The future AI strategy for Korea can be summed up in three tracks: responding to artificial general intelligence (AGI) and frontier-level models, artificial superintelligence (ASI) for science and technology, and AI agents for the public.
Vice Prime Minister and Minister Bae Kyung-hoon of the Ministry of Science and ICT said at a press briefing held on the 29th at Seoul Central Post Office in Jung-gu, Seoul, that Korea will go beyond developing AI tailored to industries where it has strengths, such as manufacturing and semiconductors, respond to competition for the world's top-tier AI models, deploy AI to solve scientific and technological grand challenges, and roll out AI agents into people's daily lives.
Frontier models refer to AI models at the forefront of today's global AI race, such as ChatGPT, Claude and Gemini. AGI is a more future-oriented concept: a general-purpose AI that understands various situations like a human and makes its own judgments to solve a wide range of problems.
Bae said, "It is time for Korea to take on the challenge of building frontier models on par with the United States and China," adding, "When the time comes within a few years for AI to create better AI on its own, the pace of technological change could accelerate sharply, so we must prepare for AGI."
An even more distant axis of competition is ASI. ASI is a higher level of AI than AGI, which broadly judges across multiple fields like a human, and refers to AI that goes beyond human experts' knowledge and reasoning to form hypotheses and find solutions on its own. In areas such as new drug development, materials discovery, semiconductor processes and basic science research, where vast data and high expertise are needed, ASI can move beyond a scientist's auxiliary tool to become a research partner.
The final pillar of the AI strategy is AI agents that the public can feel. While frontier models, AGI and ASI are about securing world-class AI technology, AI agents are the step of bringing that technology into daily life. AI agents go beyond chatbots that answer questions to AI that, under a user's instructions, handles tasks on their behalf such as issuing documents, searching for information, coordinating schedules and using administrative services.
Bae cited the government's planned public-facing AI service, "AI for all," as a representative example. He said, "AI for all will include a chatbot by default, and the differentiator is the concept that people each own an AI agent," adding, "We will also offer specialized model services for older adults or marginalized groups who may find it difficult to use AI effectively."
The government plans to unveil AI for all around Nov. and push to provide it free of charge based on government finances through 2028. Bae said, "Even after 2028, we will discuss joint investment structures with corporations so the free service can continue," adding, "We will make it easy for everyone to use AI, like Hangul and arithmetic."
◇ "Aggressive investment in AI infrastructure is needed… strategies must be flexible"
The government will also pursue AI transformation (AX) on factory floors where Korea has strengths. Bae cited manufacturing, semiconductors and industrial data as the core asset of Korea's AI strategy and emphasized semiconductor manufacturing–specialized AX and physical AI. Physical AI is AI that operates in the real physical world, such as robots, Autonomous Driving, manufacturing equipment and logistics.
Bae said, "Many experts see the AX application failure rate as over 80%. Simply introducing AI does not automatically produce results," adding, "If we can raise the AX success rate to 70%–80% with pan-government support, we can deliver many outcomes."
Investment must follow to back these AI strategies. Bae cited as the biggest change in AI over the past year the ongoing build-out of high-speed AI infrastructure centered on graphics processing units (GPUs).
Bae said, "If you look at how many GPUs are in Korea, it is still far from enough, but I don't hear many saying research results are hard to achieve because GPUs are lacking," adding, "Above all, it is meaningful that the government and the private sector have formed a consensus and that the stage for investment has been set." On the private corporations' discussions to invest in 260,000 GPUs, he said, "That figure is increasing now."
However, he also pointed to the reality that the scale of investment is still insufficient to compete with global big tech. Bae said, "Korea's total AI budget is about the level of a single U.S. big tech corporation's investment," adding, "To build a frontier-level model, we need technology, talent, data and infrastructure; while our technical capabilities have risen a lot, infrastructure investment is still lacking. Investment in AI infrastructure such as GPUs needs to be made aggressively."
He added, "In the AI era, the government's strategies cannot remain fixed plans. In five or 10 years, the benchmarks may change," noting, "Because forecasting is difficult, it is necessary to revise strategies actively and flexibly."