"National AI Team to Develop Korean-Style ChatGPT to Open Applications Within Three Months... Large Corporations Also Eligible"

"An Elite 'World Best LLM' Team Must Emerge by Year-End at the Latest"
Project to Support Research Funding for GPUs, Data, and Talent
Plans to Expand Support for CCTV and Foreign Language Data

In this case, Lee Kyung-woo, head of the National Artificial Intelligence Committee Support Group, is giving a post-briefing of the 3rd National Artificial Intelligence Committee held at the Government Seoul Office Building on the 20th. Photo by Yonhap News.

The government plans to select an 'AI National Representative Elite Team' from companies within as soon as three months to develop a world-class generative artificial intelligence (AI) model. It is expected that not only early startups and young companies but also large corporations will participate in developing foundation models of the highest global standards.

On the afternoon of the 20th at the Government Seoul Office, Lee Kyung-woo, head of the National AI Committee Support Group, stated, "The timeline for the 'World Best Large Language Model (LLM)' (hereafter WBL) contest aims to be within three months," adding, "By 2025 at the latest, an elite WBL team meeting public expectations should emerge so that we can promptly invest and produce results." WBL refers to an AI national representative project that forms a corporate consortium to pool resources and capital from individual companies, adding research funding for graphics processing units (GPUs) and talent to enhance development efficiency.

Public and private data for AI training will also be expanded. He said, "About 650,000 CCTVs operated collectively by 216 local governments capture videos of accidents, disasters, and calamities, which are necessary for AI development," and added, "Relevant ministries plan to jointly establish laws and systems to extract such event data and build training databases." He further noted, "If English-language data is needed to develop a global-level LLM, there may be provisions to allocate research funds accordingly."

Additionally, the government decided to extend the use of unstructured original data such as videos, which had so far been allowed only in autonomous driving, to other fields. To enable stable data use in AI research, special provisions for personal information use will be prepared, and the system will be improved to allow personal data use for public interest AI development, such as crime prevention. Plans were also announced to build and open synthetic data for sensitive fields like healthcare and defense, as well as specialized data for media, industry, manufacturing, and finance. Incentives will be established to encourage public institutions to actively provide pseudonymized information.

The government plans to select elite teams with AI research and development (R&D) capabilities regardless of company size. Lee said, "It will serve as a playground concept for capable early startups and young companies," but added, "There will be no restrictions on selection for large corporations, small and medium enterprises, research institutes, or academia." He explained, "Even researchers from large corporations who have LLM development experience and sufficient capabilities are naturally expected to participate in the WBL project."

The WBL project is expected to produce foundation models of the highest global standards. Lee stated, "Model evaluation will be conducted through global platforms like Hugging Face to objectively assess the level," emphasizing, "It is not about achieving a specific score but about recognition by developers worldwide." Hugging Face operates the 'Open LLM Leaderboard,' a representative LLM benchmark.

Regarding opinions that focusing on small language models (SLM) might be more efficient, he drew a line by saying, "If manpower and technology have been accumulated while developing foundation models, we envision an AI ecosystem supplied by many developers through SLM as well." Lee added, "This does not mean we will not pursue SLM. Since technologies required differ across industries such as manufacturing, distribution, and search, we will continue to invest and support these areas."

At the briefing, Kang Ki-ryong, Director of Policy Coordination at the Ministry of Economy and Finance, said, "Deputy Prime Minister and Minister of Economy and Finance, Acting President Choi Sang-mok, requested that the timeline be accelerated much faster than initially planned, and that prize money and funding be provided at a world-class, very generous level."

This content was produced with the assistance of AI translation services.