NVIDIA widens physical AI work in South Korea with Samsung, LG and Kakao Mobility

NVIDIA said on 9 November it is working with Samsung Electronics, LG Electronics and Kakao Mobility to expand physical AI in South Korea, across semiconductor engineering, robot training and autonomous driving.

With Samsung Electronics, NVIDIA is applying accelerated computing and AI physics to semiconductor materials research and process development. Some quantum chemistry calculations on Samsung's sMOL platform run up to five times faster end to end with NVIDIA's cuEST, and one chip thermal and mechanical analysis that took about 18 hours on a 32-core CPU finished in 41 seconds with GPU acceleration, the companies said.

Kakao Mobility will develop its Level 4 autonomous driving system on NVIDIA's Hyperion reference platform, based on DRIVE AGX Thor, with the Halos safety system. Early development centres on Kia's purpose-built vehicles, combining Kakao Mobility's own self-driving software and operating experience in Seoul's Gangnam district.

LG Electronics and NVIDIA have demonstrated GR00T-based manipulation with LG's CLOiD robot in a real manufacturing environment. Data collected on site is augmented with NVIDIA Cosmos, used to train GR00T models, then deployed on robots running Jetson AGX Thor and evaluated in simulation, in what the companies call a data flywheel.