The realm of artificial intelligence (AI) is expanding beyond online applications into actual manufacturing environments. Physical AI, which can perceive its surroundings and make autonomous decisions, is emerging as a new competitive edge in the manufacturing sector.
Choi Jae-sik, CEO of Inijee and a distinguished professor at KAIST, emphasized the growth potential of physical AI in manufacturing during a keynote speech at the 18th Good Growth, Good Jobs Global Forum (2026 GGGF) held at The Plaza Hotel in Seoul on September 3.
Choi identified two main pillars for implementing physical AI: robotics and manufacturing. He cited a recent comment by Jensen Huang, CEO of NVIDIA, who noted that South Korea is poised to lead the global physical AI era based on its software, AI, and manufacturing capabilities, highlighting the competitive strengths of the domestic manufacturing industry.
“The system is evolving from generative AI, exemplified by ChatGPT, to physical AI that perceives, judges, and acts in the physical world,” Choi explained. “High-quality real-time data, which serves as a vital resource for AI learning, and the manufacturing infrastructure that can be immediately applied are South Korea's core assets.”
Robots, a key component of physical AI, are undergoing continuous advancement. While past robots merely repeated actions taught by humans, recent developments have enabled them to perceive their environments and learn through experience. Choi stated that these robots must evolve into general-purpose robots that understand objectives and perform tasks independently. He emphasized that a general AI engine, large-scale learning data, and advanced physical bodies must operate as a closed loop.
“One of the most important factors for robots to think and gather experiences to act is collecting data for physical actions,” he said. “While lower levels of intelligence in physical AI are already being applied in the field, high-level intelligence that considers the entire organization remains in the research and development stage.”
Choi predicted that the transition to AI in manufacturing will follow a structure similar to the development stages of autonomous driving in vehicles. Just as driving assistance progresses from Level 1 to full automation, manufacturing AI is expected to evolve from monitoring and diagnostics to decision-making and automated execution, ultimately achieving closed-loop autonomous operations.
He mentioned that sectors such as steel, cement, and petrochemicals are suitable for applying manufacturing AI. “In the manufacturing process, we achieved a 95% accuracy in predicting temperature and environmental conditions, resulting in a 5% reduction in energy consumption,” Choi noted, adding that it is essential to gradually expand the areas of application.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.