Lee Gyu-lee is a business writer at The Korea Times, focusing primarily on IT & telecommunications, the Ministry of Trade, Industry and Energy and KOTRA. Prior to this, she has covered a wide range of cultural news, from film, television and K-pop to lifestyle and fashion.
RLWRLD CEO urges physical AI shift to solve labor crisis in Korea
RLWRLD CEO Ryu Jung-hee, second from left, joins a physical artificial intelligence (AI) session at Amazon Web Services’ AWS re:Invent 2025 in Las Vegas, Dec. 2. From left are Kevin Peterson, chief technology officer of Bedrock Robotics; Ryu; Sri Elaprolu, director of AWS Generative AI Innovation Center; Amit Goel, head of robotics and edge computing ecosystem at Nvidia; and Josh Gruenstein, CEO of Tutor Intelligence. Courtesy of RLWRLD
By Lee Gyu-lee
The CEO of artificial intelligence (AI) startup RLWRLD, Ryu Jung-hee, emphasized the need for physical AI-based robotics to deal with the labor shortage crisis facing the manufacturing sectors of Korea and Japan.
The CEO joined the physical AI trend session at Amazon Web Services’ annual technology conference, AWS re:Invent 2025, held from Dec. 1 to 5 (local time) in Las Vegas, sharing his insight on the future of robot foundation models and industrial automation. Physical AI refers to intelligence directly embedded into machines and production systems.
The panel also included Sri Elaprolu, director of AWS’ Generative AI Innovation Center; Kevin Peterson, chief technology officer at Bedrock Robotics; and Amit Goel, head of robotics and edge computing ecosystem at Nvidia.
“Korea and Japan, as manufacturing-based countries, are rapidly losing skilled labor due to the demographic cliff,” Ryu said.
“Our large enterprise customers feel a sense of crisis that if they fail to transition to automation within five years, their core businesses may disappear.… Physical AI and humanoid robots are realistic solutions to this problem. Now is the last golden time to redesign industry.”
RLWRLD has been collaborating with companies such as SK Telecom, LG Electronics, CJ Logistics and Japan’s KDDI to collect industrial data from real-world operational environments, using its signature 4D+ capture to film workers’ motion in 360 degrees.
“Unlike China, which uses brute force to acquire data by outfitting exoskeletons on thousands of people, it’s a more precise, faster and cost-effective engineering method,” Ryu said.
He also highlighted that unlike text-based large language models (LLMs), which can present incorrect and fabricated information as factual, physical AI operates only within the constraints of physical laws and hardware limitations.
“Robots are fundamentally incapable of performing actions that violate the laws of physics,” he said. “To meet the safety and reliability required in industrial environments, we need a design philosophy that is fundamentally different from that of LLMs.”
Meanwhile, RLWRLD is set to unveil a robotics foundation model that integrates vision, language and action in the first quarter of next year. It leverages a cross-embodiment architecture that maps previously learned motion data onto diverse humanoid and robotic parts.