadu, 25-year-old Nagireddy Sriramya Chandra straps a smartphone to her head before starting housework as she slices mangoes, cooks, or carries out other tasks.

[00:01:09]The device records everything she sees.

[00:01:12]The footage is known as egocentric data, first-person video designed to show AI systems how humans interact with the world around them. The recordings are then sent to AI data company Objectways, where they become training material for robots of the future. And there is an irony to the whole process. People are being paid to teach machines how to do the very tasks those machines may one day take over from them.

[00:01:38]>> I see future full of robotics and AI, but still human existence need to be there because we give like without human data and all, those can't run.

[00:01:50]>> Building a chatbot requires vast [music] amounts of text. Building a robot requires something very different.

[00:01:57]Developers need data that show how people move, how they pick up objects, how they open doors, or how they navigate a room without bumping into furniture.

[00:02:06]To collect that information, workers use head-mounted cameras, [music] video glasses, and motion sensing equipment.

[00:02:13]Some record their daily lives at home.

[00:02:15]Others work in factories or specialized studios designed to capture human movement in detail.

[00:02:21]The goal is to help AI models [music] learn how to perceive and interact with real-world environments.

[00:02:31]>> The first two days were a little uncomfortable, but after that, the speed and work precision improved automatically. There is no problem with my speed now. I work normally even while wearing the GoPro camera, and there is no change.

[00:02:48]>> At an object-based studio in Tamil Nadu's Karur, AI trainers spend hours filming themselves performing everyday tasks, holding towels, arranging objects, making beds, or organizing household items. These activities may appear routine to humans, but they represent complex challenges for robots.

[00:03:07]Every movement, every hand position, and every decision provides data that can be used to train AI systems. The company says clients are particularly interested in tasks such as cooking, making coffee, and folding clothes. Skills that future humanoid robots may one day perform in homes and workplaces.

[00:03:27]>> For per day, 90 plus videos are taken.

[00:03:30]Yeah.

[00:03:31]The next day we will change the background.

[00:03:34]Today I sit here and I can face this position means tomorrow I can stand here. Then next day I can stand here. So, each and every angle will be noted, each and every actions will be changed based upon the objects.

[00:03:50]>> The global race to humanoid robots is accelerating.

[00:03:54]Morgan Stanley forecast there could be more than a billion of them in use by the middle of the century.

[00:04:00]But before robots can work alongside people, they must first learn from them.

[00:04:05]For now, that is creating a new category of AI employment.

[00:04:11]With India emerging as a major hub for collecting, processing, and annotating the data that powers these systems. And while the long-term goal is to automate many everyday tasks, the technology still depends on human experience to get there.