01.AI founder and CEO Lee Kai-fu is on Straight Talk this week.

A pioneering force in artificial intelligence and technology, Dr Lee was formerly president of Google China and a top executive at Apple and Microsoft. He is sharing his insights on China's AI landscape and Hong Kong's role, and the best-case scenario going forward.

Check out the full transcript of TVB’s Straight Talk host Dr Eugene Chan’s interview with Dr Lee Kai-fu:

Chan: Good evening! I'm Eugene Chan on Straight Talk, and our guest is Dr Lee Kai-fu, a pioneering force in AI and technology. Former president of Google China and a top executive at Apple and Microsoft, he is now the founder and CEO of 01.AI, a rising star in China's AI landscape. A PhD in computer science from Carnegie Mellon, he is also a best-selling author, a venture capitalist investing in the next generation of tech disruptors. Named one of “Time's 100 Most Influential People”, Dr Lee blends Silicon Valley's expertise with deep insights into China's innovation machine. We're excited to have him discuss how Hong Kong can seize the AI moment. Welcome, Dr Lee!

Lee: Thank you, Eugene.

Chan: Dr Lee, before we started the show, you said to me AI is progressing very fast. But before we go any further, for our viewers' sake, can you explain to us what artificial intelligence is and what generative AI is?

Chan: AI is a field that began in 1959 and it was the study of whether we can build intelligence into software and machines, and maybe someday surpass humans, and also a study of what makes intelligence possible, so that we can understand how humans became intelligent. But that was 1959, and in the first 50 years or so, that field really didn't produce any very usable AI engines. And later with deep learning and then with generative AI, now we've really found an amazing way using the massive computing and data that's available to turn machines into really, really intelligent, in many cases, more intelligent than us already. Generative AI is the new approach that says, let's make AI read every book in the world. We now have enough data with the internet and enough compute power with the GPUs, graphic processing units, and the more it reads, the more intelligent it gets and generative means it's able to listen to a question and just answer it by generating, and because it learned to generate by reading every book in the world. And this began about two and a half years ago, with the release of ChatGPT, which was the first popular application built on generative AI.

Chan: Right. Dr Lee, thank you for giving us a very good background on what AI is all about. And before we go any deeper, so to speak, I'm sure the viewer is going to say, “Hey, I heard Dr Lee saying that surpassing humans”, that is a term that frightens some people, because we have watched a lot of movies before …

Lee: Yes.

Chan: … saying the movie, I mean, the world is run by robots. Would that happen?

Lee: Probably not, certainly not soon. And the reason is not because AI is not intelligent, but that it's still a tool. So, when we ask a question, it gives us an answer, and it doesn't have any self-awareness or desire. So, in science fiction, while the ideas are very interesting on how AI could be used, and it inspired a lot of people, including myself, but usually the dramatization of science fiction that's required to captivate the reader's attention and get people to talk about the book somehow that AI has to become sentience, has to have feelings, and we anthropomorphize the AI, which isn't happening today with the software. Software is a tool. It's like Excel. I run it, it gives me an answer. So the way ChatGPT or DeepSeek answers my question is really currently no different from Excel computing my monthly income statement, that said it is definitely exhibiting intelligence that 10 years ago, even I would have said, “Wow, this is human”. So at some point, when you imitate humans so well, even though you don't have ill intent, it can still certainly take away people's jobs and also exhibits all aspects of humanness, including potentially someday the negative aspects. But right now, there are still tools that we control.

Chan: Right. Dr Lee, I mean, you mentioned that my mind is actually running for your information that you've just said. You know, the world, with all the geopolitics happening, we've seen, unfortunately, a lot of wars all over the world, with drone attacks, with all these satellite attacks, etc. If this AI power has fallen into the wrong hands, how can we protect ourselves from it? Because it can go, I mean, even do better than humans, yeah. I mean, with less emotion. I mean, let's, let's do something nasty. What is your take on that?

Lee: Yes, I think that is a real danger. I think while AI becoming evil is unlikely in the short to medium term, evil people using AI to do evil deeds are indeed possible. So, I think AI safety is increasingly more important. Part of it means the companies building very smart AI engines shouldn't answer questions that are harmful to other people, like how to build a biological weapon, for example, or sneak a gun through an airport. Those questions simply shouldn't be answered. But I think that only provides a first level of protection. I think there also needs to be a strong enforcement of laws, not necessarily new laws, but the laws of fraud, violence, etcetera, should be equally applied to people who use AI to commit crimes, and people should be punished equally, if not harder, if they use AI as a way to proliferate such things, such as spreading rumors, being hurtful speech, targeting people, brainwashing people all the way to, you know, financial theft to stealing people's identity, credit cards to pretend to be someone else, or to even harm people's health and lives. These should be punished under current laws, and people might even be punished more if they use AI to do more harm than if they did it one one-on-one as a human.

Chan: Right, I think … and also, before we start the show, you said AI is like the word internet 10 years ago, where people think, “Oh, why should I be bothered with the internet? Now everybody's using it”. So, how do you see the everyday person find AI going to assist them and help their life better?

Lee: Yeah, I'll just describe for myself. I don't do anything without AI now. If I want to know what's new, I go to an AI-assisted engine and ask questions about current events and get targeted answers. And whenever I search for content, I don't use Google first anymore. I ask an AI engine, but with web search turned on, I would go to DeepSeek, and with an online search turned on, I can find anything I want with more accuracy. If I want more details on specific websites or something, I would still use Google. But usually when people used to go to Google, it was with the question they wanted answered, like, “what's the fastest rising stock for the last week?” Or “what did Elon Musk say during the recent events about cutting research?” Those are things people want to know now. You don't have to go to search engines and type keywords and piece together what's the best website. You just ask the question, and you get the answer.

Chan: Right.

Lee: And then when I want to write a document, I don't write it myself anymore. I give a very detailed prompt. I have Perplexity, or Google or DeepSeek, write it for me, then I would edit it and PowerPoint the same thing, I would use a tool called Pop AI, which we produce that creates PowerPoint for me. Of course, I would edit it, but it would save a lot of time. These are just some examples, because if you break down activities of everything you do, they are going to be all transformed by AI. Just like 30 years ago, everything was transformed by PC. Fifteen years ago, by phone, mobile phone and mobile internet, and now by AI, every app will be different. You have to jump in and use it now.

Chan: Thank you for telling me I'm going to start doing maybe my script next time on AI.

Lee: Yes.

Chan: In February, we had Professor Wang Kam-fai here, and we talked about DeepSeek when it first came out, and actually only in months the advancement actually has even further stunned the tech world with even more progress, like bigger models, multi-model capabilities and even global interest. How do you assess the difference, or the significance between, like the Chinese-based DeepSeek versus, say, the Americans sort of use, ChatGPT. How do they rate, I mean, ranking one another?

Lee: The two technologies are very comparable. I would say ChatGPT usually comes out with new ideas, more breakthroughs before DeepSeek. But DeepSeek always catches up within three months or so.

Chan: I see.

Lee: And also, DeepSeek's engineering and cost is significantly better than ChatGPT. So, one is better in research, the other is better in engineering. So, roughly, it's fair to say they're equal. But what's very important is that DeepSeek is open source and free, so anybody can use it, and every company can put DeepSeek on their internal server and use it internally, without the data leaking to other companies. So, I think DeepSeek's open-source approach and the free approach is really changing the way people, especially businesses, think about how they use generative AI. Because now there is a free version of ChatGPT that doesn't require sending your data to the internet, and your data is valuable to you, it may be private information, and it's safer to run inside your company.

Chan: Right. Okay, Dr Lee, let's take a short break now, and viewers, we will be right back.

Chan: Welcome back! We have been exploring the world of artificial intelligence with the founder and CEO of 01.AI, Dr Lee Kai-fu. So, Dr Lee, thank you very much for giving us a very good sound background, so that everybody will start looking into AI because you said to me that you think Hong Kong people should be more embracing of AI as compared to the mainland and the United States. When you were saying that, I mean, I know that AI adoption is becoming sort of a global thing, but the development is concentrated in a few regions. How do you say or where does Hong Kong fit in? In other words, where are we? And also you know we have some good universities, and also with fintech at the adoption, do we have the infrastructure to move from being an AI user to say, an AI creator? What do you think?

Lee: I think the potential is tremendous, but we are currently at a deficit. And perhaps the reason is that when ChatGPT took off that really got Americans and English users excited. Normally, that would have spilled to Hong Kong because of the English-speaking, but ChatGPT blocked Hong Kong from using it. So, it didn't quite share that excitement in December 2022. Then I think China developed a lot of engines, and DeepSeek really rose up to be an amazing product just starting this January. I think that it got some buzz in Hong Kong, but I have to say it is much less than the mainland. I am not exactly sure why, maybe it needs to be broadcast more, covered more, maybe it doesn't do as well in traditional Chinese versus simplified Chinese, I am not sure the reason. But doesn’t matter what the reason is because when I go to the street, talk to my friends, talk to people, I don't feel people are using DeepSeek or other large language models to the same extent that people in the US and Chinese mainland are. And that is a big problem because if you imagine a country or region that didn't embrace the internet or embrace it 2 years or 3 years later than other countries and regions, that would cost the companies to not be able to reap the benefits of saving money, making money, increasing productivity. And I agree with you that financial, it’s clearly the best sector to apply, and we actually have a number of investments that use generative AI for trading. And this is clearly the best area because when you want to apply AI, the best field is one that is purely quantitative; you don't need robots or factories, it is just numbers. And financial applications are just numbers. And also investment is an area you can measure your returns. And in asset allocation or in compliance, or in discovering fraud, money laundering, these are all tasks perfectly suited for generative AI. And I also worry that the banks in Hong Kong are not embracing this as much as the American banks and the mainland Banks. And I think it is imperative because falling behind has serious consequences. And also Hong Kong has every reason to be ahead, not behind, because of the great AI minds in the universities and the financial market that Hong Kong currently has.

Chan: Right, Dr Lee, thank you for elaborating on that point. I mean Hong Kong is sort of tasked by the central government that we have to be technologically advanced and be aware of all the development. So, with AI developing, I just asked earlier, do we have enough talents in Hong Kong to even reinvent the wheel? Because some argue that Hong Kong's economy is a small one, you can just import it. But however, would you be concerned if you rely too much on so-called foreign technology if you use ChatGPT, rather than DeepSeek? What is your take on that? Well, I think we should all used DeepSeek, and not ChatGPT.

Lee: Right, okay, good. Because ChatGPT is not made available, and if we go out of our way to use it, we don't know what the consequences are. Also these generative AI are trained on the data in the country which they are collected. So, ChatGPT has an American bias. And surely DeepSeek has a Chinese bias, but if we have to choose one, I think we are part of China, choosing DeepSeek is the right answer. And I think people should use it. It doesn’t mean that you have to use the DeepSeek app, there are many apps that use DeepSeek.

Chan: All right.

Lee: For example, Tencent Yuanbao uses DeepSeek. And our PopAi uses DeepSeek. So, there are many, many tools with it. So, I think it begins with everybody using it and figuring out where it is most useful in applying it as a first step.

Chan: Right. You know with all the recent geopolitics around, earlier you mentioned to me that actually the decoupling between the US and mainland’s development has been happening since 2022. So, with all these restriction on chips to China, is it going to stop our development with the Chinese AI?

Lee: No, I think Chinese AI stands at a great, great point now because we have a foundational technology. We basically have a national model, so we are focusing development and support for DeepSeek, whereas in the US, there are still many models fighting it out. Also we are at the stage where applications can be built, which is what I am taking 01.AI, it is building applications on DeepSeek. We have the foundational layer, it is caught up, and Chinese companies are better at building applications than American companies. If you look at TikTok versus Instagram, WeChat versus WhatsApp, Meituan versus Groupon. Chinese companies and people and entrepreneurs are just better at building apps. And we are at the point where building apps are most important. And I think we will be and should be ahead, with better, more user focused, and more populous on applications because of the foundational technology being there and the Chinese people are ready to jump on the bandwagon, they are very eager to use new applications.

Chan: Right. Dr Lee, we know that you are one of our Chief Executive’s policy advisers. And if we look ahead, if AI is the new electricity, so to say, what would be one policy that the Hong Kong Special Administrative Region government must implement to avoid being left behind?

Lee: I think attracting entrepreneurship around this new technology is most important. If you look at 15 years ago when mobile internet rose up, what were the greatest policies of the mainland that took advantage of that? It is about letting venture capital’s investment invest into entrepreneurs who disrupted the business, right? Back then things were being done differently, the companies that really succeeded are the Didi, Meituan, ByteDance, which did things differently than the old PC-based software companies. So, now there are going to be whole new crop of AI-based entrepreneurs. And ensuring that Hong Kong has a role in making sure there is a good environment, good funding for them, and then their applications don't have to be confined to Hong Kong. For B2C applications, they could build it for the mainland, they could build it for the Western market, so that is the advantage of Hong Kong. And then for the adoption of generative AI in financial markets, that could be focused on Hong Kong's financial markets, which has an impact in the whole world.

Chan: Right.

Lee: In this generative AI world, every app will be rewritten, will be done differently, will be built on AI first mechanism, in such a way that they are not recognizable anymore. So, the newcomers have a big advantage over the incumbent. And I think making sure Hong Kong is attracting and incentivizing and helping the newcomers succeed, is the best way to grab onto this generative AI market.

Chan: Right. Are we doing that right now, do you think?

Lee: I think we could do more, I think we could do more. There are standard policies, but I think this is … I know there are very good programs for attracting people and getting them funded and having incubators, but this is not a one-size-fits-all. I think there needs to be special treatment for generative AI companies, there needs to be more incentives, there needs to be compute resources provided, there needs to be provision of things that people need. It is not just going to be office space, but it is going to be GPUs, it is going to be how to have access to hiring front-end engineers and program managers. It is also about teaching people about all the new ways of building apps. Today, if a startup is going to go into generative AI, and they have humans writing all the codes, that is already wrong, right? You need to use the technology that you believe in, 80 percent of your code should be written by AI.

Chan: I see.

Lee: So, people need to know that when they want to go into this area to start a company.

Chan: But does Hong Kong have enough talents to support that, do you think?

Lee: Well, with the combination of the people coming out of universities, we can convince many of them to stay, and also, there is a large influx of entrepreneurial and engineering talent from the mainland. But they are in a new environment, so they have to get acclimated and get some assistance. But I think between these two groups of people, I think this is really enough to get started.

Chan: Right, I am going to ask you the last question. Imagine Hong Kong in 2030, what is your best-case scenario if we seize the AI opportunity wisely?

Lee: The best case scenario is that Hong Kong can be the place where a lot of Chinese AI companies and engineers and talents can build applications for the whole world. And it is recognized as one of the top entrepreneurial cities for AI, just as today, Beijing, Shenzhen, and Hangzhou are. If Hong Kong could be that fourth city, I think that would be tremendous. Secondly, I think if the financial institutions will really start to embrace this, leveraging Hong Kong’s financial center and strength, and the presence of companies from China and non-China, and unifying a financial model in areas that can have a big impact.

Chan: Okay.

Lee: That would be the second thing.

Chan: Right. Thank you, Dr Lee, for those fascinating insights for us. Today, we have seen that Hong Kong isn't just an AI spectator. With its deep expertise in finance, logistics, and even legal services, our city has a chance to pioneer AI application that transforms industries. The future isn't just about adopting AI, it is about shaping it. Until next time, I am Eugene Chan, and thank you for watching Straight Talk.