■ Byeng Dong - Youn, CEO of onepredict
Over 99% of on-site data is discarded without being utilized.
Limitations include a lack of failure/defect data and contextual information.
The AX gap between large corporations and SMEs is continuously widening.
"We must capitalize data, connect AI agents, resolve the AX gap,
and make the 'AI Native Factory' a new export industry."
While Artificial Intelligence (AI) is changing the competitive landscape of the manufacturing industry, over 99% of the data generated at domestic manufacturing sites is discarded after a single use. As the gap in manufacturing AI capabilities leads to differences in productivity and competitiveness, it is pointed out that on-site data must be transformed into AI assets and production processes redesigned around AI. Ultimately, to maintain competitiveness as a manufacturing powerhouse, building an 'AI Native Factory' and resolving the manufacturing AX (AI Transformation) polarization are urgently needed.
In an interview held on the 26th at onepredict in Banpo-dong, Seocho-gu, Seoul, Byeng Dong - Youn, a professor of Mechanical Engineering at Seoul National University, emphasized, "Although Korea is a global manufacturing powerhouse, there is a severe lack of 'usable manufacturing data' for AI to learn from. If we cannot transform on-site data into usable assets, the future of our manufacturing powerhouse cannot be guaranteed." Professor Youn received his Ph.D. in Mechanical Engineering from the University of Iowa, served as an assistant professor at the University of Maryland, and joined Seoul National University in 2010, where he currently heads the SNU Industrial AX Center. In 2017, he founded the industrial equipment predictive maintenance company onepredict, advancing AI-driven factory solutions and pushing for a KOSDAQ listing next year.
Professor Youn cited the lack of failure and defect data, the disappearance of skilled workers' tacit knowledge, and data disconnection across processes as the biggest challenges for manufacturing AI. Because factories have a high proportion of normal operations, there is a lack of anomaly data for AI to learn from, and even when anomalies occur, the causes and action results are often not systematically recorded. The experience of skilled workers detecting anomalies through machine sounds, vibrations, smells, and temperature changes is also disappearing as they retire. He explained, "Vast amounts of data are accumulated in equipment, sensors, Manufacturing Execution Systems (MES), and Quality Management Systems (QMS), but contextual information explaining the background of occurrence, process conditions, and equipment status is lacking. Connecting data across processes and creating an environment where AI can continuously learn is the starting point of manufacturing AI."
In the case of China, a competing manufacturing nation, the government is taking the lead in building high-quality industry-specific datasets and common AI platforms, rapidly applying AI and humanoid robots to automobile, electronics, and steel factories. They are also transplanting the AI factory model to overseas production bases, such as in Europe. The US and Germany are also stepping up to overcome the manufacturing crisis by integrating generative AI and autonomous AI agents into decision-making for production, quality, and supply chains.
The solution presented by Professor Youn is the 'AI Native Factory.' It is not just about adding AI solutions to existing factories, but a factory operating system that integrates production, quality, maintenance, safety, energy, logistics, and supply chain data, with AI agents supporting decision-making. It involves accumulating sensor, video, equipment, and operational data in a format that AI can learn from, and then connecting task-specific AI agents to systems like MES. He said, "It's not an unmanned factory, but a structure where AI handles repetitive monitoring and analysis, while humans focus on strategy formulation, creative judgment, and responding to exceptional situations," adding, "Job transition training for on-site workers must be carried out alongside performance sharing based on productivity improvements."
He emphasized that the government's manufacturing AX policy should focus on data capitalization, training experts, joint computing infrastructure, demonstration facilities, and building sustainable operating systems just as much as on distributing GPUs and equipment. While large corporations have their own data infrastructure and specialized personnel, SMEs struggle from the very stage of defining which process problems to solve with AI. Therefore, he requested the government to include the costs of data updates, AI retraining, and maintenance required whenever products and processes change into the support system. He urged that even after AX support for a company ends, the government should evaluate whether the system is actually operating and spreading to other production lines and partner companies. "The manufacturing AX polarization leads to a gap in industrial competitiveness," he stressed. "The government must provide a common data platform, specialized personnel, and demonstration environments as public infrastructure."
Professor Youn also proposed fostering the AI Native Factory as a new export industry for Korean manufacturing. Going beyond exporting finished products like automobiles, batteries, ships, and electronics, this is a model where manufacturers, construction companies, automation firms, and AI companies form a consortium to jointly export everything from factory construction to automation equipment, AI software, and operation services for quality, maintenance, energy, safety, and supply chains. He explained that continuous revenue can be generated by providing AI software and maintenance services even after the factory's completion.
For this, he said it is necessary to establish manufacturing data governance and standard contracts that clarify data rights, security, and the scope of secondary use. "In the field, the focus should not be on the number of robots, but on how quickly data can be connected to decision-making within the factory," he emphasized. "Jim Snabe, Chairman of the Supervisory Board of Siemens Germany, also mentioned that they are 'benchmarking Korea's manufacturing AX.' We must proceed with more confidence and accelerate the transformation to AI Native Factories and the overcoming of AX polarization."
...
...
...
...
While Artificial Intelligence (AI) is changing the competitive landscape of the manufacturing industry, over 99% of the data generated at domestic manufacturing sites is discarded after a single use. As the gap in manufacturing AI capabilities leads to differences in productivity and competitiveness, it is pointed out that on-site data must be transformed into AI assets and production processes redesigned around AI. Ultimately, to maintain competitiveness as a manufacturing powerhouse, building an 'AI Native Factory' and resolving the manufacturing AX (AI Transformation) polarization are urgently needed.
In an interview held on the 26th at onepredict in Banpo-dong, Seocho-gu, Seoul, Byeng Dong - Youn, a professor of Mechanical Engineering at Seoul National University, emphasized, "Although Korea is a global manufacturing powerhouse, there is a severe lack of 'usable manufacturing data' for AI to learn from. If we cannot transform on-site data into usable assets, the future of our manufacturing powerhouse cannot be guaranteed." Professor Youn received his Ph.D. in Mechanical Engineering from the University of Iowa, served as an assistant professor at the University of Maryland, and joined Seoul National University in 2010, where he currently heads the SNU Industrial AX Center. In 2017, he founded the industrial equipment predictive maintenance company onepredict, advancing AI-driven factory solutions and pushing for a KOSDAQ listing next year.
Professor Youn cited the lack of failure and defect data, the disappearance of skilled workers' tacit knowledge, and data disconnection across processes as the biggest challenges for manufacturing AI. Because factories have a high proportion of normal operations, there is a lack of anomaly data for AI to learn from, and even when anomalies occur, the causes and action results are often not systematically recorded. The experience of skilled workers detecting anomalies through machine sounds, vibrations, smells, and temperature changes is also disappearing as they retire. He explained, "Vast amounts of data are accumulated in equipment, sensors, Manufacturing Execution Systems (MES), and Quality Management Systems (QMS), but contextual information explaining the background of occurrence, process conditions, and equipment status is lacking. Connecting data across processes and creating an environment where AI can continuously learn is the starting point of manufacturing AI."
In the case of China, a competing manufacturing nation, the government is taking the lead in building high-quality industry-specific datasets and common AI platforms, rapidly applying AI and humanoid robots to automobile, electronics, and steel factories. They are also transplanting the AI factory model to overseas production bases, such as in Europe. The US and Germany are also stepping up to overcome the manufacturing crisis by integrating generative AI and autonomous AI agents into decision-making for production, quality, and supply chains.
The solution presented by Professor Youn is the 'AI Native Factory.' It is not just about adding AI solutions to existing factories, but a factory operating system that integrates production, quality, maintenance, safety, energy, logistics, and supply chain data, with AI agents supporting decision-making. It involves accumulating sensor, video, equipment, and operational data in a format that AI can learn from, and then connecting task-specific AI agents to systems like MES. He said, "It's not an unmanned factory, but a structure where AI handles repetitive monitoring and analysis, while humans focus on strategy formulation, creative judgment, and responding to exceptional situations," adding, "Job transition training for on-site workers must be carried out alongside performance sharing based on productivity improvements."
He emphasized that the government's manufacturing AX policy should focus on data capitalization, training experts, joint computing infrastructure, demonstration facilities, and building sustainable operating systems just as much as on distributing GPUs and equipment. While large corporations have their own data infrastructure and specialized personnel, SMEs struggle from the very stage of defining which process problems to solve with AI. Therefore, he requested the government to include the costs of data updates, AI retraining, and maintenance required whenever products and processes change into the support system. He urged that even after AX support for a company ends, the government should evaluate whether the system is actually operating and spreading to other production lines and partner companies. "The manufacturing AX polarization leads to a gap in industrial competitiveness," he stressed. "The government must provide a common data platform, specialized personnel, and demonstration environments as public infrastructure."
Professor Youn also proposed fostering the AI Native Factory as a new export industry for Korean manufacturing. Going beyond exporting finished products like automobiles, batteries, ships, and electronics, this is a model where manufacturers, construction companies, automation firms, and AI companies form a consortium to jointly export everything from factory construction to automation equipment, AI software, and operation services for quality, maintenance, energy, safety, and supply chains. He explained that continuous revenue can be generated by providing AI software and maintenance services even after the factory's completion.
For this, he said it is necessary to establish manufacturing data governance and standard contracts that clarify data rights, security, and the scope of secondary use. "In the field, the focus should not be on the number of robots, but on how quickly data can be connected to decision-making within the factory," he emphasized. "Jim Snabe, Chairman of the Supervisory Board of Siemens Germany, also mentioned that they are 'benchmarking Korea's manufacturing AX.' We must proceed with more confidence and accelerate the transformation to AI Native Factories and the overcoming of AX polarization."