Discussions to expand the utilization of artificial intelligence (AI) in the overall field of research and development (R&D) have gained momentum. Researchers have proposed urgent solutions to the lack of AI research infrastructure, such as graphics processing units (GPUs), and institutional shortcomings, while the government expressed its commitment to large-scale support.
On 4th, the Ministry of Science and ICT held an 'AI for S&T experts meeting' at the Korea Institute of Energy Research (KIER) in Daejeon, discussing AI utilization methods with experts from various fields including bio, semiconductors, and nuclear power.
Minister Bae Kyung-hun said, "As AI can change many scientific fields, applying AI well can create an environment capable of producing Nobel Prize winners," and mentioned, "We will open 20-30% of the 50,000 GPUs we aim to secure by 2028 to academia and research institutions."
Experts attending the meeting noted that AI can accelerate the development of new bio drugs, lead R&D innovations in semiconductor design optimization, and enhance the design and safety of reactors. They suggested that active government investment is needed for this.
Professor Kim Jong-ho from the Korea Advanced Institute of Science and Technology (KAIST) stated, "Depending on how AI is applied in the semiconductor field over the next 10 years, the fates of Nvidia and Samsung Electronics will diverge," and added, "In a situation where engineers are gradually decreasing, I believe that the country that effectively utilizes AI will survive."
There was also a discussion about the practical difficulties felt in the research field. Professor Kim Woo-yeon from KAIST's Department of Chemistry said, "The process of purchasing GPUs is too complicated." For example, if one already owns a GPU, one must explain why additional purchases are necessary. He pointed out that when expenses exceed 500 million won, a specialized manager is required, but individual research labs cannot hire such managers.
Yoo Yong-kyun, head of the AI Application Research Lab at the Korea Atomic Energy Research Institute (KAERI), mentioned, "To remain competitive, AI technology must be integrated throughout the entire cycle of next-generation reactor design and licensing. However, there is a lack of evaluation systems and policies to encourage cooperation between related departments," and added, "Even if selected for a project utilizing AI, a separate application is needed to use GPU resources; it would be nice if it could be resolved all at once."
Lee Jae-hyeon, head of the Energy AI and Computational Science Lab at the Korea Institute of Energy Research (KIER), stated, "Even if cloud (virtual server) resources are utilized, the aging of research lab facilities limits the speed of data transfer," and mentioned, "To have suitable infrastructure, it costs 100 million won per building, and at least 500 million to 1 billion won at the institute level."
The Ministry of Science and ICT plans to establish and operate the 'AI for S&T industry-academic-research expert task force (TF)' chaired by First Vice Minister Gu Hyuk-chae, in connection with this discussion. Based on the results of the TF's operation, the ministry plans to establish a national strategy for AI for S&T and systematically promote policies.