AI Giant Dinosaur US vs China chasing fast, the only hole for South Korea to survive?

■ Broadcast: YTN Radio FM 94.5 (09:00-10:00)
■ Host: Reporter Cho Tae-hyun
■ Air date: January 17, 2025 (Friday)
■ Talk: Cho Sung-joon, professor of industrial engineering at Seoul National University

* The text below may differ from the actual broadcast content, so please check the broadcast for more accurate information.
◇ Reporter Cho Tae-hyun (hereinafter referred to as Cho Tae-hyun): Let's continue with the second part of YTN Radio Live Economy. In the second part, we will continue to talk about in-depth AI-related stories. I think many of you will remember the mid-20th century, but it was called the Cold War era. At that time, the U.S. and the Soviet Union competed as if the world were going to collapse over the development of nuclear weapons. But the composition has also changed in 2025. The theme has changed a lot, too. We will continue to talk about it with Professor Cho Sung-joon of Seoul National University. I just got a text message from the professor. The listener said, "Artificial intelligence technology is developing amazingly fast. I hope it will evolve further into a technology that helps humans." Actually, some people seem to be worried about Terminator or Matrix. Is there any chance of that?

◆ Cho Sung-joon, professor of industrial engineering at Seoul National University (hereinafter referred to as Cho Sung-joon): We can't say we can't, but we have to control it systematically and in various ways to prevent that from happening.

◇ Cho Tae-hyun: Okay. I don't think you need to worry about that yet. I don't know if such an era will come or not. Let's get into the real story. In the first part, we looked at artificial intelligence in the big picture. In the second part, let's look at the competition structure. The country with the strongest AI in the world is the United States, right?

◆ Cho Seong-jun: Yes, that's right.

◇ Cho Tae-hyun: What do you see as the reason why the United States has such an absolute weight in AI?

◆ Cho Seong-jun: I look at three things. The first one is, of course, a person. AI is also from the United States. And the country that created the digital computer that all humans use today is the United States. So actually, when computers first appeared in 1946, people already started talking about AI. I called it the Great Brain then, too. It was a huge machine almost the size of our studio with a vacuum tube.

◇ Cho Tae-hyun: The machine is about as good as today's calculator.

◆ Cho Seong-jun: Yes, but I paid quickly. And it makes perfect memories. Isn't that what people are most weak about? Because I don't remember well and I can't calculate quickly. That's why I'm so respectful. So, what we're doing easily is that of course, he's going to do it easily. But after doing it for decades, it turned out that he was not good at what we were good at. Anyway, in that process, countless people study and research AI, and this is the power of the United States. So there is a lot of quantity and quality.

◇ Cho Tae-hyun: I know, I failed, but all these things have been accumulated.

◆ Cho Seong-jun: It's been accumulated all the time. Another interesting thing is that America is a country of immigrants. So all the talent from all over the world is flocking to the United States. Second, the U.S. just pays an unlimited salary if it is capable. The boss gets this, but you can't get more, there's no concept like this. The boss is the boss and the best engineer. Then you'll get twice as much as the boss.

◇ Cho Tae-hyun: In fact, there is a saying that the outflow of manpower is serious in Korea.

◆ Cho Seong-jun: That's right. That's the power of the United States. The second is funding, after all. Korea has chaebols, and Korea is not a country without money, but there are various types of money in the United States. The money that's been spent on AI right now is Silicon Valley's money. The characteristic of money in Silicon Valley is that it is very gambling compared to Texas or New York.

◇ Cho Tae-hyun: So there are risks, but when you succeed, you can make a big hit.

◆ Cho Seong-jun: It's the money that can make a big hit. The reason why the money was concentrated in Silicon Valley is that the people who started the business in the past are young. This is the money that people in their 40s and 50s have. What does the money look like? That's how he earned it. Since I earned it like this, of course, if my juniors push me again, I'll burst it like I popped it. I'm very deeply rooted. So those things become a virtuous cycle, and of course, if you put 10 in, all 9 will be ruined. But one thing explodes. I essentially don't have this type of money in any country in the world. There is a rich man, but there is no such gambling rich man. That's the second feature of the United States. The third characteristic is that the whole society is very adventurous. You get a lot of the feeling that it's embedded in American DNA. If you think about it with the Americans, it's the descendants of people who came by boat to the unknown world in the past. It's a history of pioneering. Will the bear or the Indian be waiting when they go? But beyond that, this way of thinking that started from the floor in a barren place is deeply rooted. In that sense, AI is actually going to a new world by boat. What happens to our society if we use AI like this? Of course, we worry about this. But these people are like that. Use the line. And if there's a critical problem or something bad while using it, let's punish him. 'He's doing a bad thing with AI.' Then let's catch him, take all his property, and put him in prison for about 50 years. There's this kind of judicial system. On the other hand, our country is in an atmosphere where each other forgives each other. Since the punishment is weak, it's too hard to punish what we're thinking anyway, so let's not use it from the beginning. I think it's a social way in Korea to restrict it hard and prevent it. On the other hand, in the U.S., do whatever you can, do whatever you want. But you know you can't do bad things, this is it. That's why I'm just using it.

◇ Cho Tae-hyun: So it was very surprising to hear that you built a huge language model without knowing that what you said earlier would be like this. These are people, money, culture. Everyone seems to have a comprehensive impact. Let's continue the story of the United States a little later. Is China the next country after the United States?

◆ Cho Seong-jun: Yes, that's right. It's China. China is the human resource I mentioned earlier. It's 1.5 billion won, so it's a lot. Of course, it's not enough than in the United States, but it's still enough to make it. 90% of the AI papers I read recently are Chinese authors.

◇ Cho Tae-hyun: Does it come out that much?

◆ Cho Seong-jun: I can only say that it's pouring out, so shouldn't I read some papers that aren't Chinese authors? Open AI, Microsoft, Google. Even though they belong to a place like this, their names are all Chinese names. So not only the Chinese in China, but also the Chinese in the U.S. It's in China again. Chinese who moved to the U.S. A lot of different Chinese people are participating in a lot of R&D,

◇ Cho Tae-hyun: China is working hard on technology, led by the state.

◆ Cho Seong-jun: That's right. There is no private capital for money, so the state pushes it. Third, society as a whole is not curious about society there. If you do it, you can do it.

◇ Cho Tae-hyun: Because it's an authoritarian country.

◆ Cho Seong-jun: I'm just pushing you because it's going to be a one-man dictatorship. He's a big brother. So there is no privacy issue. So you can get the data indefinitely. And we have all the CCTVs installed on the road, and we have a lot of CCTVs, but we can't use them all. But they use everything in China.

◇ Cho Tae-hyun: What surprised me a little bit before was when the big data team was created at the Bank of Korea, a few years ago. At that time, I asked the National Statistical Office to give me the data, but it was blocked by related laws and was recovering the data. I remember being very surprised by the Bank of Korea, but it doesn't happen in China, there is no such thing. So, we are making progress, so what is the gap between the United States and China now?

◆ Cho Seong-jun: Well, it's hard to say in numbers, but I think it's just about 100 to 60.

◇ Cho Tae-hyun: Then the U.S. is definitely in the lead.

◆ Cho Seong-jun: Yes, he's got the upper hand. I'm also making GPT in China. That's why people don't use it often. Because using GPT means that my data goes to GPT. If I ask questions or, for example, an executive who will make important decisions for Korean companies asks questions about his or her company, I have to tell you my story. Then the data is transferred. But it's very inconvenient to ask that question to China. I think that's why people don't ask questions in China. But if you provide it for free, that's a little possible.

◇ Cho Tae-hyun: So China is following, but there is still a gap with the U.S., but the U.S. is still very allergic. Reactions that keep China in check and prevent it from securing itself at all. Why do you think you react so strongly?

◆ Cho Sung-joon: Not only AI, but actually, Trump's first, Biden, and Trump's second, and if you look around the world now, China is the second country that competes with the United States. I think the U.S. traditionally uses a strategy of keeping the second-in-command below 10 if it is 100. There was a time when Japan was beaten once in the 1980s, and in the 1990s, the Soviet Union self-destructed on its way up, and China had to press a little bit, but one of them is particularly bad in terms of technology, and since this is the future, the United States of course is trying to beat China's development in these technology fields, including AI. So I think AI has come up as a very main target under a big overall plan.

◇ Cho Tae-hyun: If you look at the stories that keep coming out, they were there last night. It's on the bio side, and last night, AI, bio, quantum computer, etc. seem to be trying a lot to block it, but I have a question here. Is this something you can't do to block?

◆ Cho Seong-jun: It can be difficult. First of all, I think we can delay it.

◇ Cho Tae-hyun: We can delay it.

◆ Cho Seong-jun: It's causing a delay. a delay operation So, if you think about what will happen next year in 10 years, China is a country with a lot of contradictions. It's coming out little by little, so wouldn't you think that? If we get along so well for 10 years, there may be hope that self-destruction will occur and China will be broken down. Anyway, we're going to block it and keep developing, so we're going to keep a gap.

◇ Cho Tae-hyun: So when you look at big tech companies, most of the big tech companies that are leading the AI era are in the United States. Which company do you think is leading the way right now?

◆ Cho Seong-jun: Google in terms of technology. Actually, the first thing I mentioned about Transformers and language models to make is Google. Actually, Google was the first to make this AI. Didn't the people who came out of it just go to open AI and launch it there and make a mega and pop it like that? Is Google actually a bad old time? I made such an AI once and put it in because it made me talk rough, but the open AI was very smart and I took care of it to prevent him from talking rough. So, it is ChatGPT that prevents hate speech, prevents discrimination, and then prevents people from answering, and hangs safety devices behind it. At first, Google didn't do that. Another thing Google hesitated about is that they are one of the search markets and the company makes money by searching, right? But when AI comes out, the search dies. Then it becomes my own flesh, so I had no reason to put another product on the market that I'm cutting off the popular product, and I kept hesitating and thinking about it, and then the open AI exploded. But I'm following him hard.

◇ Cho Tae-hyun: That's why you're saying that Google is leading the way.

◆ Cho Seong-jun: Technology. Technology itself is like that, but technology has spread a lot. People who created OpenAI in the early days quit and created another company, so there are a lot of companies like that here and there. Meta, a company that has Facebook and Instagram, is also working very hard on AI. The meta-feature is that we're making a lot of open-source, free software. AI, so the technology there is also good.

◇ Cho Tae-hyun: Because people like free stuff.

◆ Cho Seong-jun: Many other start-ups have a lot of technology in this AI technology right now.

◇ Cho Tae-hyun: I think it's something that everyone touches at least once in Silicon Valley. However, it is a little different from big tech companies, and I think Nvidia is also a company that is receiving attention in the stock market that cannot be left out when it comes to AI. As Nvidia has been evaluated as a very suitable product for AI learning, demand is increasing and performance is improving a lot. What kind of demand for GPU chips will continue in the future?

◆ Cho Seong-jun: Yes, it's overwhelming. It's overwhelming right now. Originally, GPUs are called G graphics. So, computer game players said that GPUs are better than CPUs to change the screen quickly, so I made GPUs and put NVIDIA into game PCs. What's interesting is that one of the AI researchers took too long to learn AI with CPU, but I think that person plays a lot of games. Someone tried it because they thought it would be faster if they learned it with GPU.

◇ Cho Tae-hyun: In a way, it's a change of thoughts.

◆ Cho Seong-jun: It's a change of ideas. I wrote it down as a game, but I just tried to use it, but it got really fast. Then, as GPUs began to be used for AI, AI could not learn without GPUs and could not be infested (conclusion drawn). In the meantime, Nvidia developed technology well and today it exploded like this. Then, what will happen in the future? I think the turning point has come to the device I mentioned earlier. So, it goes down to the device, and as I said before, it needs to be very small. You can't put AI like the dinosaurs in the data center into your smartphone and not into robots. It has to be small, so now the existing architecture has to be completely changed. If Nvidia raised its hand and said, "We prepared it when we had the technical problem of what kind of hardware we should burn a small AI again," then we could say that we should keep our dominant lead and use our own devices when another semiconductor company appears. You can't tell that.

◇ Cho Tae-hyun: The future is not fixed.

◆ Cho Seong-jun: Yes, so anyway, there is no company that can compete with Nvidia right now. It's almost a monopoly right now.

◇ Cho Tae-hyun: To explain the background a little bit, the CPU has good performance, but it has to be a little slow because it operates sequentially. GPU's performance may be slightly poor, but it is called a parallel operation, so it is calculated at the same time, so you can understand the concept that this is a little more advantageous for artificial intelligence. Now, let's talk about Korean companies. The listener asked, "Is there a company that makes AI software in Korea? If so, what is your international competitiveness? ” You sent me a text. How much is it? Do you have it?

◆ Cho Seong-jun: We have Naver. Naver has created an AI called Clova X. So Clova X's strength is that she learned a lot in Korean. GPT such as ChatGPT says that even though Korean documents are less than 1% of all documents, they are good at Korean, but Clova X Naver has created AI that is better at Korean by learning more Korean.

◇ Cho Tae-hyun: Can we pick Naver as the company that is showing the best performance among the companies in Korea?

◆ Cho Seong-jun: Yes, so the company that makes AI software is actually Naver. It is almost the only LLM and

◇Jo Tae-hyun: In terms of software.

◆ Cho Seong-jun: But there's hardware again. When it comes to hardware, we can't say without Samsung Electronics and SK Hynix. So if there is a GPU you mentioned earlier, the memory should always be next to the GPU, but I told you earlier that the memory is processed in parallel with the memory next to the CPU. Then, since it's parallel, a huge amount of data has to go into the GPU and then come out. Instead of such a so-called 10-lane road, we need to create a 100-lane, 200-lane road. Then, can't there be a lot of cars going back and forth at the same time? HBM allows a lot of data to flow back and forth in that way. This is Samsung and Hynix, and in fact, Samsung Electronics is behind Hynix, but in fact, is the capacity of the factory to supply this huge amount? In this regard, many companies will wait for Samsung Electronics to become soon.

◇ Cho Tae-hyun: But as you said, I have some hardware strengths in Korea, but you're actually saying that Naver is the only one in software. Then, should we say that such things as competitive presence in the AI field are not that high except in terms of hardware?

◆ Cho Seong-jun: Yes, that's right.

◇ Cho Tae-hyun: How much do you diagnose the current situation?

◆ Cho Seong-jun: So hardware is the world's best top, and the software side is actually a foundation model for one Naver event, while the U.S. has 60 to 80 pieces. Of course, that's not all, but if you count like that, we're one. So someone ranked me in the world, and my country is 8th or 7th. But actually, I got 100 points in the US and 60 points in China. It's less than 10 points from third place. So, it's actually the same as whether I'm in third place or fifth place or seventh place.

◇ Cho Tae-hyun: Then we can say that it's a very urgent time to catch up and follow these things. Is there any movement in the government?

◆ Cho Seong-jun: Yes, the government is working very hard. The same goes for our Ministry of Science and ICT. The Ministry of Commerce and Industry and the Ministry of Science and ICT are investing a lot of money in R&D to create and develop these things. The Ministry of Commerce, Industry and Energy is focused on how to use this in companies or in manufacturing sites. Also, the Ministry of Employment and Labor is not a matter of how to educate, how to teach AI to people in companies, and how to educate our growing children about AI. There are a lot, and the Ministry of Public Administration and Security is actually related to how to open up the data held by public institutions and create industrial value by learning with data. Of course, how can the Ministry of Health and Welfare use our medical data?

◇ Cho Tae-hyun: That's right. That's because you have the most data.

◆ Cho Seong-jun: The Financial Services Commission also said, "How can we use this financial data? AI is a hot topic in all departments of Korea, as all departments in the country are involved in all fields, so AI is emerging as a hot topic in all departments.

◇ Cho Tae-hyun: That's how the government is making a lot of effort. Recently, we will invest KRW 1 trillion in funds to develop AGI and other things I mentioned earlier. Various policies are coming out, but as a professor on the field and a person on the front line, what part do you want to be most disappointed about? What would you like to say if there are any parts like this that you want to improve quickly?

◆ Cho Seong-jun: Unfortunately, I mentioned earlier that there are two major pillars of development and utilization, but in fact, we are making big companies such as Naver or Samsung Electronics. It's a secret, but I'm making it internally.

◇ Cho Tae-hyun: No, because I think he'll definitely prepare it.

◆ Cho Seong-jun: So Samsung Electronics is making it for its own use, not for external services. Naver is making it for external use, so for such large companies, what the government can do is actually help remove regulations. On the other hand, if we move on to utilization, it will be the work of the Ministry of Employment and Labor or the Ministry of Education as I said earlier, but young people, and we are in our 30s, 40s, 50s, retired 60s, and 70s who are currently working for the central companies of Korea, and the whole nation is actually no longer computer-blind, right? Just as we use all our phones well and there is no such thing, we can use 'AI blind' well if we resolve it quickly. So even if we don't develop well, we can use it well. Because development is good, but even if development is not the first place, if we use it well, Korea's GDP will rise. Society is supposed to develop.

◇ Cho Tae-hyun: We also need to find a way to make good use of it.

◆ Cho Seong-jun: Because we use Word PowerPoint Excel. Aren't they all foreign products? It's all made by Microsoft, but all organizations in Korea have used these software to increase productivity, and Python is a language made in the United States, but many people code for it and increase productivity, so you can use AI if you have it in that dimension. But if you don't know how to use it, you can't use it whether it's made in Korea or the United States. Now, if we can't use it, for example, we have one company that we develop in the world, and if we can't use it, then only that company makes money, and it's not very fun in Korea as a whole.

◇ Cho Tae-hyun: Utilization is as important as development. As expected, he also mentioned the improvement of the regulation that always comes out in the aftermath of the regulation. So far, we have talked about various stories surrounding the AI industry with Professor Cho Sung-joon of the Department of Industrial Engineering at Seoul National University. Thank you for talking today.

◆ Cho Seong-jun: Thank you.

#AI #US #TechnologyDevelopment #Investment #GPU #Nvidia #China #ChatGPT #ChatGPT #ChampionCompetition #Naver #SKHynix #Samsung Electronics #Google #OpenAI #SiliconValley