[Public Voices] The Real Opportunities and Challenges Presented by ChatGPT

[Asia Economy] The emergence of ChatGPT has stirred up the internet and news media. Stories abound about ChatGPT producing writing at a college student level, or providing optimal prompts tailored for another AI service specialized in image generation (Midjourney) to create stunning pictures. With Microsoft announcing a new search service based on this technology, public interest and expectations have reached a peak.

Since AlphaGo, which defeated 9-dan Lee Sedol, various AI technologies have been introduced, but what makes ChatGPT different? First, it seems to truly understand human language. Large Language Models (LLM) like ChatGPT are trained on text-based data. This includes most of humanity’s knowledge assets such as novels, specialized books, scientific papers, newspapers, historical records, and internet documents. It is thus closer to ‘Artificial General Intelligence (AGI)’?the ability to understand and respond to a wide range of issues like a human.

The most important impact of ChatGPT is that it lowers the threshold for practical AI use. Since ChatGPT is already a trained model, no new training is required. It only needs to be properly equipped with suitable pre- and post-processing software to be effectively utilized. In other words, it is a single AI model that can be used for (almost) any problem. With just an appropriate user interface, an English education chatbot can be launched immediately, and by adding a module linked to key economic data, it can automatically generate market status reports. The effort, investment, and development time required to create AI models tailored to each function are drastically reduced. For AI startups, this is like a gold mine of opportunity. This is why we expect companies equipped with AI technology to creatively solve business needs to thrive in the coming years.

Naver, Kakao, and LG are also developing large-scale AI models. Competing with the giant US IT companies on model size or relying solely on Korean language specialization is unlikely to succeed. In 2 to 3 years, ChatGPT itself will become outdated technology. Large corporations must focus on creating new AI that overcomes limitations no one has properly solved yet. Three major challenges can be identified: First, the issue of reliability?distinguishing whether ChatGPT’s answers are factual, imaginative creations, or incorrect responses based on faulty data. Explainability, which transparently reveals the reasoning process and source data behind answers, is a crucial task. Second, there must be continuous and effective methods to update already trained large AI models with newly generated, up-to-date information. Third, the enormous computational resources required to train and operate large AI models must be optimized immediately.

All competitiveness comes from talented personnel, so securing and nurturing talent in this field is urgent. It is encouraging that the government is expanding coding education and setting a good direction for mid- to long-term workforce development. However, recently emphasizing semiconductor talent development raises concerns that other fields might be perceived as less urgent. The core of future industrial competitiveness lies in artificial intelligence and software (SW). I want to emphasize once again that the greatest mismatch between future workforce demand and supply is in this field.

Lee Sang-gu, Professor, Department of Computer Science, Seoul National University

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