Sungwon Ahn, Head of AI Research: "Korea Excels in Manufacturing and Finance... Domain-Specific AI Models Needed"

"For physical AI, represented by robotics and autonomous driving technology, to become established across the domestic industry, there is a need for 'high-performance domain-specific AI models' from a software perspective."

Sungwon An, Director of the Software Policy Research Institute, is giving an interview with The Asia Business Daily. Photo by Dongju Yoon

원본보기 아이콘

Sungwon Ahn, Head of the AI Research Division at the Software Policy Research Institute, stated in a recent interview with The Asia Business Daily, "We need to develop domain-specific models for sectors where we excel and have abundant data, such as manufacturing, finance, and healthcare, and target niche markets like the Middle East."

High-performance domain-specific AI models are those that learn data from particular industrial sites and synchronize the process of 'perception-reasoning-execution' in real time, allowing the model to adjust and refine its responses according to the situation. Since securing high-quality, refined data is the key factor determining the competitiveness of physical AI, it is important for Korea to leverage its strengths rather than competing head-to-head with the U.S. and Chinese big tech companies, which are ahead in technology and capital.

He added, "Since each company collects and holds data differently, achieving data standardization (integration) is realistically not easy," emphasizing, "The government should take an active role in establishing standards."

Sungwon Ahn referenced past attempts at data standardization at Seoul National University Bundang Hospital and Ajou University Hospital, explaining, "Each hospital uses different data formats and has intertwined interests, making data sharing difficult." This is because sharing could expose everything from patient personal information-such as how many surgeries were performed on patients with certain diseases and what the success rates were-to hospital trade secrets.

However, he believes that the positive effects of data standardization are too significant to ignore. He stressed, "Whether data is sold or distributed for free, if it can be shared, it will help reduce the trial and error experienced so far," adding, "This will contribute to the advancement of the entire domestic industry." He particularly noted, "This is helpful for small and medium-sized enterprises and startups, which have difficulty accessing data," and explained, "Discussions on standardization are taking place in the field so that, even if the number of robot joints differs, one program can control all movements."

He also highlighted the importance of network technology for full-stack development. Sungwon Ahn commented, "Nvidia is expanding its direct influence beyond graphics processing unit (GPU) design to building internal data center networks," and explained, "The network acts like a traffic light, reducing bottlenecks in the software (CUDA library) that maximize GPU performance. As AI models become more complex, dependence on the Nvidia ecosystem is likely to increase, so a strategic approach to this is needed."

Regarding engineering capabilities, he stated that engineers should be architects with creativity and the ability to design and oversee entire systems. As demand for entry- and mid-level technicians declines, it is necessary to possess the ability to judge the quality of AI output, understand data and processes, and have interdisciplinary, convergent knowledge.

Sungwon Ahn said, "There is increasing demand for 'domain convergence talent' who can reflect the voices from the field," and elaborated, "For example, in a chemical process company, it is not enough to know only AI technology; one must also know the hazardous components of chemicals, optimal temperatures, and other basic knowledge to determine what data needs to be collected." He added, "At present, robots cannot yet match the intuition of on-site workers with 20 to 30 years of experience, and people still prefer face-to-face services in areas like healthcare. To pass down know-how and expertise, it is important to ensure a continuous inflow of skilled personnel."

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.