"I estimate that it must have cost at least 200 million won."

This was the assessment of Jung Kyuyeong, CEO of C7 Plannings, a video production company, after watching a video about Jeju Island. But how much did it actually cost to produce this video? The answer is "virtually zero." The video was created as a class project by students enrolled in the "Data-Driven Next-Generation Digital Healthcare Artificial Intelligence (AI) Solution (AI Healthcare)" course at the Asia Economy Education Center. The entire video was made using AI. Naturally, there were no production costs. AI Healthcare is a government-funded, tuition-free vocational training program (K-Digital Training, KDT) supported by the Ministry of Employment and Labor.

원본보기 아이콘

원본보기 아이콘

원본보기 아이콘

Jung's estimated cost reflects what it would take to shoot, edit, and add computer graphics and audio to the video without using AI. But how much do experts think it would cost if AI is used for the work? Another video production company CEO, Jung Juhun of RSMID, estimated, "It looks like they used about four AI programs," and set the price at "between 3 million and 10 million won." The actual team consisted of four students majoring in life sciences, sports medicine, microbiology, and resources and environmental engineering. None of them had studied computer programming or video production.

Um Soonhyun, the lead instructor for the AI Healthcare course, said, "When using AI, there is little difference between someone who majored in Korean literature and someone who majored in computer engineering." He added, "There is also no difference between a developer with ten years of experience and one with just one year. In other words, it's like developing with 1,000 assistants. In development work, programming skills are now less important than logical and integrative thinking abilities."

He continued, "Now, the person who can program best is the company CEO." The reason is that only the CEO can give precise instructions based on a clear understanding of what is needed. "For developers, logical and integrative thinking skills are now more important than math or programming abilities." The Asia Economy Education Center, which used to teach how to follow instructions well, has now started teaching how to give good instructions. The educational policy was changed to foster the kind of talent needed in today's world. People have been saying for a year or two that AI will change the world. But AI has already changed the world?we just haven't fully realized it yet.

Last February, Andrei Karpathy, co-founder of OpenAI and former head of AI at Tesla, posted on X (formerly Twitter) saying, "There is a new software coding (software development) method that I call Vibe Coding."

원본보기 아이콘

There is a word that symbolizes this changed world: "vibe coding." This neologism, which appeared early this year, is shaking up the global software, content, and startup ecosystems. Last February, Andrei Karpathy, co-founder of OpenAI and former head of AI at Tesla, posted on X (formerly Twitter) saying, "There is a new software coding (software development) method that I call Vibe Coding." The word "vibe" comes from "vibration" and is usually used to mean atmosphere, feeling, or aura. For example, "That cafe has a vibe you can't find anywhere else."

Until now, coding meant that developers had to learn computer languages and input them into computers via keyboard. To program, you first had to learn a computer language based on mathematics. Of course, it took a lot of experience to use it properly. Many nights spent hugging the keyboard were necessary before you could finally produce something that could be called a program.

But vibe coding is about simply telling the AI what you want to make, as if you were texting a friend, and having it generate the program. You program using natural language?the way you speak every day. Karpathy said, "I completely surrender to the AI's vibe and even forget that code exists." He added, "If there's a problem, I just have it remake it. These days, I hardly ever touch the keyboard." If programming in the past was about mathematics, vibe coding is a combination of rhetoric and logic. Put simply, you develop programs using everyday language?natural language.

In the Book of Genesis, God says, "Let there be light," and there was light. Now, when a developer says, "Let there be a program," a program appears. The most important things for a developer now are creativity, imagination, intuition, and sensibility. The grueling process of writing error-free code is gone. Instead, you have to imagine a concept that never existed before?like light in a world of darkness?and explain it logically to the AI. This is why creativity has become even more important.

Has the programming environment really changed this much? When Karpathy first posted about it, many thought he was exaggerating and mixing in some hope. But about a month after he posted on X, the term "vibe coding" was all over the news again.

In March, Garry Tan, CEO of Y Combinator, said that thanks to vibe coding, "Ten people can now do what used to take 50 to 100 developers." Y Combinator is the world's most famous startup accelerator. Sam Altman, CEO of OpenAI, was once also the head of Y Combinator. Accelerators, also called startup planners, find promising early-stage startups, invest in them, provide workspace, and teach business strategies. In other words, they are like a mother to startups and know them better than anyone else.

CEO Tan attended a demo event for startups in their portfolio and said that 25% of participating companies do 95% of their coding with AI. In other words, a significant number of top startups now have developers who simply converse with computers instead of writing code themselves. The human's job is to check whether the result matches the instructions, and if not, to keep explaining until the outcome is satisfactory. Naturally, fewer people are needed. The days when people couldn't start a business because they had ideas but no programmers are quickly disappearing.

원본보기 아이콘

원본보기 아이콘

원본보기 아이콘

The Korean software industry has also ridden the huge wave created by AI. Oh Chiyoung, founder of Jiran Jigyogroup, one of Korea's leading software companies, solved a problem that had plagued his company since its founding in 1994 thanks to AI. Jiran Jigyogroup, which has more than 20 domestic and overseas subsidiaries and affiliates, has struggled with a shortage of developers for the past 30 years. The world is vast and there is much to develop, but Korea has always lacked software engineers. However, Oh said, "After actively utilizing AI, productivity increased by 30%."

With more development manpower available, the first thing they did was handle development requests from their Japanese subsidiary in-house. Previously, they often outsourced development to external companies because they lacked people. But now, things have changed. The Japanese subsidiary, which knows the local situation well, acts as project manager, while Korean group engineers and AI handle the development.

The future looks even brighter. Oh explained, "If you use AI for existing tasks, productivity increases by 1.3 times, but if you design and work with AI from the beginning, productivity increases by two to three times." This means that new projects that used to require two or three people can now be handled by one person and AI. Jiran Jigyogroup's motto is "Dream, Challenge, and Relentless Progress." From the company's perspective, they can now pursue projects that were previously just dreams because there was never enough manpower.

원본보기 아이콘

Of course, AI is not positive in every respect. There is a catch to the idea that you can develop software just by speaking. In 2023, Karpathy said, "The most popular programming language is English." The natural language he uses for vibe coding is English. Lead instructor Um has a similar view. "When it comes to vibe coding, your major doesn't matter much. Sometimes, people who majored in Korean literature use AI better than computer science majors. But there is a difference in results between those who are good at English and those who are not."

The natural language that AI understands best is, strictly speaking, English. Most of the people who developed major AI systems used English. The programming languages used to develop AI, such as Python, C++, and Java, are also in English. You have to use English to communicate smoothly with AI. Korean is also a natural language, but AI translates Korean into English before processing. This is where the "vibe" differs. Some say this is a problem that can be solved as AI improves. However, there is a gap between native speakers and those who learned a foreign language that is hard to overcome.

There are also employment issues for software majors. According to global job portal Indeed, if the number of job postings for software developers in the US in 2022 (when ChatGPT was first released to the public) is set at 100, the number in 2024 dropped to 35.8. US developer job postings have been declining steadily since the peak in 2022, when the AI development boom was at its height. The current number of postings is less than one-third of the peak.

The same thing is happening in Korea. According to recruitment platform Jinhaksa Catch, the number of new IT developer job postings in the first half of the year dropped from 995 in 2023 to 684 in 2024, and to 564 this year. Oh Chiyoung advised software majors, "Study and research programming at school, but learn AI for employment." AI is ultimately software, so you need to understand programming for deeper development. It's similar to how English majors study literature but also learn business or everyday English for employment.

원본보기 아이콘

However, industry insiders assess the impact of AI on Korea as positive. Oh Chiyoung suggests recalling the time when Korea served as a global testbed for internet technology during the internet boom. The core internet technology was developed by the US Department of Defense, but in terms of applied technologies, Korea achieved outstanding results that were hard to match, earning the title of IT powerhouse.

The IT companies born at that time helped elevate the Korean economy to the next level. Oh explains that a similar phenomenon could happen again in the AI era. He said, "After productivity increased thanks to AI, Japanese startups even inquired about establishing research centers in Korea." This means that Korea's proven applied technology capabilities are now attractive to foreign companies.

It is also noteworthy that many people who have encountered AI early on are considering starting their own businesses. Lead instructor Um said, "While teaching AI, I also started thinking about entrepreneurship." After working at ETRI (Electronics and Telecommunications Research Institute) and serving as CTO of Korea Fujitsu, he retired from the field and began mentoring the next generation. The reason he started dreaming of entrepreneurship is simple. "After just two or three months of learning AI, you can produce talent ready for the field. The money, manpower, and time needed to start a business have drastically decreased. Now is the opportunity."