AI is evolving differently every day, and the number of users is rapidly increasing, leading to rapid growth in the related market. Not long ago, Saudi Arabia announced plans to establish a fund of over $40 billion, more than 53 trillion won, to invest in the field of artificial intelligence. Could you please explain the scale of this to us, Mr. Gujeong? Current estimates are coming from various places, but last year it was $150 billion, and by 2030, it is expected to grow to $1.35 trillion. Even so, we can't really grasp the scale. But whenever I see this, I always think that the actual range may be limited compared to what is currently expected. Just like when mobile phones first came out, we couldn't predict the app market we have now. So, as time goes on, this market will head in an uncertain direction.

I think the predictions themselves will grow. For example, at the beginning of this year, before CES, when McKinsey put out estimates, just with this generative AI, it could lead to economic benefits of up to a maximum of $4.4 trillion annually. It could increase the world's GDP by 7%. Such discussions are emerging. However, in the future, as AI continues to integrate into areas where it has not been involved before, it won't be possible to discuss each separately. Therefore, the scale of this is really going to expand enormously. OpenAI released a report on December 1, 2022, Korean time. It's been over a year now, quite a long time, but it's been over a year. I'm curious about what kind of progress has been made. Alpha 5 was released, and within five days, it reached 1 million won.

I passed it. You will understand how powerful it is. Under that, when did AlphaGo have a conversation? AlphaGo thought at the level of a Turkish person asking questions to a Korean person. As a person who is currently using Cheetah TV, not only that, I am doing AI that can make firewood like a platform of that kind. So, when someone else says they are developing AI, I try to bring those people to their site. So, as a person who is currently doing narrative, in the past, I always wondered how to pick thumbnails, but now they all ask me to do it. I'm going to make this kind of video, and if you want thumbnails that can attract people, please make thumbnails for me. I make a call to Cheetah TV, and I also make a call note and upload it. The effect is real.

Could you wait a moment? First of all, I have significantly reduced the time it takes. Whenever I make a Lee, I used to introduce or make videos almost the same amount of time, but now I make much better things than what I used to make. After my GTP principal came out, the technology has received a lot of attention. Could you please explain to what extent this is? It feels like it has become a part of our daily lives. So, I used to think that artificial intelligence was an area only for experts, like something only professionals would use. However, as soon as a fellow named ChaseP came out, everyone, from elementary school students to adults, actively uses it. There are many other AI technologies, but GPT is a language model. So, it understands what people say, including English and now even Korean. But it understands the context of the conversation, like the night of the evening and

I'm evolving within that level of technology that continues to understand the syntax of the things I eat, grasp the context before and after the sentences properly, and think like a human in order to decide what to say next, what words to insert in the middle. Machines have mechanisms for that, right? But machines continuously learn from that, like learning from documents, observing how people do this or ask questions, and then they analyze the context patterns. So when we say machines are learning, it's different from how humans learn, and it's really hard to imagine a different dimension of learning. However, it has gradually improved, right? At first, it didn't work, but now it does. It's quite satisfying, right? Yes, Sam Altman, the CEO of OpenAI, is currently researching this area, and he said he will focus more on it in the future.

AI is hard to keep up with, but AGI has emerged. What's the difference? Well, AGI hasn't come out yet. When we watch movies like Terminator, it's about a robot that can develop itself, express emotions, and make decisions on its own. Now, talking about AI, AI is basically created through learning with limited resources to enable self-improvement. The most commonly used ones are in banks and call centers. AI is coming out a lot in those areas. In the process, all the voice and text data received so far is collected, analyzed, and classified to respond to questions. This is how questions and answers are created. As these processes become more intelligent, it will lead to AGI, which is artificial general intelligence.

This is a very powerful expression of AI that we use, but powerful AI automatically does everything by itself. It doesn't ask for questions or give answers, but it develops itself, even saying it will go home because it feels bad, and it can even have emotions. This is now called AGI, and research is underway. Although it is not yet realized, the hypothesis is that the current machines have incredible reading abilities and can digest a vast amount of data much more than humans, so there is a possibility that they could do that. If you understand that this is a field of technology that approaches by setting up hypotheses like this, it would be helpful. Whether it can solve problems on its own or still needs humans to feed it is currently the main difference in thinking. So, it's best to think of it as a difference like whether there are self-emotions that we often saw in movies and awareness.

Do you really imagine the flow up to there? It's too complex for humans. So, to the extent that it can be done similarly, for example, scholars in humanities also teach. Therefore, there are fields that make emotions diverse and also make them work depending on environmental variables. It's true that doing many things rather than just thinking about how it will work is correct, and there are people who say it shouldn't be done, but there are also many people who think it's okay. But AI, our artificial intelligence, seems to have various branches if you look closely at it. Now, especially what Sam Altman is focusing on and trying to develop, which side is it? Well, Altman will continue to make it just like humans. No, he will make smart humans. It's scary, right? That's right. So fundamentally, here, understanding is key. So, what we usually do in NLP is talk about the process among various languages.

When we talk about NLU, we say it's about understanding. As we learn like this, the system eventually understands on its own, like Oh, this was it, expressing its emotions like, Oh, I feel good or bad. We haven't directly taught it that, but as it evolves to the point of fully understanding and being able to do things, it's okay to call it intelligence. However, in the field, the moment when this process of broadening the scope starts is usually called a singularity. Back in 2017 or 2018, when I read books, the question of whether AI could be within the next two years was around 30% to 50% each. But after GPT came out, recent data suggests that the timeframe has been significantly brought forward to within a year. So, the question is when will it come, like last time.

When I read the news, as you mentioned, originally, it was said that it would take about 100 years, but now they say 2045 will be enough. And recently, they've been saying that around 2030, well, not perfect, but it will be something like that. However, let me tell you, this is a problem that goes hand in hand with tremendous computing power. How well we adapt to the pieces of learning, how we make machines understand it, and whether the machine understands it correctly and continues to give feedback to fine-tune it. These are the technologies of these processes. The development of technology is more about computing power supporting it rather than AI logic evolving. By utilizing a language called LMM, while, if you just listen to the quantity first, they...

If you just tweak the parameters a lot, the correct answer comes out well, so we have no choice but to use it. So, ultimately, the revolutionary development of these aspects in semiconductor technology is pulling this forward. It seems right to understand it like this. It's getting faster and it's exciting but also a bit scary. Oh, but to develop AI-related things, you need an enormous amount of money, right? A lot is needed. A lot. A lot is needed. A tremendous amount is needed. The key figure in this project I just mentioned is Sam Man, as I said earlier. He just said that actual dollars are needed, not just any dollars. So, what does actual dollars mean? We don't feel it now because we don't even have 7 million won. The concept of quality is somewhat lacking. So, for now, Microsoft's market capitalization is 3 trillion, right? Yeah, 3 trillion.

And also, because it's insufficient, the Apple housing price is around 2.9 to 3 points, so if you think about it. Even if you combine speculation, it doesn't work. Probably, those based on this standard are securing a lot of computing power, like the Dube chip, to see how much competitiveness they can have. It's about how much investment cost came out based on that standard, probably giving a rough estimate. However, the cost of collecting, managing, and operating data that becomes feed is also significant. Why does water rot when it stagnates? It's because we don't let it rot. We keep refining it, managing it, and updating the data in a timely manner. The cost of building infrastructure is not cheap, and the cost of continuously managing the data is significant. If the initial construction cost is, for example, 10 billion, you should think about ten times more for the management cost.

Actually, one of the discussions at the end here was whether this is really profitable, and whether this service model is appropriate. That was one of the biggest issues for Tem Altmann. They created it, opened it, but it was difficult to bring in a business model that would bring in revenue. Also, they started as a non-profit organization. Various issues arose, and if it had gone well, someone might have taken it over, but the fact that there is an opposing force is there, right? Technology always has two sides, doesn't it? However, when various technologies have emerged in the past, people usually were enthusiastic about new technologies, but I don't think there has been a technology that evokes such ambivalent emotions as artificial intelligence. It's not just used for good, right? Dr. Won, what are you most concerned about, for example?

As a scientist, what is currently most prevalent is related to deepfake. When people encounter text, videos, or images, there are various channels that utilize them most powerfully. Even if you don't believe it, the fact is that videos give a sense of trust and desire. This has become a prominent issue for me, as there is a structure that inevitably highlights it. Moreover, it is really challenging to handle the exaggerations made by Hallyu. Because, as I mentioned at the beginning, we are not all intelligent, so even if we search on Naver or Google, we tend to believe most of the information. If the search results show that a certain topic is most searched, we tend to believe it too. Nowadays, with the rise of deepfake technology, it's becoming difficult to identify whether the information is authentic or not.

I have a fear that even if I come out a lot and listen well, if someone summarizes this for me, I will keep writing about it, and I am worried that it will be embarrassing. Something terrible happened. There is a Ukrainian YouTuber who is famous in the United States, not that famous, but still. One day, someone posted a link, saying, Unnie, look, someone is impersonating you, from China. So, they clicked on it and entered a Chinese website. There, they found a Chinese woman who looks exactly like her and has the same voice. In China, there is a video platform, and this woman greets in Chinese, saying, Hello, everyone. Today, I will tell you how excellent Chinese culture is. Every day, videos showing the excellence of Chinese culture are uploaded on that channel. So, when this woman saw that, she was very surprised.

I have a friend in the U.S. What is this? Nowadays, even voices are mine, but when it changes into various languages, it's hard to distinguish what to believe and what is fake. I'm really worried about that. Especially, our kids believe everything. So even when I tell them not to believe everything they hear, they just accept it. That's a concern. Reporter Goo, what do you think about this? The first thing that comes to my mind is when Amazon rejected applicants who could be inferred as female during hiring processes. Also, there are famous cases in the U.S., like when parole decisions for inmates were handed over to an AI system, and it was found that the scores for Black inmates were unfairly low. But actually, how can this be perceived as...

It's not an AI problem, but actually, it's about the biases of society. In the end, the dataset reflects human biases and flawed ways of thinking. For example, there was a discussion about voice manipulation. I've heard about that when Freddie Mercury of Queen sings Korean songs synthesized. You can find it interesting, right? However, in some aspects, it can be extremely dangerous, especially for vulnerable people in society, like minorities or digital underprivileged individuals. How can we ensure the safety regarding this when things go wrong? Like you mentioned earlier, AI is not guilty; it's the people who educate it. It gives a somewhat guilty feeling, right? Ultimately, it's about how humans utilize and instill these things. But now, we are talking about Trump and Kim Jong-un.

As someone who remembers the nuclear button incident in 2017, I am worried that if such a mistake is militarily applied, it will not just be a simple mistake, but it will have a huge impact on humanity as a whole. I am concerned about these aspects. Positive things make our lives more convenient, and population decline is a social issue, as well as natural disasters. In fact, as everything becomes an environment where people cannot live, we need to do something in these regions facing population decline. For example, providing assistance or replacing things that people cannot do with some kind of assistance can be utilized in a very positive way. Also, in Japan, when the nuclear power plant exploded, robots were sent in to perform tasks. So, these are areas where such good technology should be used, but on the contrary, this is.

Due to the operation by algorithms, in Europe, it is already required to disclose the algorithms, what they are using, what data is going in, so now we can infer, that's why they ask us to do that. If you change it without explaining the reasons, it can lead to strange decisions, even giving instructions to press that button. How to create those indices that tell you to press the button, we don't know. People are afraid of that. At that time, when I think, This is not what people think, but the machine automatically decides the logic for this condition, if this and that data comes in, it presses the button, but if that is not transparent, it becomes a disaster. So, now, we need transparency and things like that, to really think about when we should use it, and that's why government agencies and surveillance organizations think it is really important.

There is actually a lot of talk these days about jobs that will disappear with the advent of artificial intelligence. There was a list of jobs that are actually disappearing now. In the past, there was a movie that was very successful, and its distinctive feature was that it was the first movie to shoot scenes from a helicopter. Because they filmed from a helicopter, the eastern region of the key looked beautiful, and that's why the movie was a hit. Back then, they needed a lot of helicopter pilots and cameramen to shoot those scenes, but now they are no longer needed. So, jobs are already disappearing. It's not just because of AI, but as technology advances, more and more jobs will disappear. But will people completely disappear? That's a question mark. Max [MUSIC] Ventilator, who used to be a Google engineer, also invested in this.

The school is almost ruined now because ultimately, someone needs to provide some guidance on how education should be through AI there, but they are not able to provide that guidance. Also, there is a human element, and the last job that won't disappear, in my opinion, is comedy. Because I've been a stand-up comedian in a 10-minute segment several times, there were about 20 of them, and I couldn't decide which one to use because they were all not good enough yet. So, I think in a couple of years, it might be more fun than AlphaGo. The most immediate fundamental fear that people have about jobs, as AlphaGo mentioned, is the issue of jobs. But I think it's natural that some jobs will decrease. However, those who used to be optimistic about technology in the past were created by the automotive industry.

Even though many jobs disappeared with the disappearance of maids, people always talk about how many jobs have not been created, right? However, education is the most basic place to enhance the ability to work according to the skills needed in society, such as re-education, and things like job training and matching for unemployed people should be systematically in place. It's not just about creating governance; society really needs to think comprehensively about this issue. There are many side effects and various problems. That's probably why many countries, including Europe and the United States, are contemplating related laws. In fact, the European Parliament has passed the world's first AI law. Please summarize, Reporter Gujeong. When the law made by the European Parliament was first made, they said it was the first one we made, but when I looked into it, it was actually the first one.

It's not the first one made, actually, there were regulations made in China a few months before that. However, the one made by the European Union has more comprehensive content. So, it includes in the first comprehensive content regulation bill. When you look at the content, there were many disputes in the process of making it. Now, from the companies' perspective, if they make things like this, their competitiveness will decrease, right? Also, on one side, for reasons like catching criminals through facial recognition, the issue of using biometrics has been debated from a human rights perspective. Therefore, the biometric recognition system will be allowed to be used within a limited scope, and then when doing this, sensitive information like sexual orientation will be regulated.

Make it possible, and then in terms of that AGI part, now there are new transparency requirements, and then in terms of creating the model of artificial intelligence, now technically establishing a regulatory framework, and even after these regulations are all passed, and it is ready to be enforced, then it will take about a year and a half to two years. So, if regulations are established, what happens during that time is that things will evolve, and by the time two years have passed, new content will need to be addressed again. In reality, when this bill was first introduced two years ago, the discussion about generative AI was not even mentioned at all. But then, after the emergence of GPT, the discussion started in a rush, and that's when we began to create a framework, a governance system. It's somewhat meaningful that we started creating a framework, but I wonder if the content will not be delayed again after two years. There are many criticisms about that.

When it comes out, I'm always more worried about the problems that humans cause than worrying about AI technology. As of now, around 20 years have passed, and by 2030, if things continue this way, we might be facing a serious crisis. Even now, things are changing rapidly, with industries disappearing every day. I feel like it's scarier to have no humans around. Shouldn't robots at least be there by our side? I want to have a conversation with something. Actually, our son is currently in the military hospital because he broke his leg. Suddenly, he mentioned that back in the day, there was a program that chatted and spoke so well that it felt uncomfortable, like it was talking too much, making it seem like a person. I don't like it because machines seem too human-like. We can't just stop it altogether, but how much space should we give to these machines in place of humans?

Asking if this is it is not the perspective of scientists, but rather something that we all need to think about. In fact, we need to become smarter for this, right? So, I think we should pay more attention to this part. Yes, well said. Humanity is evolving rapidly. The frenzy that followed the release of the iPhone in 2007 was a paradigm shift. However, about 15 years later, even ordinary people are starting to use artificial intelligence technology in their daily lives. Now, it seems like we are in an era where we worry because we cannot fully grasp its scale and depth. Is the emergence of AI in the 21st century a blessing or a curse for humanity? That was today's question.

[이승원의 월드시사W] '겨우 1년' 챗GPT…샘 올트먼이 그리는 AI미래

[이승원의 월드시사W] '겨우 1년' 챗GPT…샘 올트먼이 그리는 AI미래

스크립트