“The achievements of DeepSeek in China demonstrate how crucial talent is in securing competitiveness in artificial intelligence (AI) technology. Currently, the number of AI researchers in Korea is only 21,000, which is merely one-twentieth of China's.”
Choi Yang-hee (70), president of Hallym University, met with ChosunBiz at the office of the National Science and Technology Advisory Council in Jongno, Seoul, on the 7th. He stated that securing domestic AI talent is urgently needed. According to the Paulson Institute at the University of Chicago, China accounts for 47% of the AI researchers in the top 20% worldwide, followed by the United States with 18% and Europe with 17%. Only 2% are from Korea.
Choi served as a professor in the Department of Computer Engineering at Seoul National University for 29 years and dedicated efforts to train software (SW) talent. He has held positions as the first chairperson of the Seoul National University AI Committee and as minister of the Ministry of Science and ICT. Currently, he is vice-chairperson of the National Science and Technology Advisory Council and an outside director at KT.
Choi noted, “Korea is currently facing a serious crisis in terms of quantitatively and qualitatively securing AI talent,” adding that “the number of AI researchers is significantly insufficient compared to global competitors.” The following is a Q&A.
—The DeepSeek shock has underscored the importance of securing AI talent worldwide.
“Despite strong checks from the United States, China has achieved results comparable to OpenAI's latest AI models through the power of its homegrown talent. This serves as a proof that AI competitiveness ultimately relies on the capabilities of outstanding talents. In Korea, the trend of top students gravitating toward medical schools is worsening each year, making the future outlook for securing AI talent grim. About 40% of domestic AI-related graduate students are migrating overseas. Consequently, the domestic AI workforce is continuing to decline.”
—What about the startup environments in the United States, China, and Japan?
“The United States is not too worried about securing AI talent as it absorbs AI talent from all over the world. It has the world's top-level AI talent. China is implementing a strategy to expand its domestic talent pool to enable homegrown talents to thrive in AI development. It produces 1.5 million engineering graduates each year. Universities like Tsinghua University, Zhejiang University, and the University of Science and Technology of China are nurturing indigenous AI talent, reaching a level of 'self-sufficiency.' Starting in 2020, Japan has transformed into a net importer of AI talent due to its innovative visa systems and systematic support policies. 'Sakana AI,' founded by foreign entrepreneurs from Google, has rapidly grown into an AI unicorn in Tokyo.”
—What are the reasons Korea struggles with securing AI talent?
“There is a significant disparity in the compensation systems for talent. Major U.S. tech companies such as Google offer starting salaries of up to $1 million (about 1.45 billion won) for doctoral-level AI researchers. This is vastly different from Korea's large corporations. Even OpenAI reportedly offers starting salaries of $865,000 (about 1.25 billion won) for doctoral-level AI researchers. Additionally, a corporate culture that emphasizes tenure over performance and a lack of long-term career growth are also problematic. Many private companies in Korea prohibit concurrent employment, but if 2-3 concurrent jobs were allowed, I believe that overseas AI researchers might return to Korea.
The second reason is the weakening of the educational (university) ecosystem for nurturing and retaining talent. The poor research environment and low compensation at universities are leading talented individuals to migrate overseas.
The third reason is Korea's inadequate AI startup ecosystem. The small scale of the investment ecosystem, disconnection from the global investment market, and strict AI data regulations have rendered the AI startup ecosystem fragile. It is difficult for domestic startups to secure competitiveness, causing AI talent to leave Korea.
—How are the startup environments in the U.S. and China?
“The U.S. and China actively support AI startups and are expanding their AI ecosystems. Companies like OpenAI, Anthropic, and Cohere in the U.S. are receiving large-scale investments from venture capitalists (VC) and major corporations to ensure sustainable growth. China also provides vigorous government support for AI startups, and large corporations such as Baidu and Alibaba are increasing their collaboration with emerging AI companies.”
—What is the current level of Korea's AI technological competitiveness?
“According to the recent Global AI Index 2024 ranking released by Tortoise Media, the United States and China are solidifying their global leadership based on overwhelming investment, a large number of AI talents, and government support. Using the United States (100) as a benchmark, China recorded approximately 53.88, making it the second-highest country for AI technological competitiveness. South Korea (27.26), Singapore (32.33), the United Kingdom (29.85), France (28.09), and Japan (20.31) are contending with scores in the 20-30 range.”
—What are your views on whether domestic corporations should secure competitiveness through AI technology collaboration with advanced overseas firms like Kakao and KT, or independently develop their technologies?
“Both strategies have their advantages and limitations, and considering Korea's realities, a dual strategy is necessary. The term 'sovereign AI' refers to acquiring data sovereignty and independent technological capabilities, which is crucial. Even if domestic corporations collaborate with advanced foreign technology firms, they need to possess their own technological capabilities to have bargaining power and respond to sudden demands for fee increases from overseas firms.
Since representative models like ChatGPT primarily utilize U.S. data, the low level of proficiency regarding Korea's unique language and culture is another reason why securing independent AI technological capabilities is necessary. There is a risk of leaking key information in advanced and defense industries; therefore, major countries are also actively working on sovereign AI development. However, it is also necessary to adopt a strategy that rapidly incorporates the latest technologies through collaboration with overseas firms to secure global competitiveness.
While developing measures to enhance digital security and sovereignty, it would be beneficial for Korea's AI competitiveness to actively cooperate with overseas big tech firms to increase productivity and to promote the development of competitive global services in a balanced manner.
—Domestic corporations are rapidly advancing the adoption of generative AI.
“AI has already established itself as a key driver across various industries, becoming an essential element that influences corporate survival and competitiveness. This year is expected to mark a turning point in the practical application of AI. According to Forrester Research, over 70% of companies worldwide are expected to adopt AI this year. The recently announced collaboration between Kakao and OpenAI is expected to accelerate the adoption of generative AI among domestic corporations.
Noteworthy is the emergence of 'AI agents.' Through AI agents, companies will acquire the ability to autonomously detect and solve problems, going beyond simple task automation. The productivity gap between companies that adopt AI and those that do not will increase.
—Are regulations an obstacle to the advancement of domestic AI technology?
“When regulating AI, a balanced approach is necessary that secures accountability and reliability without hindering technological innovation. Currently, Korea is reported to have higher regulations compared to the European Union (EU) and Japan. It is known that Korea has 2.6 times more AI regulations than the EU and approximately 1.8 times more than Japan.
Last December, the Basic AI Act passed the National Assembly's plenary session, and it is deemed important to specify the regulatory scope and requirements in future enforcement decrees. It is necessary to resolve business uncertainties through effective regulatory provisions. Differentiated regulations considering the characteristics of each industry should be introduced, and improvements such as extending the period for faculty startup leave and expanding regulatory sandboxes should also be addressed.
—What advice would you give regarding the educational direction for nurturing AI talent?
“From elementary school through university and graduate school, educational programs should be designed from a long-term perspective. Instead of limiting education for AI talent to a narrow technical field, foundational literacy education should precede. Since AI is applicable in a wide range of fields, the ability to utilize it is also crucial. The term AI talent feels overly restricted to just AI. In some ways, software talent seems more appropriate. Thus, I oppose the term AI graduate school, as software capabilities form the foundation of AI. It is necessary to create long-term programs for nurturing software talent starting from elementary school.