Professor Wong Kam-fai of the Chinese University of Hong Kong, an artificial intelligence R&D expert with 40 years of experience, is on Straight Talk this week. Professor Wong talks about how the future will be driven by AI development, and the importance of openness in developing AI tech globally.
Check out the full transcript of TVB’s Straight Talk host Dr Eugene Chan’s interview with Professor Wong Kam-fai:
Chan: Good evening! This is Eugene Chan. Today's episode of Straight Talk is my 200th show, and I would just like to thank all the viewers for your support. I have truly loved hearing from you. Your feedback is invaluable, and, so, please keep sharing your thoughts with us. This week, we have Professor Honorary Wong Kam-fai with us. Professor Wong has over 40 years of research and development experience in AI, and is a pioneer in artificial intelligence research in Hong Kong. Currently, he is the Associate Dean of External Affairs at The Chinese University of Hong Kong as well as vice president of the Hong Kong Professionals and Senior Executives Association. Professor Wong is also a member of the Hong Kong Legislative Council and the National Committee of CPPCC. He has received numerous awards, including the 2022 Qian Weichang Science and Technology Award and the 2024 St Francis Prize. Professor Wong, welcome to Straight Talk again.
Wong: Good evening!
Chan: The reason why we invite you today is we want to talk about a very popular subject. The title of the show tonight is “Cracking the code: What makes DeepSeek unique?” We're going to talk about DeepSeek. So, maybe you can start by telling us what DeepSeek is and how is that different from other generative AI, such as ChatGPT?
Wong: Oh, I think I better use it in a more layman term. Well, this kind of software, generative AI, is very useful in our daily lives. Say, for example, you can treat it as a smart personal assistant, for example, like your secretary. So, imagine that every time, for example, you have to write a speech, you will ask your secretary to do it for you. And now your personal assistant is even smarter. Well, it can actually write programs for you. It can actually write the PPT for your presentation, and many more. And that's why it's called generative. I mean, the conventional artificial intelligence is mostly on decision making, but now that the generative AI systems can actually help you to generate, be it your speech, be it a program. So, this is why that generative AI is so welcomed by the public. And it all actually started in the year 2022 … where ... well, ChatGPT came in, and everybody was actually wowed, so pleased about these sorts of services. But the ChatGPT was actually produced by OpenAI, but they have a way of doing it. The way that they do it is: the bigger the better. So, if you ... if I want to have a more capable ChatGPT system, they said, “Oh, you need to have a stronger or higher computational power”, and, therefore, you have to invest more in hardware.
Chan: Right.
Wong: Okay. And what makes DeepSeek so exciting, which came in right at the beginning of the year, is it's actually something generated by a bunch of kids.
Chan: Really?
Wong: Yes, you know, with about five to eight years’ experience, you know, just college graduates, and then they are actually graduated in China. You know, these are Chinese students, and what they did was solve the same problem as ChatGPT, but with a cheaper price.
Chan: Right. So, KF, so you're saying that the unique part of DeepSeek, compared to the ChatGPT that you said it came in 2022, it is cheaper, and anything else that makes it more unique?
Wong: Well, it is cheaper with comparable performance as well as it is open source. I mean, the open source part is very important, because that makes this particular system universally applicable to everyone. Well, you know, we'll talk about ... I hope that we can talk about that later in the Paris AI summit there. For example, there is one attribute that this so-called agreement tries to promote, which is inclusion. Inclusion means that, well, I'm poor, does it mean that I cannot use this sort of brilliant software? I mean, when you think about it previously, when the software itself is so expensive, then the underprivileged people, or even the poor people, may be deprived from using it.
Chan: Right, before we move on. You mentioned the term open source. I'm sure the viewers would like, you know, we hear this all the time. What is meant by the word open source?
Wong: Well, firstly, open is the idea of open to promote open innovation. The idea of open innovation means that, you know, research, creativity, innovation is built on top of each other, like LEGOs. So, if I do not open my idea or open my creation or invention, then nobody can actually build anything on top of it.
Chan: I see.
Wong: So, let's say, for example, our Nobel Prize winner in 2009 in physics, our professor, Charles Kao, what did was his research invented the basics of fiber optics. And then, without fiber optics, we would not have the internet, and without the internet, we would not have our social media now, and without social media, we would not have our digital economy. So, that's why Charles Kao was very unselfish and opened that creation, opened that patent up, opened that intellectual property up, so that we can incrementally build on top of it for something good, for humanity.
Chan: Professor, when you said it is not open source for ChatGPT is it the reason why some people, somebody said to me, in Hong Kong, we can't get access to ChatGPT.
Wong: At some point, because of the conflict between America and China, some of this advanced technology, and including advanced AI technology, were not allowed to export to China, including Hong Kong.
Chan: Right. KF, I also read that in DeepSeek. Will you say that is more beneficial to the Chinese-speaking community, because I read much of the successes due to the incorporation of Chinese characters during its pre-training phase. Can you enlighten us on this?
Wong: Well, the way that you built such a system is based on what we call training. Training means that you collect a lot of past information, and then you feed it to the machine, and they learn the knowledge behind it. So, then think about it. I mean, in ChatGPT, or ChatGPT, the background knowledge itself is built from so to say, you know, Western culture, okay, so when you try to ask previously when you try to ask a question, kind-of related to the history of China, you may not get the right answer. Okay, so that's why we think that DeepSeek is very disruptive, including opening up, you know, the China community, especially from my profession, the education community, and then now we can make good use of this system to learn more about Chinese culture.
Chan: Professor, just now you said, right at the beginning of the show, we treat this generated AI as your own personal assistant. That means, who should be the primary users of such a system?
Wong: Anyone.
Chan: Should it be like the businesses, the individuals or even corporations?
Wong: Anyone.
Chan: Anyone can use it.
Wong: Actually anyone, anyone can use it. The examples that I gave, like secretary today. So, this is for business, and for example, students, they can use it for writing their thesis. I think you heard a lot about it, be it positive or negative, a lot of students were using ChatGPT to do their research.
Chan: Right.
Wong: It can be positive depends on the way that you use it. It can also be software programmer, as I said, it can generate code for you …
Chan: Right.
Wong: … or even architects, they can have the initial building plan for you, and even these days with a system called Cora, Asora, that you can actually produce videos.
Chan: Really?
Wong: But beware: there is the negative side of it, even the gangsters or the bad people can use it to build a better deep fake. And there's a lot of cheating behind that. That can also make good use of generative AI technology too.
Chan: Right, we often also talk about business to business, B2B, or even business to consumer. How do you see the advantages of using DeepSeek in such an environment?
Wong: Well, for B2C, then it is what we call a personal assistance. Well currently, the personal assistant, when we talk about personal assistant, we will think about it for big entrepreneurs, it will not be that case anymore. Everyone, I said, everyone will be able to use, will have a personal assistant. So, before you go on a bit of a holiday. For example, you can consult your personal assistant to find out what to do. Okay, that's the B2C part, and for the B2B part, as I said, your secretary, okay, your business development officers, you can actually ask your DeepSeek or ChatGPT to suggest what you can do. For example, if I were to go into the market in China, I probably would go for DeepSeek to find out what I can do.
Chan: Right, professor, let's take a short break now, and viewers, we will be right back.
Chan: Thank you for staying with us. Professor Wong from the Chinese University of Hong Kong has been sharing with us about the latest breakthrough in generative AI DeepSeek. Professor, you have already touched on it, although in a very short time. You have given a very clear explanation that you can treat DeepSeek as a personal assistant, even at all levels. But with all new technology, they always come with risks. You mentioned about the, what we call, the ethical concerns a bit earlier about, say, even the criminals can use it for a better deep fake. What will be the potential risk or, as I said, ethical concerns that I mentioned, to something like DeepSeek?
Wong: Well, whether you like it or not, in the future, in five or 10 years, people believe that when we are working on AI, our vision is to have what we call digital humans, or we have artificial intelligence that can replace humans. This is something that we talk about a lot. And whether you like it or not, it is coming. It is becoming more prominent, this sort of problem. So, think about that, in the future, what would happen in the real world is that you will sort of coexist with digital humans. So, just talk about humans, when I do something, when I do my work, when I do my job, we have a code of conduct. We have to be ethical, we have to be inclusive, we have to be open, we have to be transparent. These sort of elements are something that we have to remember, we have to put them into practice. But now, think about the future, you are co-working with a digital human. What if this digital human is not behaving the same way?
Chan: Professor, I just, as you are talking, I just remember watching some movies that humans are run by robots one day. Is it possible?
Wong: Yes. Well, people are worried about that. But that could happen. But therefore, just like education, just like ethics education that we are doing, what we try to do in university, we try to train a decent man or decent person. So, at the current moment in time, we have to think about digital human. We have to educate, you know, decent digital human as well.
Chan: So, as a parent or as someone like yourself, would you advocate everybody learn about this?
Wong: Obviously. As I said, whether you like it or not, artificial intelligence will be a tool. Or even, I sometimes put it, it is just like English or Chinese, like a language that we have to practice in our life.
Chan: Right. KF, let's move to the actual real world with this DeepSeek coming to our knowledge. I have realized that it caused quite a stir in the Silicon Valley. So, how does this impact the race of AI supremacy?
Wong: It is very destructive. Well, actually…
Chan: Destructive is quite a negative word?
Wong: It is positive or negative.
Chan: Right.
Wong: So, there are some people saying that DeepSeek is a Robin Hood, okay? What happened was in the OpenAI time, in America, they tried to say that the bigger the merrier. So, if you want a bigger system, a more powerful AI system, you need more competition power, you need to buy machines. Therefore, you can see that in Wall Street, in the market, actually a lot of money actually pour into the AI community, the AI economy. But when you really actually look at it at this point in time, many, including OpenAI, these companies are not making money, they are still burning, burning money. But now DeepSeek come in, and they say that I can do the same job as you, only with 5 percent of your investment. So, the investors is actually jumping up and down …
Chan: I am sure …
Wong: … thinking that whether I should still invest in OpenAI or not. And then you can see that there are some people are thinking seriously about moving somewhere else.
Chan: So, what actually happened, say, in a place like Wall Street? Was there a huge reaction there?
Wong: There was a huge reaction. Well, when I said Robin Hood, it means that, well, these people who are actually getting the money in the AI community are only 1 percent or 2 percent. And there are many people, 99 percent of the people are actually, you know, not making any money at all. I mean, the actual company itself is still burning, we don't see any effect. But these people who are actually speculating with this or publicizing this AI investment are making a lot of money. And they should not, people think that they should do something proper, you know, permanent productive, but they are just doing investment making … it is a money game, that’s all.
Chan: So, obviously the investors have huge concerns. So, we have also heard of reports that some US companies are banning the use of AI tools. Does DeepSeek face a similar push back as well?
Wong: Sure, sure, sure. But all out of … sort of business competition reasons, it has nothing to do with the technical part of it. I mean, as far as I am concerned, being a scientist myself, as I said, open innovation is very important. If our invention, if our research results are not open up, then it would not be sort of incremental, and that would hinder the advancement of sciences and arts.
Chan: Right. KF, I am going to move to an area that you have just mentioned briefly in the first part of the show. We talked about the recent AI Summit in Paris, which highlighted growing divisions among global leaders on how to regulate AI. We have talked about the importance of it. And with the US and the UK refusing to sign a declaration on the fact that it should be inclusive and sustainable AI. So, what do you make of these different approaches, I would say? And how might that affect the global AI development because looks like people are looking at it from a different angle?
Wong: I personally am a little bit disappointed about their decisions. I mean it is not just me, I mean if you look at the news, you know many country leaders also raised their concern about the US and the UK pulling out, opted out. So, it is all about inclusion, it is all about openness, it is all about sustainability. And I said many times about inclusion is that technology is neutral and everyone in this world should have the rights to use advanced technology. Why should someone rich can use it and someone poor can't? So, inclusion is very important. And the openness part, as I mentioned again, open innovation, without the basics, open, you will never have advanced sciences. And thirdly, sustainability is actually closely related to openness as well. I mean, sustainability, we are talking about science sustainability. So, if I want science to advance, then it would never be able to do that without openness.
Chan: Right. KF, I mean we hate to guess what is behind everyone's motives behind why they are not signing the agreement.
Wong: Oh, but there he said it.
Chan: So, what will happen?
Wong: He said it, he said it. He said that over-regulated would affect the development of AI research. I think that that is nonsense. I mean, the whole thing, the agreement itself, as we just talked about, is about openness, it is about inclusion. That is where we are opening things up.
Chan: Right. KF, since you mentioned that, I mean, everybody must have their own reason to justify what they do. So, what you are saying that is obviously you don't agree with what those countries are …
Wong: Just the two countries.
Chan: Just the two countries. So, let's say if they continue along that line, what is going to happen between the cooperation between, say, the Chinese mainland and the rest of the world?
Wong; Well, everyone has the vision of globalization, has the idea of global village, not only in the commercial world, but also in the scientific world. By pulling it out, you are doing depolarization. So, there will be, you know, two zones, the USA and the UK being one zone in the AI community, and the rest, which is no good because when we talk about sciences, it is humanity. It is not just two countries or two regions. I mean, let me give you an example, an example of the Americans just recently pulled out from the World Health Organization. So, what would happen is that, say for example, one day there is a virus happening in the United States. They knew it fine, but because they were not in the WHO, therefore they could not tell the rest of the world. And what if one American went out. And then just like what happened to us in SARS, only one person went out and it affected the rest of the world, then that could happen. Similarly in AI, what would happen is that, as you mentioned there are good people, there are bad people. What we want to be open is because if we come across bad instances or bad habits or bad wrongdoings, we will publicize it, we will share with the rest of the world.
Chan: KF, last area we want to sort of touch on is … thank you for your sort of enlightenment on the rest of the world, how is this going to affect Hong Kong? Because I know that the Hong Kong government has invested over HK$100 million last year to cooperate with local universities to develop a Hong Kong version of ChatGPT. With this advent of DeepSeek, what would you recommend now?
Wong: Well, I think both are complementary. We cannot wait, and therefore, we started with ChatGPT, a Hong Kong version of it. I think it comes in the right time, we should do it. But as we all know, technology advances very rapidly, including the inception of DeepSeek. Now DeepSeek coming in, what we have to do is to study and see how DeepSeek technology can be complementary to our Hong Kong ChatGPT system because it will be to our advantage. As we all know that DeepSeek is more cost effective, especially under this climate of the budget, we have a very tight budget.
Chan: Right.
Wong: Therefore, if we can introduce a cheaper solution, it would be very helpful.
Chan: And helpful to make Hong Kong an I&T hub too, right?
Wong: Yes, of course, of course.
Chan: Right. Professor, I think that is all the time we have for today. And thank you very much for sharing your insights with us. As we have heard today, AI is evolving rapidly, bringing both incredible opportunities and significant challenges. Global cooperation is certainly crucial to shaping the future of this powerful technology. As Andrew Ng, founder and lead from the Google Brain Project said, “Artificial intelligence is the new electricity”.
Thank you for joining us and have a good evening!
