Teaching Artificial Intelligence to Move in the Real World
Artificial intelligence has made significant progress in mastering various tasks online, but it still requires assistance in learning how humans physically move around in the real world. To address this, a growing global army of trainers is helping AI escape computers and enter living rooms, offices, and factories by teaching it how humans move.
In an industrial town in southern India, Naveen Kumar, 28, works for a startup that creates physical data used to train AI. His job involves folding hand towels hundreds of times, as precisely as possible, while wearing a GoPro camera on his forehead to capture exact point-of-view footage of how a human folds. The carefully choreographed movements aim to capture all the nuances of what humans do, such as arm reaching, fingers gripping, and fabric sliding, to fold clothes.
A robot practices for the 100-meter race before the opening ceremony of the World Humanoid Robot Games in Beijing in August. (Ng Han Guan / Associated Press)
The Rise of Physical AI
Large language models that power chatbots such as ChatGPT have mastered using language, images, music, coding, and other skills by analyzing vast amounts of online data. However, data on how the physical world works, such as the force required to fold a napkin, is harder to obtain and translate into something AI can use. As robotics improves and combines with AI that knows how to move in the physical world, it could bring more robots into the workplace and the home.
Companies like Tesla, Boston Dynamics, and Nvidia are leading the charge in developing the next generation of robots. Nvidia projects the humanoid robot market could reach $38 billion over the next decade. Other companies, such as Encord and Micro1, are working on teaching AI how to act in the real world by having humans guide robots remotely or collecting human demonstration data.
Robots are displayed at
Image Source: www.latimes.com
