[Anchor]
Nvidia and LG are strengthening their physical AI alliance. Following last week's meeting between Nvidia CEO Jensen Huang and LG Group Chairman Koo Kwang-mo in the United States, Jensen Huang's daughter has visited LG Electronics' robot training site.
Reporter Jeong Seong-jin explores the reasons behind this accelerated collaboration.
[Reporter]
Following human hand movements, humanoid robots sort items and lift heavy steel plates.
This is LG Electronics' data factory currently under construction in Seoul, also known as a robot training center.
Madison Huang, a senior director responsible for product strategy for physical AI at Nvidia and the daughter of Nvidia CEO Jensen Huang, visited the facility.
This came just four days after Jensen Huang and LG Group Chairman Koo Kwang-mo met in the United States to sign a memorandum of understanding on physical AI cooperation.
The robot training center is the culmination of the robot alliance between LG and Nvidia.
The structure involves LG providing robot bodies and data from manufacturing and logistics sites, while Nvidia supplies robot brain software and synthetic data technology to accumulate robot data, which is then used to advance physical AI technology.
Through this process, the amount of robot data expected to be secured by the end of this year is projected to reach 100,000 hours, equivalent to about 12 years' worth.
[Park Ki-bum / LG Electronics, LG Data Factory Department: By combining high-quality data secured by LG Electronics in the field over decades with Nvidia's solutions, we will complete a world-class data virtuous cycle system.]
While large language models like ChatGPT can utilize vast amounts of pre-accumulated documents and images for learning, physical AI requires the creation of new action data from real robots.
China is already operating dozens of robot training centers across the country backed by government support and private sector leadership, and robot companies in the U.S. and Germany are also building data training facilities.
[Han Jae-kwon / Professor of Robotics, Hanyang University: Physical AI ultimately requires data on human movement. A competition is now underway over how to secure that data.]
To preempt the physical AI market, projected to reach 1 trillion dollars a decade from now, domestic companies face the task of securing independent technological capabilities while strengthening cooperation with Nvidia.
(Video by: Cho Chang-hyun | Video Editing by: Choi Jin-hwa | Design by: Chang Sung-beom | Footage provided by: Nvidia)
※
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.