The US media often gets caught up in self-centered narratives, but we firmly believe that learning about global AI affairs, cultural nuances, and political variations can benefit us and our readers a lot.
In today's edition, Ben Eum presents an in-depth historical analysis of AI in South Korea, coupled with key highlights from Generative AI events in 2023. Pay special attention to our takeaways and conclusions to stay well-informed. South Korea has demonstrated remarkable tech-savviness; observing trends in this country could provide crucial insights.
Introduction
The match between AlphaGo and Lee Sedol 9th Dan in March 2016 (officially named the ‘Google DeepMind Challenge Match’) could perhaps be counted as one of the first notable events in the history of artificial intelligence in South Korea. To the citizens of a country intimately familiar with the game of Go, this event was an emotional roller coaster that started with 'confidence' and 'curiosity' and ended in 'shock' and 'confusion'. Following the dramatic 4:1 conclusion of this historic AI-versus-human Go match, media, experts, and the press competed to share their opinions on the meaning of the match. Thereafter, 'artificial intelligence' became a national agenda, deeply ingrained in the minds of Koreans as a 'new technology' with the potential to change the world.
If that match could be compared to the 'First Impact' from the famous anime 'Evangelion,' then the shock to the Korean market and industry brought about by the generative AI services led by ChatGPT, released by OpenAI on November 30, 2022, could be described as a 'Second Impact’ which surpassed the former in both quantitative and qualitative aspects. ChatGPT, which attracted a million users in less than a week of its launch, quickly spread its influence in Korea. A survey by the Korea Chamber of Commerce and Industry in March 2023 showed that one in three Koreans had used ChatGPT at least once. As of April 2023, there were already 2.2 million users in Korea according to OpenAI's data. Not only did the stock prices of companies related to generative AI services soar, but major corporations and startups alike have developed and launched various generative AI services within the year. The government has also recognized the significance and importance of generative AI technology, declaring a project for the 'AI routinization for the entire population' and planning and investing in a budget to establish a foundation for nurturing hyper-scale AI.
In 2024, the investment and interest of stakeholders in the Korean market for generative AI are expected to deepen and expand further. Most importantly, AI, which was a national agenda but could only be a proprietary technology of some major corporations, giant tech firms, and a few startups, has come closer to everyone in the form of ‘services not as 'technologies'.
Dream of ‘K-AI’: Driving country transformation and AI leadership with sovereignty
Anyone interested in culture and entertainment might have heard of terms like K-Pop and K-Drama. Korea has added a new interpretation based on its history and culture to various genres that began overseas, systematizing them to create content with Korea's unique color and distinction, which is loved worldwide.
In the field of AI as well, two unique ‘K-AI’ approaches reflecting the current situation and context of South Korea can be identified. One is the 'national-level full investment and dissemination of generative AI as a tool for survival and transformation', and the other is the 'pursuit of vertical integration of the generative AI industry centered on telecom companies, conglomerate groups, and platform companies and securing AI sovereignty'.
National-level full investment and dissemination of generative AI as a tool for survival and transformation
Since the Korean War, South Korea's rapid economic growth has been fueled by policies like fostering talent, industry support, low taxes, and minimal welfare. Yet, post-COVID-19, global uncertainties from US-China trade conflicts and Brexit have raised recession fears. Recognizing this, there's a push for innovation in key sectors like shipbuilding, chemicals, and automobiles through digital and AI technologies, aiming for a fundamental industrial and business model overhaul to counter prolonged low growth.
Moreover, South Korea faces challenges due to drastic demographic shifts, particularly as its low birth rate has led to a total fertility rate of 0.78 in 2022, marking it as the world's fastest-aging country. This situation necessitates urgent responses to the dual issue of a shrinking workforce and rising welfare demands, critical for the nation's sustainability.
The South Korean government recognizes that adopting and spreading AI technology in all industries is a key element in restoring the vitality of the South Korean economy and solving social problems. In December 2019, it announced the 'National AI Strategy' to build a world-leading AI ecosystem and continue investing to transform the country into one that utilizes AI well. Major tasks include establishing AI innovation clusters at key locations nationwide to expand large-scale high-performance computing resources, promoting investment in fundamental AI research and preemptive development of next-generation intelligent semiconductors, and implementing AI literacy and interdisciplinary education throughout the educational curriculum, as well as establishing AI graduate schools to nurture AI specialists.
Furthermore, in April 2023, the Ministry of Science and ICT announced the 'Plan to Strengthen the Competitiveness of Hyper-scale AI' which, based on the achievements in the AI field thus far, plans to strengthen the competitiveness of domestic large language models (LLMs) and nurture generative AI based on LLMs as a future strategic industry. To accelerate the AI routinization for the entire nation, it plans to start with an investment of approximately 400 billion KRW (~$300 million) in 2023 and continue with sustained investment over the next 10 years.
Securing AI Sovereignty: Telecom companies, conglomerate groups, and platform companies pursuing vertical integration in the GenAI industry
With active government support, the Korean industrial sector also conducts aggressive experiments and investments to secure global competitiveness in generative AI, develop new businesses, and expand overseas.
Korea, alongside the United States, China, the United Kingdom, and Israel, is a key player in hyper-scale LLMs based on its national language. Notably, only the U.S., China, and Korea offer comprehensive commercial services. These efforts are driven by telecom and major platform companies, traditionally instrumental in Korea for pioneering innovative digital, mobile, and AI technologies. Leading firms like SK Telecom, LG, and Naver have been advancing hyper-scale AI language models, actively shaping their AI ecosystems in 2023.
Their ecosystem construction invariably exhibits the aspects of 'multi-LLM' and 'vertical integration' and the status of key operators' investments and partnerships in the generative AI field is as illustrated below:
As seen in the illustration above, except for Naver – which can be considered the leader of South Korea's hyper-scale AI language models – multi-LLM strategy aims to enhance the scalability and competitiveness of their platforms through partnerships with various open-source/closed-source LLM developers, aside from their own hyper-scale AI language models.
There is also vertical integration that includes inference-focused AI hardware companies such as Sapeon, Rebellions, and Furiosa AI, under the cooperation of the government and cloud platform operators.
In 2023, these players conducted quite a few PoCs (Proof-of-Concept) for various generative AI services targeting numerous clients in the finance and manufacturing industries. Most experiments are being conducted in the context of B2B in areas such as intelligent search, knowledge discovery, recommendation engine, AI call center and chatbot, and report generation.
It seems that 2024 will be the year when B2B-oriented generative AI solutions and services are expected to spread in both domestic and overseas markets.
These players are prioritizing the B2B market over B2C due to the uncertain success in generative AI B2C services globally and the challenge of competing with giants like OpenAI in training costs. The trend is to first build experience in generative AI business and services through B2B, focusing on monetization in the short to mid-term. Given Korea's role as a key player in the global manufacturing industry, targeting specialized generative AI for this sector, and focusing on developing cost-effective, high-performance technologies and innovative services is seen as an effective short-term strategy. Companies like Naver in Saudi Arabia and KT with a Thailand partner are already establishing their overseas B2B presence with hyper-scale AI language models.
Challenges Ahead
To maintain the competitive edge of the generative AI industry and lay the groundwork for future growth, several challenges need to be handled:
First-generation vs second-generation services.
Generative AI services currently developed by key players and startups are often 'first-generation,' focused on partially enhancing existing value chains and work processes. Many startups in South Korea are in reality just GPT-wrappers, lacking unique features.
The real challenge lies in innovating and adapting services that fully exploit generative AI's unique capabilities and potential economic impact. Reflecting on the early smartphone era, where many initial services merely replicated internet browser workflows on smaller screens without optimizing for new user experiences and smartphone form factors, we recognize a similar pattern emerging in AI.
The future likely belongs to 'second-generation services,' which could encompass agent technologies functioning as assistants or co-pilots, attuned to users' comprehensive needs.
South Korea's AI talent development.
Another key challenge is cultivating and securing talent and resources capable of envisioning and developing these new, disruptive service models, rather than building them. Consequently, the roles of business developers and UX designers, adept at harnessing generative AI's ability to provide probabilistic, interactive responses, unlike technologies that yield predictable outcomes from one-way user commands, will become increasingly vital.
Another issue that needs to be looked at from the talent perspective is that South Korea is a 'net outflow country for AI talent'. According to the 2020 analysis of AI talent movement between major countries by Element AI, South Korea's AI talent inflow index is -0.3, indicating that AI talent is flowing out of the country. Leading AI countries such as the United States (6.2), the United Kingdom (1.6), Germany (0.8), and Canada (0.6) are all inflow countries without exception.
Activating collaborative research ecosystem.
In comparison with leading countries that aggressively pursue research collaboration with other countries, South Korea's global collaboration in AI research is quite inadequate. Countries like Ireland, Singapore, and France, which have a comparable scale of researchers to South Korea, have a global collaboration count of 7~15 in terms of paper authors, while South Korea is very low at about 3. South Korea should not remain in a 'Galapagos' but instead create an ecosystem that actively participates in the global research community.
Creating a successful case in the global market for AI semiconductors.
Not only corporate subsidiaries and startups such as Sapeon, Furiosa AI, and Rebellions, but also Samsung Electronics-Naver through partnerships, are entering the AI semiconductor business. However, there are difficulties in securing distinctive competitiveness against global companies like NVIDIA, and the success of continued funding for some AI semiconductor startups in 2024 seems to be uncertain. Securing the leadership in the low-cost, high-performance AI semiconductor industry is very important as it’s a key element of vertical integration for the growth of South Korea's hyper-scale AI language ecosystem.
Appropriate social consensus for the risks of generative AI.
Companies need to prepare proactive risk management and response measures. At the end of 2023, after meeting with numerous major corporations, I found that many companies are preparing budgets related to generative AI for the official development and release of services in 2024, following numerous PoCs conducted in 2023. However, a considerable number of companies have hit a wall, such as delaying service launches or undergoing full reevaluation due to risks such as bias, privacy, security, and hallucination, which they only began to feel during the PoC stage.
Discussions on AI ethics checklists and certification/verification systems are indeed underway in South Korea. However, considering the general election in 2024, it’s hard to expect that harmonious national standards or regulations to guide corporate risk management will be established quickly. Companies that pursue business on a global scale, in particular, need to invest in proactive AI risk management systems and tools at the organizational level, business process level, and technology platform level to secure global competitiveness for their generative AI services.
Generative AI Key Events in 2023
Conclusion & Key Takeaways
Considering various factors such as market size, digital infrastructure, AI research and development capabilities, AI literacy, and market readiness, South Korea is well-positioned to be a reference country that could lead the growth of the generative AI industry in the Asian region in the future.
Moving past the year 2023, which could be termed the hype phase of generative AI, 2024 looks to be the year where generative AI is truly considered a viable business. Whether the vertically integrated ecosystems will be successfully in operation with generative AI at their core, and will indeed be able to develop and roll out profitable killer services, remains to be seen. Additionally, amid the political calendar centered around the general elections in April, it will be a key point of observation whether the discussions on AI risks, currently academic, will materialize into concrete standards and risk prevention systems, and whether this process will see the emergence and continued investment in competitive LLM development and generative AI service startups, enabling their growth.
Previously in the Global AI Affair series: 6-month report (from January 2023 to early July 2023):
