Recently, many advanced countries, including the United States, China, Canada, Japan, and France, have announced that they will invest trillions of dollars in artificial intelligence budgets at the government level. Naturally, with so much news like this coming out, questions are being raised in Korea as well. “Why isn’t Korea spending money on artificial intelligence? What is the government doing?” I would like to explain this and point out the problem. Personally, I think this issue is very important.
In fact, South Korea is allocating a lot of money to artificial intelligence.
This may sound surprising, but it's true. In fact, even at the central government level in Korea, the combined artificial intelligence budget allocated by local governments, public corporations, etc. exceeds trillions. The articles below report the budget prepared by government ministries related to artificial intelligence in the 2024 budget.
“Isn’t that enough?” Why do people in the actual industry think that Korea does not support artificial intelligence? Why can't I feel it? The reason is simple. This is because most of this budget is being allocated to places that are far from actual artificial intelligence.
As the world has witnessed, hegemony in the artificial intelligence market is currently coming down to “Who has more AI chips and who creates a better general model?” The United States and China are securing chips that were previously imported for gaming, either by disassembling them or via Southeast Asian countries. In fact, even if you place an order with NVIDIA now, it will take several months to receive the correct product. (This is also why I said that we should continue to buy Nvidia from 2022.)
In any case, the NVIDIA product has the best artificial intelligence learning efficiency. However, we have failed to secure this infrastructure nationally. In the recently built artificial intelligence center and wherever you go, most of them are around V100 or A100. Even this isn't much. There are only a few, and some are not even operational.
I have seen many cases where models from 5 years ago are being used. Of course it's better than nothing. However, as those of you who know, this makes a difference of at least 3 times to as much as 12 times in the learning speed of artificial intelligence models . This is a gap that is difficult to bridge through ingenious software or learning methods. As I said before, as model learning becomes a battle of scale and speed, high-performance infrastructure is now becoming a necessity.
So, the government is building data centers in cooperation with regional organizations, but if you look at the budget, the money is not entirely being spent on artificial intelligence facilities. In the meantime, we are building roads, welfare facilities, and running other projects under the name of artificial intelligence budget. Are artificial intelligence startups in Silicon Valley, New York, Tel Aviv, or Paris doing well because they have easy transportation?
The problem is not the budget, but how and who spends it.
We live in the cloud era, and there are many ways to assign and use models with just one click. According to a person who recently said that spending money like this could cause problems later on. In addition, if security issues are raised, they try to build their own facilities instead of the cloud, and then they cannot obtain the latest infrastructure. They buy and build old models that can be purchased, but no one uses them, and only places that want to bid for government projects or small companies come in. The bridle is created naturally.
In addition, these budgets are currently being budgeted separately by the central government, each ministry, local governments, and public enterprises. There are various groups involved in this work, all issuing separate bids. Ultimately, this is bringing about Galapagosization. We need to pool our budget and talent nationally, but that is not happening right now. This is not something that can be done just by setting a budget, nor can it be done by just hiring people.
Is this simply because they are paid a lot so they leave? This is because there is no place in the country that properly supports and supports research. There are many cases where you cannot do research even if you want to. Personally, I think reducing the R&D budget is something that comes from this context, but cutting it off to prevent budget leaks is close to being an understatement. Did you actually catch it? It wouldn't have meant much. Recently, every week I keep seeing talk about government-level fund creation, investment attraction, and bipartisan action in China, Japan, Taiwan, and the United States.
I know that there are several people in Korea who are making similar efforts. However, there doesn't seem to be much interest in it in places where decisions are actually made or budgets are prepared. When was the last time Korea was called an IT powerhouse? Is it still an IT powerhouse? Rather than a voucher industry or a consumptive business that is in a hurry to meet performance, shouldn't we prepare an environment that is sustainable and can operate even after 5 or 10 years?
When to Concentrate
There is a nationally popular project in Korea. These are ‘semiconductors’, ‘displays’, and ‘secondary batteries’. These three are industries that the government takes care of to the extent that there are separate customs laws. Industries that involve all three are electric vehicles and smartphones. The reason why Korea developed so quickly is because it was prepared in the 90s and 00s. How is it now? These technologies are also being pursued at a frightening pace in China, Japan, and Taiwan. CATL, the world's number one company, is looking at the next step in the lithium-ion battery market with sodium-ion batteries, and Japan's Panasonic, Toshiba, and Sony are also sharpening their skills in displays and semiconductors. There is nothing to say about Taiwan.
In Korea, the government talks less about the software industry than you might think. There are companies like Naver and Kakao, but they are often regulated rather than actually supported in Korea. In the case of the Naver-Line Yahoo incident that occurred this time, there was actually nothing the government could do. Well, I need something done to say something, but since I haven't done anything, I have no choice but to just watch. The same goes for startups. In fact, you may have learned a lot from investing in startups during the Park Geun-hye and Moon Jae-in administrations, but that has not been carried over. Collaboration is happening at the private enterprise level. Leading the way are Naver, Samsung, and SK.
I think we need to lead or support this type of construction and infrastructure industry at the national level. In fact, if it is difficult to break the artificial intelligence development ecosystem created by NVIDIA from the outside, you can approach it by creating compatible chips or chips that can accelerate. In fact, creating a new standard when it has already become the standard does not have much meaning and will only accelerate the Galapagosization.
Shall we just talk about digital government for now? Oh, there is already “e-government”? It is difficult for anyone who has used e-government even once to say that it is convenient and good, even if it is empty words. Also, since this is used by central ministries, it is often unavoidable to use it. Even if it comes down to each local organization or agency, it is organized and used independently. If this happens, the budget will be leaking again. We did a lot of things like digital government and DX, but we just did it, but we didn't do it sustainably.
Personally, I'm always curious. Is this a problem in the budget planning stage, a problem in selecting a company, or a problem in supervision (audit)? Even now, time and money are passing by and we are missing important decisions like sand in our hands. I don't know if this article will reach them, but we really need to think about what we will do in a few years. The longer this is delayed, the more people will be deceived and dragged into artificial intelligence services, and the wrong people will make money. It's already happening.
In fact, we have previously discussed whether the National Assembly (politics) is interested in this, but both the executive and legislative branches should think about it. The judiciary needs to do a better job of monitoring this. We already learned why the separation of powers is a valuable system and why our country chose it in middle school social studies textbooks. I hope the system returns to normal.
Summary
We are now at an important inflection point. We hope that with the right decisions and implementation, Korea can leap forward into an AI powerhouse.
There are several problems with South Korea’s artificial intelligence policy and budgeting:
The budget is allocated but not used for actual artificial intelligence development.
Failure to secure the latest AI chips and infrastructure
Inefficiency due to dispersed budget execution
Talent drain problem
Lack of long-term vision
To solve these problems:
Establishment of AI strategy at national level
Efficient and focused use of budget
Securing the latest AI infrastructure and creating an environment for in-house chip development.
Creating an environment to attract and retain talent
Investment and policy establishment from a long-term perspective
Please leave a variety of opinions in the comments. I, too, may only be speaking from a narrow tech industry perspective. I believe that better decisions can be made when diverse perspectives come together.
2
떨
떨어지는 사파이어 시내
안하는게 맞습니다. 해결책이라고 나온 것들이 실제 해결책이 맞나요? 국가에서 주도를 하면, 그 인프라는 어디에 설정하고, 어떻게 공유를 할 건가요?
그 예산은 다른곳으로 돌리는게 맞다고 봅니다.
3
Haebom
예산을 안쓰는 것도 방법일 것 같습니다. 국가적 결단이 있어야 한다고 생각합니다.
개인의 입장에선 더더욱이 자유롭게 결정해도 될 것 같은데 후대를 위해서 대한민국이라는 나라는 어떤 결정을 해야하는가? 하는 생각에 몇자 적어 봤습니다.
적어도 사기업 중심으로라도 재편된다 하더라도 연구개발에 투자할지, 가져다 쓰는 것에 집중할지 같은 결정 말이죠. 개인적으로 클라우드 시대때 한 번 국가적으로 결정을 빠르게 명확히 못했고 그 부담을 지금도 지고 있다고 생각해서 더 길게 적은 것 같습니다.
해결책이라고 나온 것들이 실제 해결책이 맞나요?
국가에서 주도를 하면, 그 인프라는 어디에 설정하고, 어떻게 공유를 할 건가요?
그 예산은 다른곳으로 돌리는게 맞다고 봅니다.