Pod Insights Decoder Recode Decode: 'AI Superpowers' author Kai-Fu Lee About this episode Kai-Fu Lee, the CEO of Sinovation Ventures and former president of Google China, talks with Recode's Kara Swisher about his new book, "AI Superpowers: China, Silicon Valley, and the New World Order." In this episode:00:55 - Lee's background & Google China03:32 - Why he left Google05:41 - Why American companies struggled to compete in China09:46 - It's not all because of the government12:23 - Investing in artificial intelligence18:42 - What "AI Superpowers" means21:15 - Data and privacy in China vs. the West25:17 - Where AI is going next30:17 - How Lee thinks about American tech companies33:09 - The impact of AI on jobs40:10 - The political implications of those job changes43:30 - The responsibilities of tech creators and investors
Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript 0:00 Speaker 00 Companies shouldn't have to choose between speed and control. Amex Corporate connects cards with flexible spending controls and real time visibility to help teams adapt faster. Built to simplify, designed to scale. This is Amex Corporate. Get started at go.amex slash Amex Corporate. Terms apply.
0:24 Speaker 01 Your heart can tell you a lot about your health. Apple Watch Series 12 measures your heart rate every five seconds with the most accurate heart rate sensing and wearable. So your vitals app now with heart rate variability can tell you when something is off. And your readiness score can let you know when to rest and when to push. Hear the story in every heartbeat with Apple Watch Series 12. The features described are for wellness purposes only and not for medical use. iPhone 11 or later required. Based on Apple conducted study of heart rate accuracy August 2026. Visit apple.com slash Apple Watch Series 12.
0:55 Speaker 02 Hi, I'm Kara Swisher, editor at large of Recode. You may know me as a super advanced AI that is displacing the jobs of 20 other tech journalists in San Francisco, but in my spare time, I talk tech and you're listening to Recode Decode from the Vox Media Podcast Network. Today in the red chair is someone I've known for a long time, Caifu Lee, the CEO of Synovation Ventures and former president of Google China. He's also the author of a very important new book about artificial intelligence called AI Superpowers: China, Silicon Valley, and the New World Order. Caifu, welcome to Recode Decode.
1:25 Speaker 03 Thank you, Kara, or AI, should I say?
1:29 Speaker 02 So there's lots of things I want to talk about with you, but let's get people up to speed on who you are. I've known you for a long time and I remember when you were hired, I think I wrote a story in the Journal at the time that you were hired. So talk a little bit about your background so people can get a sense of where you've come from.
1:42 Speaker 03 Sure. I grew up in the US, Columbia, Carnegie Mellon PhD. And then I ran multimedia at Apple, followed by SGI, Microsoft, where I started Microsoft Research in Asia.
1:55 Speaker 02 Right.
1:56 Speaker 03 Back at headquarters, worked for five years in Redmond. And then I went to start Google China in 2005.
2:03 Speaker 02 Right. So talk about how you got there. So because you had obviously a very storied career, you had a lot of great spots. Multimedia at Apple was a critical job. That was back in the 90s or?
2:12 Speaker 03 In the 90s, 90 to 96.
2:14 Speaker 02 Yeah. Which was their recovery period, really?
2:17 Speaker 03 A difficult period.
2:18 Speaker 02 Difficult, yeah. Kind of stuff. So you had a long time Silicon Valley. Why did you do China for Google? You went over there in what year?
2:26 Speaker 03 In 2005.
2:27 Speaker 02 2005, which is early in in Google's when they were involved in China.
2:31 Speaker 03 It was the initial entrance.
2:33 Speaker 02 Right.
2:34 Speaker 03 But I was at Microsoft Research China in 98.
2:38 Speaker 02 Right.
2:38 Speaker 03 So that gave me the experience.
2:40 Speaker 02 Sure.
2:40 Speaker 03 And that's presumably why Google tapped me.
2:42 Speaker 02 And what was your goal there to do? What was at the time when they were entering and then they exited? But talk about your goal. What was the goal for Google there?
2:49 Speaker 03 Well, the goal was to build up a local presence, win as much market share as we could, and stay true to the corporate values.
2:57 Speaker 02 Right.
2:57 Speaker 03 And all three of which we accomplished.
2:59 Speaker 02 Right, and you located servers outside of China. There were all kinds of different things you did. Can you talk a little about that? Because it's going to be relevant to what we're talking about later.
3:06 Speaker 03 Sure. Well, the Chinese laws required some servers to be present in China, so we had that. The majority of the servers were outside. And then there were certain commitments that Google made in order to do censorship and be present in China.
3:22 Speaker 02 Right.
3:23 Speaker 03 Things like providing an explanation whenever something was removed.
3:26 Speaker 02 Right, at the bottom.
3:27 Speaker 03 At the bottom, and also providing an uncensored search, and also not store personal information in China.
3:36 Speaker 02 Right, which they didn't. They didn't allow people to register, essentially.
3:39 Speaker 03 That's right.
3:40 Speaker 02 Correct. And one of the interesting parts was putting that, saying this was if someone was doing a search, at the bottom saying, things were left out of the search due to laws of China, correct? Something like that.
3:50 Speaker 03 Yes, and actually, all search engines ended up doing that.
3:53 Speaker 02 Right, right. And it was because the idea was that you didn't pretend that you weren't censoring things, correct?
3:58 Speaker 03 That was the idea.
3:59 Speaker 02 That was the idea. Now, was it a Chinese government idea or a Google idea?
4:02 Speaker 03 Both were OK with it.
4:03 Speaker 02 Both were OK with it. Now, so you were there until how long?
4:07 Speaker 03 For four years.
4:08 Speaker 02 Four years.
4:09 Speaker 03 Hmm.
4:09 Speaker 02 That, were you there during the pullout or?
4:11 Speaker 03 No, I left three months before that.
4:13 Speaker 02 Right.
4:13 Speaker 03 And I had no idea that was happening.
4:15 Speaker 02 Oh, you didn't, really?
4:15 Speaker 03 No.
4:16 Speaker 02 No, you weren't aware of that?
4:17 Speaker 03 Well, you saw the later reports that things, they saw things in November and decided to leave in December.
4:24 Speaker 02 Right, there was meddling by the Chinese government.
4:26 Speaker 03 That was the allegation.
4:28 Speaker 02 Allegation, OK.
4:28 Speaker 03 Allegation, yeah. And I left earlier in September.
4:31 Speaker 02 Hmm, hmm. And why did you do that?
4:34 Speaker 03 Well, I saw that the entire entrepreneurial landscape was just burgeoning.
4:39 Speaker 02 Right, in China.
4:40 Speaker 03 And I lost all of my young, super smart staff. They were all into startups, all doing very well. VC industry was starting, the Chinese internet market as an independent market was taking off.
4:53 Speaker 02 It really was right then.
4:53 Speaker 03 So it was an exciting thing, I thought I wanted to be a part of that.
4:56 Speaker 02 So you wanted to escape too, at the same time?
4:58 Speaker 03 Well, I didn't quite use the word escape.
5:00 Speaker 02 Yeah, but you, but you, Google, it was a big job, running Google China.
5:03 Speaker 03 It was a tremendous opportunity, but also frustrating at times.
5:08 Speaker 02 Right.
5:08 Speaker 03 But I compare that with having the freedom to invest in companies and help young entrepreneurs.
5:14 Speaker 02 Right.
5:14 Speaker 03 That seemed more fun.
5:15 Speaker 02 So let's set the table then, because before that, China wasn't seen as the entrepreneurial engine that it is today, correct? Or it was just start, that was right around when it was really becoming clear.
5:25 Speaker 03 I think that's reasonable to say. Actually, China's entrepreneurial energy started in the late 90s with the portals.
5:32 Speaker 02 Right.
5:33 Speaker 03 And then later the search engine Alibaba were launched in the early 2000s.
5:37 Speaker 02 Initially, they were thought of as copies, correct? That was the, that was the generalized feeling, that they copied US innovations.
5:44 Speaker 03 And that's not an inaccurate statement.
5:46 Speaker 02 Yeah, at that time.
5:47 Speaker 03 At that time.
5:48 Speaker 02 And so there was an Alibaba that was like an Amazon, there was a, there was everything that was in the, there was a Baidu that was like a Google.
5:55 Speaker 03 Yeah.
5:55 Speaker 02 And various things like that.
5:56 Speaker 03 Yeah.
5:57 Speaker 02 And many people had felt that that was the way it was going to be for China, that we're going to be fast followers, essentially.
6:02 Speaker 03 I think that's the assumption, because in Silicon Valley, you know, copying is frowned upon, and it's viewed that once you copy, you always copy.
6:10 Speaker 02 Right.
6:11 Speaker 03 But I think those turned out to be wrong assumptions.
6:14 Speaker 02 Yeah, absolutely. And then at the same time, Silicon Valley companies were having troubles and were operating in China. Can you talk about why that was? eBay had a disastrous run. Yahoo was only successful because it bought a Chinese company, or it had a stake in a Chinese company.
6:29 Speaker 03 I think the core reasons are numerous. First, the US headquarters thought of China as just another market. So just take the product, it should work. It worked in Europe and Japan, should work in China. But China was substantially different. Secondly, some of the companies wanted to make money too early and too soon. And I think, thirdly, the heads of these organizations, generally speaking.
6:54 Speaker 02 They were placed in China.
6:55 Speaker 03 Obviously, there were exceptions, but generally speaking, were no match for the local entrepreneurs.
6:59 Speaker 02 Right.
7:00 Speaker 03 The entrepreneurs, you know, they own 80% of the company. This was their one thing in their life that was going to make or break their whole career and future. They worked 12 hours a day, seven days a week.
7:11 Speaker 02 Right.
7:12 Speaker 03 They did whatever it took to win. And then the multinationals had a normal, professional lifestyle.
7:18 Speaker 02 We're going to come in, we're going to.
7:20 Speaker 03 Yeah.
7:20 Speaker 02 Business development.
7:21 Speaker 03 Do business, do it the standard way, do it the corporate way. Don't offend the headquarters, do what headquarters wanted. And don't contradict the headquarters and hopefully get a promotion back to the headquarters in three years.
7:33 Speaker 02 Right.
7:34 Speaker 03 And that mentality just had no chance of success.
7:37 Speaker 02 So talk about why the Chinese market was different, because they did treat it. They had been, had success going into Europe or Germany or wherever in those ways. What was the difference of the market?
7:46 Speaker 03 Well, there's a difference then and difference now. The difference now is even more dramatic. It's almost like a parallel universe. So all the practices and assumptions you have in the US will fail. For example, if you're an app, you would expect to promote using Facebook, Snap, and so on in the US. But in China, none of those worked. The US is very well segmented companies with Google, Amazon, Facebook, even each having clearly what they did.
8:15 Speaker 02 Their land.
8:16 Speaker 03 Yeah, their lay of the land, their piece. In China, everybody was competing with everyone. No one had any market for sure. And you had to know the dynamics of what was happening and make the right bets. For example, Alibaba had the entire payment that would seem like a phenomenal choke point. But all of a sudden, in. One year, Tencent took almost half of that away from them.
8:41 Speaker 02 Right.
8:42 Speaker 03 So it's what I call in my book gladiatorial kind of competition. So if you want to play in that game, first you have to be a gladiator.
8:52 Speaker 02 Right.
8:52 Speaker 03 Then you have to know how to work with the other gladiators.
8:55 Speaker 02 Right.
8:55 Speaker 03 And read the tea leaves on who's going to win.
8:57 Speaker 02 Right.
8:58 Speaker 03 Things change so much.
8:59 Speaker 02 You can't treat it like it's a US like a US market. It's like its own it's its own market rather than a subsidiary market.
9:06 Speaker 03 That's right. And also another huge difference is the Chinese companies go heavy. Right, the American companies like lightweight tech platforms. Chinese companies are willing to hire 600,000 people and to lower the costs of something. For example, you know, compared with Yelp, OpenTable, all very light platforms. Meituan in China, they brought in 600,000 people to ensure the delivery of a takeout order is goes down to something like 70 cents per delivery. And that completely changed the way Chinese people eat.
9:41 Speaker 02 Right.
9:42 Speaker 03 So that led to a very different model than what OpenTable and Yelp did. Those companies left the restaurant industry alone. Meituan basically disrupted the restaurant offline industry.
9:54 Speaker 02 Right, right.
9:55 Speaker 03 So how does an American company learn to play in that kind of tough, tenacious, disrupting market that left nothing alone?
10:04 Speaker 02 And go to any length, in other words.
10:06 Speaker 03 And competition can be very tough dealing with, you know, challenges in the press and also users who are unhappy and false rumors being spread and those are all part of doing business in China.
10:20 Speaker 02 And what about the government?
10:22 Speaker 03 Well, the
10:22 Speaker 02 That's what they always point to, the government's not going to let us succeed here.
10:25 Speaker 03 Yeah, well, the government actually plays, I think, at this point, a very modest part of difficulty of American companies going in. Now, obviously, some companies need to get a license.
10:37 Speaker 02 Right.
10:38 Speaker 03 But as we can see, Google's now got a Waymo license, it seems. Facebook is trying, maybe, yeah. And I think, you know, it's not impossible to get a license. But my question is, even if you get a license, can an American company really learn to thrive in that environment?
10:56 Speaker 02 In that environment.
10:57 Speaker 03 Now, when I went as Google, the environment wasn't that tough and tenacious. And I was I would say a little different from the typical multinational leader. I disagreed with headquarters at times and made decisions that I thought was good for the company. And we had our arguments and then we had some success. We gained market share from nine percent to twenty-four percent and on the way to become a billion-dollar subsidiary. So the numbers were going in the right direction and I thought I was going to be the only one who may have a chance to have at least a significant minority share. And then after Google, Uber, I think, had a chance.
11:39 Speaker 03 I think Travis and his team were tenacious also and that fit the Chinese spirit. But ultimately, dealing with all the local issues, they ended up still losing to DDD.
11:51 Speaker 02 Right, right. When I say the government is that the government, one of the arguments that Silicon Valley companies make is that the government advantages Chinese companies. How would you answer that?
12:00 Speaker 03 I think that's very minimal because there are obviously licenses they can grant or not grant. Beyond that, China's part of WTO. I think at this point is also, I don't really see anything that they have done to, in the recent ten years, that would show this. I think, I guess you could argue that American companies weren't going to succeed anyway, so they don't have to do anything. But in any case, I think the main issue today
12:34 Speaker 02 Was how they compete.
12:35 Speaker 03 Is just that it's too hard.
12:37 Speaker 02 It's too hard.
12:38 Speaker 03 And I would also say that Chinese company coming to America would be equally hard.
12:43 Speaker 02 Right.
12:44 Speaker 03 It's just that the two ecosystems are so different, they're bound to continue to live in their independent, parallel universes.
12:52 Speaker 02 All right, let's get to AI in your book. Talk about the premise of your book, what you were trying to do. Since then, you've been investing, give some examples of what you've been investing in at your venture company.
13:01 Speaker 03 Okay, well, we manage a total of about $2 billion and AI is our largest portfolio, about a third or so is in AI.
13:10 Speaker 02 When did you start doing that investing?
13:12 Speaker 03 Four years ago.
13:13 Speaker 02 Four years ago.
13:14 Speaker 03 I think most of China caught the AI fever about two years ago. So we were ahead because we saw deep learning was going to start making headways. So we have about forty-five investments in AI. And we have five unicorns that are totally valued at about $23 billion.
13:33 Speaker 02 So what were you looking for? Why were you ahead of the curve on that?
13:37 Speaker 03 Well, we saw, for example, deep learning was going to make a big difference. We were very big on deep learning, computer vision, very early. And then we were among the first to go into autonomous vehicles. We saw that AI for finance was going to be a big segment. And then hardware and semiconductors was going to be an important area for China. So those were our fundamental bets and the makeup of the five unicorns that we have.
14:06 Speaker 02 Well, explain why, though. Give me a deeper question. Why did you think that was the bet to make?
14:11 Speaker 03 OK, so deep learning was the single biggest breakthrough in AI that made a machine learning possible on huge amounts of data with minimal human intervention. And it didn't need humans to tell features. It would discover them as long as there was enough data. And China had so much data. So there's going to be, somewhere it's going to tip. And we also saw that it's the people in computer vision that invented deep learning. So it's likely that computer vision would be the first area to tip, not speech recognition or something else. So we made big bets in that area. The semiconductors, well, we saw the Nvidia pricing.
14:53 Speaker 03 And we knew that the Chinese companies would want alternatives. And there are ways.
14:59 Speaker 02 High prices, you mean?
14:59 Speaker 03 Well, they sell the same product for very high prices for display versus AI. And that high level of margin, I think, leaves room for local competition.
15:11 Speaker 02 To compete with Nvidia?
15:13 Speaker 03 Right, not, I think it's hard to compete completely because Nvidia is a powerful company. But if you take one segment of the compute, let's say the inference, not the training, or make it cheap in cell phones, devices, toys, where China is strong. So that were, those were our investments in semiconductors, in AI acceleration. Autonomous vehicles, I think, was an area, there was a large number of people who.
15:42 Speaker 03 Decided to bet in the space. And we found a couple of really, really good teams. And we actually made four investments in autonomous vehicles, not counting the sensors. And I continue to think that will be the largest disruption. It might take a little bit of time.
15:57 Speaker 02 Right.
15:58 Speaker 03 So we made four very good investments, one of which has become a unicorn. And the speed at which progress is made in that space is phenomenal. I think two years ago, three of the four companies started. And one could say they were eight years behind Google.
16:14 Speaker 02 Right.
16:14 Speaker 03 Today, I think they're about two years behind Google.
16:16 Speaker 02 Yeah, absolutely.
16:17 Speaker 03 So they're catching up very fast. The next two years may be harder, but they've caught, in two years, they've caught up six years.
16:24 Speaker 02 Right.
16:25 Speaker 03 And then finance is the lowest hanging fruit because finance is a numbers game. And if AI is an objective function that optimizes profitability, lowers costs, improves margin for loans, credit card frauds, banks, insurance companies, that seemed like a no brainer because you didn't have warehouses, manufacturing plant. You just plug in the algorithm and money comes out. You're printing money.
16:50 Speaker 02 Right.
16:50 Speaker 03 So we backed a couple of companies in that.
16:52 Speaker 02 All right. We're here talking with Caifu Li. He is an investor in China. He's the CEO of Sinovation Ventures. His new book is called AI Superpowers. And when we get back, we're going to talk more about what that means.
17:02 Speaker 04 Fall tastes better with better ingredients from Whole Foods Market. No added growth hormones or antibiotics in their meat department means slow braised short ribs are even more satisfying. All seafood is wild-caught, like their vibrant sockeye salmon, or verified responsibly farmed. And all sweets, including fall-themed cookies, are made without high-fructose corn syrup. Spice up fall at Whole Foods Market.
17:32 Speaker 05 What if we told you there was one thing you could eat to reduce your risk of death by about a quarter? Interested?
17:38 Speaker 06 And what if we added that this magical substance can dramatically lower your cholesterol, improve your mood, and keep your weight stable?
17:46 Speaker 05 No, it's not protein. It's actually fiber, the least sexy, most overlooked nutrient of all.
17:52 Speaker 06 But what is fiber? How could one nutrient do all those amazing things? And how can you get enough? Is a powder you stir into some water going to do it for you?
18:01 Speaker 05 All that, plus the strange tale of the gorilla poop-obsessed father of cornflakes in the latest episode of Gastropod. Find Gastropod and subscribe wherever you get your podcasts.
18:13 Speaker 02 We're here with Caifu Li. He is the CEO of Sinovation Ventures. He also ran China for Google for many years and worked for a number of other big tech companies in the US. His new book is called AI Superpowers. And the bottom part is China, Silicon Valley, and a new world order. Talk about what you mean by superpowers. You've done investments. You talked about your investments in the area, and you did it early and often. Talk about what you mean by superpowers and why that's important.
18:41 Speaker 03 Yeah, I actually meant three things. Primarily, I meant China and US will be by far the world's AI superpowers.
18:49 Speaker 02 Right.
18:50 Speaker 03 Because they'll possess the greatest value companies, people, data, and IP. I also mean the companies such as Google, Facebook, Amazon, Alibaba, Tencent will be superpowers because they started early and they benefit from the virtuous cycle of AI. And the third meaning is that AI itself is a superpower in that it will create a wealth-generation engine that we've never seen, but also potentially a job displacement engine that we have to deal with.
19:25 Speaker 02 Yes, what we're talking about, yeah. So talk about the first one. One of the things that a lot of people, I've been talking to a lot of people about is this idea of China moving forward rather quickly in AI, well past the US. A lot of having to do with data. Someone was asking me why Google's going to China, or I asked someone that, and they said, "Data." That's it. Data is the problem. They don't have enough. China has more. And with some of the, we'll talk about the surveillance and facial recognition things, where they just have data coming out of their eyeballs in China, essentially comparatively. So talk about that competition, because a lot of people feel the US is going to Decline in that area pretty quickly because of the lack of data.
20:01 Speaker 03 Yeah, so I clearly believe in the power of data because deep learning simply works with more data.
20:08 Speaker 02 Yes.
20:08 Speaker 03 You take any three or four variants of the algorithm, one thing for sure is you pump a lot more data at it, it works better.
20:14 Speaker 02 Right.
20:15 Speaker 03 That's the primary reason that speech recognition, vision and others have improved a lot. And China has so much more data, not just in terms of number of people in market. For example, China's fully connected with mobile payment.
20:29 Speaker 02 Right.
20:29 Speaker 03 So 700 million people, most of the Chinese population, can pay each other with two buttons on the phone with almost no commission and as little as 15 cents. And that level of universality of transaction will create so much data that can be used by Tencent and Alibaba.
20:49 Speaker 02 For more insights.
20:50 Speaker 03 For mining insights, targeting and so on.
20:53 Speaker 02 Right.
20:53 Speaker 03 But can also be used by individual merchants or apps who have transactions. If you had a retail store, before you had faceless people who bought stuff.
21:04 Speaker 02 Right.
21:04 Speaker 03 Now you know who bought what and you can suddenly do much better inventory prediction, sales forecast and so on. So that's just one example, but if you move forward to, you know, China has 10 times more takeouts than the US for food delivery, it has 300 times more in shared bicycle rides. It has, I think, four times more in shared car rides. And all these standard numbers are larger, more than the ratio of the internet population. Simply because the usage of mobile was stronger, more sticky in China.
21:42 Speaker 03 But it's also going offline because a number of health clinics and autonomous stores, autonomous fast food, shopping malls, and of course airports, train stations, are hooked up with all kinds of sensors. And the sensors might track motion, heat, cameras, microphones, and these will send up data.
22:07 Speaker 02 Lots of data.
22:08 Speaker 03 Yeah, they may, yeah, they will send up data, not raw data, but data that's relevant, right? Such as a user picked up this product and. Frowned and didn't buy it.
22:18 Speaker 02 Well, people have, there's more nefarious uses of that data, obviously. But one of the things that they were talking about is that in this country, they wouldn't. The allowance of sensors and facial recognition is going to see a much rougher road because of consent, all kinds of things, where that's not the case in China. And so they can suck up so much data about people's movements, their faces, their activities that they couldn't do here, that companies like Google and others are hindered. Most people feel it's a good thing that they're hindered. Talk about that.
22:48 Speaker 03 Well, I think the Chinese users have a stronger willingness to exchange the capture of certain data if, in return, there is value to be provided. For example, greater security, lower crime rates, or possibly convenience. I do think the Chinese people care about privacy. There are people raising awareness. But I think at the current level of deployment, people accept the trade-off that is being offered.
23:21 Speaker 02 And I think the worries about the government observing people is well-founded, correct or not?
23:28 Speaker 03 I think that's a very popular feeling in the West, yes.
23:34 Speaker 02 And which would hinder the West from collecting data, in a good way, if people in the West think that.
23:41 Speaker 03 Well, I think the people in the West would generally not adopt the widespread sensors used by government. Therefore, also, the private companies would have a hard time putting it in stores.
23:56 Speaker 02 Right.
23:57 Speaker 03 I don't know how Amazon Go is perceived, because they have those cameras.
24:01 Speaker 02 Yes, they do. Not, people are wary of it. People are wary of Nest. People are wary of all the different. Products that they are trying to insert in the home and everywhere else. And there's always some complaint about them.
24:13 Speaker 03 Yeah, well, I think there's also a lot of good that could come from this deployment.
24:17 Speaker 02 All right, talk about that.
24:18 Speaker 03 Well, in hospitals, you can prevent sick people, elderly people from falling down. You can call alarms for them. Crime rate, we've talked about. Autonomous stores, you can essentially turn an offline store into an online store by capturing user preferences. And at schools, I think parents in China are likely to give consent for cameras at school, not for surveillance, but for giving teachers feedback on where how to improve the kid's performance, where they might be getting lost. So hospitals, clinics, and elderly homes and so on.
24:58 Speaker 03 So there are applications I think the West would find useful as well. But I think those would be would have a hard time getting launched here.
25:08 Speaker 02 Right, and the mentality is definitely different, which I think it depends on what you think about it. So talk a little bit about where AI is going. And so here you have all these major investments. You've made billions of dollars in investments in this area, as have many others. Where are the trends going in AI?
25:26 Speaker 03 Well, we see four waves of AI, all really just at the very beginnings. One is using AI in internet. The other is using it in businesses. So for business intelligence, banks, insurance companies, and basically all corporations to automate the middle person out of the system. The third level, third wave, we think is adding eyes and ears, what we were just talking about, the sensors. And we think that will create a lot of new applications that didn't exist before.
25:55 Speaker 02 Right.
25:56 Speaker 03 Because previously, those data, visual data, were discarded. They became transient. But now they could be captured and something could be done, you know, smart cities and so on. And then lastly, we see autonomous AI. And that is when AI gets arms and legs. They may not, they may be wheels, but they can move around, manipulate in factories, manufacturing, in farms for picking fruits, in commercial applications like washing, washing dishes. And also eventually in the home for education, toys. And, and also eventually there will be housework robots.
26:33 Speaker 02 Right.
26:34 Speaker 03 All of these will happen along with autonomous vehicles that will begin in non-public roads, then going into highways, and then going into all the roads, and then going from L three to L four. And that will lead to another wave of changes. So we see great investment opportunities in all four waves. And no doubt there will be a fifth, sixth, seventh wave, which we just don't know what they are.
26:58 Speaker 02 What are you thinking that might be?
27:00 Speaker 03 Well, I think a delegation interface with a smart assistant that has an infinite memory to enhance us, you know, without any of this hardware intrusion, but just as an augmentation of our physical limitation.
27:15 Speaker 02 Okay.
27:15 Speaker 03 That could be.
27:16 Speaker 02 That means not something, remember Google had the eyeball thing and the ear thing? Not that, that it would be part of your.
27:24 Speaker 03 Have you seen the Black Mirror?
27:25 Speaker 02 Yes, of course, yeah. Yeah, oh, yeah, yeah, the eyes being.
27:28 Speaker 03 Yeah, that one, that it remembers everything you saw and heard.
27:30 Speaker 02 Yeah, through your eye.
27:31 Speaker 03 And can index it.
27:32 Speaker 02 Right.
27:32 Speaker 03 Obviously, that has also dystopian and worrisome outcomes.
27:36 Speaker 02 The guy took it out at the end.
27:38 Speaker 03 Yes, but, but it may not be that instantiation. But it's a, I mean, we people are limited in our ability to remember. That's our faultiest part of our brain.
27:48 Speaker 02 Right.
27:49 Speaker 03 And computers are perfect.
27:50 Speaker 02 I was just saying this to someone the other day, something I wanted to, I wish I could have remembered this.
27:53 Speaker 03 Yeah.
27:54 Speaker 02 My son was asking me about it, something, and I was like, I just don't remember it.
27:57 Speaker 03 Happens all the time.
27:58 Speaker 02 I have a vague memory of it, or, or if at all, if it's the correct memory or if it's not the correct memory. It was really, it was an interesting discussion I ended up having with my son.
28:07 Speaker 03 Right.
28:08 Speaker 02 And beyond that.
28:09 Speaker 03 Well, beyond that, whether we believe in AGI, general AI, or not, I'm in the camp that feels that's very distant, many decades away.
28:19 Speaker 02 Explain that for who?
28:20 Speaker 03 Well, AGI means having AI be like humans, have the common sense, cross-domain ability to reason and plan, and then one step further, maybe even with self-awareness and emotions.
28:33 Speaker 02 Cyborgs.
28:35 Speaker 03 Cyborgs, right. So, some people, like Ray Kurzweil, project that's coming in 2029.
28:42 Speaker 02 It's not.
28:42 Speaker 03 I think not. I think the number of breakthroughs that will be needed would be probably another dozen deep learning level breakthrough.
28:50 Speaker 02 Right.
28:51 Speaker 03 And if you look at the last 60 years, we've had one deep learning level breakthrough. So, when will the 12 come? Maybe 720 years.
28:59 Speaker 02 So, of the ones you're talking about, all these different AI waves, which, give me timeframes on a lot of them, on the first four.
29:06 Speaker 03 The first three have already happened. Internet AI is all around us. Business AI is being implemented, but requires a large amount of data, so it's going to be, you know, the large banks that start and so on. The visual is, I think, is starting in China, less so in the US. The spoken is happening in both China and US. Amazon Alexa is a perfect example, their Chinese equivalent of that. Autonomous vehicles, I think, are already happening in non-public roads.
29:38 Speaker 02 Sure.
29:38 Speaker 03 Shuttles, forklifts, smart robots, Kiva inside Amazon and the Chinese equivalent.
29:44 Speaker 02 Kivas. People never pay attention. I pay a lot of attention to Kiva.
29:47 Speaker 03 Sure, sure.
29:49 Speaker 02 I absolutely, when they bought that, I was like, oh, wow, that's an interesting change.
29:53 Speaker 03 And I think the natural next step is to have the, another robot pick out.
29:56 Speaker 02 100%.
29:58 Speaker 03 Yeah, 100%.
29:59 Speaker 02 I was at the Amazon warehouse in Kent, and they were using the Kiva robots, and there was a guy, the Kiva robot would bring over the stuff, and then the guy would pick it out, and I said, "You're finished." And he's like, "What?" And then I'm like, "I don't even want to explain what's about to happen to you."
30:12 Speaker 03 That's right.
30:13 Speaker 02 Before I get to the issue of work, how do you look at US companies right now, having worked for all of them? What do you think where they are?
30:21 Speaker 03 Well, I think the Silicon Valley style of entrepreneurship is still the world's leading way of building companies. It's vision-driven, focuses on tech, tends to go deeply to solve problems, and has strong culture and value that makes them are built to last. So I have a lot of respect for all the companies I work for. I also think China is emerging with a kind of a new way to build companies of running as fast as you can, and always assume someone's gonna eat your lunch, and so you better eat theirs first. And it's a very scrappy, tenacious competition.
31:02 Speaker 02 It used to be Silicon Valley, correct or not?
31:04 Speaker 03 They did.
31:05 Speaker 02 Was it? Did you not think we're too soft, right?
31:07 Speaker 03 I think there were days when Microsoft was pointed and considered that way.
31:13 Speaker 02 And then?
31:14 Speaker 03 And then Microsoft became kinder and gentler, right?
31:16 Speaker 02 Right, yeah, they had to. They were forced into it by the government, yeah. Was that a bad thing?
31:21 Speaker 03 I think it's not good or bad, it's just different. Because China is such a big market, and there's so much capital going in, and the people who've made the right investments have made so much money, and the market kept growing, at least up till now, all those things incentivize this behavior. Also, keep in mind Deng Xiaoping about 40 years ago said he will change the economic system by letting some people get rich first. So there is a rush for the Chinese in the last 40 years to be among the first, because otherwise you may not be among those.
31:55 Speaker 02 Right, right, absolutely.
31:56 Speaker 03 So I think all of those pointed is almost a system constructed with just the right elements. And that's what in the last 10 years turned it from a copycat to an innovative country.
32:10 Speaker 02 Are US tech companies now too soft? Are they too rich? Do they have too many private planes?
32:15 Speaker 03 I don't think that's the issue. Maybe that's an issue. I think the Chinese tycoons buy just as many private jets.
32:24 Speaker 02 Yeah, that's a fair point.
32:25 Speaker 03 Yeah, but let's say... We found an internet population in Mars. And we're gonna land the two top entrepreneurs in Silicon Valley and two top from China and let them compete to see who can win the Martians. I would bet on the side of the Chinese entrepreneurs.
32:41 Speaker 02 Because?
32:42 Speaker 03 Because they're faster, more tenacious, more understanding of user needs and willing to build products for user needs rather than what Steve Jobs says. You know, I look in the mirror and that's my user understanding.
32:54 Speaker 02 Right, right, that's usually, that's really interesting also.
32:56 Speaker 03 Brilliant, but it's hard.
32:57 Speaker 02 But you know, the kombucha from the Silicon Valley entrepreneurs would be better. And the various foods and things like that. I'm only teasing.
33:05 Speaker 03 Yeah, it could be.
33:05 Speaker 02 So let's talk about the impact of AI on jobs because I think this is something that American companies or tech companies are facing. This idea that what they're doing, right now, they're in the midst of, did they kill democracy? And they were just having hearings, for example, we're just listening to. But one of the things has been this discussion of the future of work.
33:23 Speaker 03 Right.
33:23 Speaker 02 So what does AI mean for that from your perspective?
33:27 Speaker 03 Well, I, I think there are a lot of simple, um, simple kind of one sentence answers. None of which I think are right.
33:35 Speaker 02 Okay.
33:35 Speaker 03 Some people said it's all purely a human amplifier. Some people said, oh, it's just going to make us better. Others said, no, it's gonna take all the jobs away. I think it'll be different depends on what jobs you have. If we look at what AI cannot do, there are really two main things. One is creative jobs, you know, jobs like yours, jobs like scientists, storytellers.
33:58 Speaker 02 I agree.
33:59 Speaker 03 You know, artists and so on. And the other are the compassionate people who really have created a human to human connection.
34:06 Speaker 02 Right.
34:06 Speaker 03 Trust.
34:07 Speaker 02 Until they get the robot eyes right, and then they will.
34:10 Speaker 03 Well, I think the robots will always mess up. And when robots mess up, they mess up badly.
34:16 Speaker 02 See, I think we should put robotic things in people and then not, instead of making robots try to be like people, we should make people try to be like robots. It's a big thought, actually. Just think about it. No, think about it. We spend all our time trying to get a robot to open a door and everyone goes, wow, it opened a door.
34:31 Speaker 03 Yeah, yeah.
34:32 Speaker 02 Why bother? Just put robotic things in people.
34:35 Speaker 03 That's right. I, my favorite example is elderly care. All these people building robots to take care of old people. We should take care of our parents.
34:44 Speaker 02 Yes, we just can't get people to do that.
34:45 Speaker 03 And if we don't, then we should hire a person to. None of our parents want to be taken care of by a robot. So we should, yeah, I, I agree with you, I get what you're saying.
34:53 Speaker 02 But I'm saying put robotic things in the caregivers. Anyway, it's a bigger idea.
34:58 Speaker 03 I see.
34:58 Speaker 02 I'm too genius for you. I'm too gen-
35:00 Speaker 03 No, I get it. I get it. So you want the human caregiver, but maybe with super strong arms.
35:05 Speaker 02 Exactly.
35:05 Speaker 03 That can prevent the fall of-
35:07 Speaker 02 Yeah, like instead of spending our time trying to get a robot to open a door.
35:09 Speaker 03 Yeah, yeah.
35:09 Speaker 02 Like, come on, why bother? We know a man, a man or a woman can open a door.
35:13 Speaker 03 Okay.
35:13 Speaker 02 So that's already solved. So, in any case.
35:15 Speaker 03 Okay.
35:16 Speaker 02 So now we give them eyes and ears and things like that. An extra lifting and ectoskeletons and things like that.
35:21 Speaker 03 I see.
35:22 Speaker 02 Yeah.
35:22 Speaker 03 I see. Right, so coming back to my thoughts. So the human connection is hard. So what will happen? So if we make a quadrant of four types of jobs. I think the lower left quadrant is the low compassion, low empathy, low creativity.
35:38 Speaker 02 Right.
35:39 Speaker 03 All those jobs will be taken by AI.
35:41 Speaker 02 Name those jobs. Those are factory jobs.
35:43 Speaker 03 Oh, beginning with, yeah, factory jobs. Starting with inspection, going into assembly. Starting with dishwashing, going into flipping hamburgers and simple cooking.
35:53 Speaker 02 There's a robot hamburger place here in San Francisco now.
35:55 Speaker 03 Right, absolutely.
35:56 Speaker 02 Called the Creator.
35:57 Speaker 03 And we, we invested in a, in a Chinese noodle robot too.
36:01 Speaker 02 Okay.
36:02 Speaker 03 And then there are also white collar jobs that will be replaced.
36:06 Speaker 02 So where sector is that? That's same in the, that's in the same sector?
36:09 Speaker 03 Yeah, basically traders, they're already gone, right? And then a city name said they're going to remove 10,000 of their operational staff. And even before that, telesales, telemarketing, customer reps. The jobs currently outsource to India. And the current manufacturing jobs are outsourced to China.
36:28 Speaker 02 Right.
36:28 Speaker 03 Both sets of those will be challenged. I would argue white collar ones first because that's software only.
36:34 Speaker 02 Right.
36:35 Speaker 03 So that's a lower left quadrant.
36:36 Speaker 02 Name those jobs. Lawyers, possible doctors.
36:39 Speaker 03 Not, not yet. No. The, the lower left corner, those are more the inspection, assembly line, telemarketing, customer service. Those are in danger because they were low human touch and low creativity.
36:50 Speaker 02 And data rich.
36:52 Speaker 03 Data rich, exactly. And routine in nature. And then the upper, upper left quadrant would be lower creativity, but lots of compassion. I think those jobs will flourish. And in fact, I think migration needs to go from lower left to, to upper left. So for example, the doctor's job will probably change because the medical, medical diagnosis will become very, very good.
37:19 Speaker 02 Right, through AI.
37:20 Speaker 03 Uh, through AI. And then the doctor is more of a human connector. And then maybe just four years of college is enough. Maybe nurse practitioners can become doctors. Maybe there's more training about how to comfort and how to tease out from the patient. What are you really feeling?
37:38 Speaker 02 Because diagnostics should be done by AI, like radiology, for example. I use, always use the radi, like you don't need a radiologist. They're not accurate compared to the AI.
37:47 Speaker 03 That's right. Well, it takes time, but eventually they'll be displaced in terms of.
37:53 Speaker 02 All of diagnostics, it seems.
37:54 Speaker 03 Eventually, take twenty, thirty years. Uh, but one segment at a time. So doctors can become this compassionate profession. And we might have ten times more doctors because the cost of medical care will go down. Poor people can access it. And then you can still have real super experts that you pay a lot of money for. But most healthcare, uh, so more doctors could be employed. But not the same kind of doctors today. The same could be applied to many other areas like professionals. You know, um, wealth planners, teachers in particular, you probably want.
38:28 Speaker 02 Talk about that.
38:29 Speaker 03 What I think a lot of what teachers do are routine. So grading homeworks, grading exams, giving us exams, giving the same lecture and then again, again. Those can be done by AI or MOOC. And what the teacher should do is one-on-one, targeted, finding out what your passion is, guiding you, coaching you, become your mentor for life. And that could be one-to-one. That could be homeschooling. That could be, um, public school, but one-to-one ratio. So teacher numbers could blossom. So I and also, I think teacher and doctors are more require a lot of training. There will be other less trained jobs. For example, elderly care. We're going to have a lot more older people. People over eighty requires five times as much care.
39:13 Speaker 03 And we want people to take care of them. And elderly care is a very difficult to fill job because it's not paid well, nor does it have a high social status. And I think those need to be changed so that when people come off the assembly line and telemarketing jobs, they could move into either an elderly care type of a job or a teacher type of a job depending on their aptitude.
39:38 Speaker 02 Right.
39:39 Speaker 03 And then the right side are the creative jobs. So that's a sigh of relief. We're OK. The creative jobs without too much compassion, empathy needed. They can use AI as a tool. Scientists can find more drugs with AI filtering for them. And their power will be amplified symbiotic combination. And then the upper right will be high empathy and high creativity. And those are what will make humans shine.
40:06 Speaker 02 What does that mean politically and for the countries, these countries, including China? Because there's a lot of, you know, rote manufacturing jobs there. There's a lot. And you've thrown people at it, which is why it's lower cost and which is why Apple and others have moved manufacturing there. What happens? Because there is a social crisis that can result in this.
40:26 Speaker 03 I think the bigger crisis is social more than financial because it's not just a question of losing a job and getting some social welfare, UBI to pay for you. It's that people have attached the meaning of their lives to their work.
40:40 Speaker 02 Right.
40:41 Speaker 03 And when the work is gone, there's so is the meaning. So I think it's imperative that governments start to understand how to redistribute the money so that there is enough money to take care of this set of people who need to make a transition.
40:58 Speaker 02 Right.
40:59 Speaker 03 And then there needs to be retraining and incentives put in place so the migration can happen.
41:05 Speaker 02 Do you think government has that capability? I don't.
41:09 Speaker 03 I think.
41:09 Speaker 02 They don't know how Facebook works.
41:11 Speaker 03 I think very few governments have that capability. I think we can start with something small. For example, rather than give everybody a tax break, give it to those who homeschool. Give it to those who are doing volunteer work. Give it to those who take early retirement, but put their time in a socially meaningful. So there can be smaller steps. Vocational training should change. We should have fewer auto mechanics courses, but maybe more plumber courses. Because plumber is not a job that robots can do soon. And we already know which jobs are going to be on the decline. So the vocational schools should follow this projection of job increase or decrease
41:56 Speaker 03 based on automation. So those are things that can be done by any government. I think potentially governments like China may be able to make bigger steps with distribution. Chinese government has historically been effective in pushing one segment to another. The Chinese agriculture to manufacturing shift was done faster with a lot of chaos, but still more effectively than probably any other country. So there might be a different sample.
42:27 Speaker 02 How do you assess the US's commitment to this? I don't think there's any. I don't even.
42:31 Speaker 03 Well, I don't think the current administration acknowledges that this is happening. I've seen some top officials say the AI job displacement is 50 years away. And that would be worrisome.
42:44 Speaker 02 That was the Treasury Secretary.
42:46 Speaker 03 That was.
42:47 Speaker 02 That was the Treasury Secretary. But he's an imbecile. So what do you do then? Does that bring, again, the US behind again? Because it'll be losing these jobs, because that's the way it's going to go, without any preparation for the future.
42:59 Speaker 03 Well, my guess is that when there will be some profession that's suddenly disrupted and millions of people are out of their jobs, and then that will wake up the government. That's my guess. I don't know, but possibly the danger is it might be one of the outsourced jobs. So that the paying is in India, not in China.
43:22 Speaker 02 Right, right, right, absolutely. So finishing up, let's talk a little bit about what that means for the creators of these technologies. Do you have a responsibility as someone who funds them to figure that part out?
43:34 Speaker 03 I think I do. And all of us try to contribute differently. There are those who advocate the UBI. I do not.
43:42 Speaker 02 Universal basic income.
43:43 Speaker 03 Yes.
43:43 Speaker 02 Why is that?
43:44 Speaker 03 I do not, because I think that doesn't solve the meaning problem. I think you're just giving people to money as an anesthesia for paying.
43:54 Speaker 02 That's a really good way to put it.
43:55 Speaker 03 Doesn't get them over the really true problem. And I think all of us are thinking of ideas, and that's good. And I think all of us are willing to contribute, whether it's by taxation, or by donation, or by foundations. And I think those of us in investment, we could look more at investment that create jobs. Still make money, but maybe not as much as the AI companies. So I think it's imperative for all of us to do what we can without expecting the government to do it all. Which is why I wrote this book, so that the awareness would be there. But the call to action is up to the individuals.
44:32 Speaker 02 Up to the individual companies. Do you think they have that commitment?
44:36 Speaker 03 I'm optimistic. I'm optimistic they do, because in Silicon Valley, I think all the talks about UBI suggest that people
44:43 Speaker 02 Are thinking of it.
44:44 Speaker 03 want to do something.
44:45 Speaker 02 Yeah.
44:45 Speaker 03 Whether we think that one will work or not, it's still very respectable that they are thinking ahead.
44:52 Speaker 02 And when you look forward 20 years, what are the jobs you think will be the most important?
45:02 Speaker 03 I think the creative jobs will be the most important. I think a lot of people who could have been creative were stifled over the last 100 years because maybe they weren't the highest paying jobs.
45:17 Speaker 02 Right.
45:17 Speaker 03 They were forced into some relatively more routine job that paid more. But I think now we can really have a chance to release our potential in creativity.
45:29 Speaker 03 And I think the empathetic jobs, human jobs, will also be important because that's the only job type that can absorb the exodus and the displacement that will happen in the routine jobs.
45:41 Speaker 02 Yeah. Yeah. Last question, I'm gonna put you on the spot. If you were to pick one US company and one Chinese company that you find, a large one or a small one, we'll do each of those that you're most impressed with right now. What would they be?
45:53 Speaker 03 Most impressed US company would be Alphabet, Google.
45:57 Speaker 02 Because?
45:58 Speaker 03 Because of its aspiration, because that it tries to stick with its values. You know, we may or may not agree with it, but it tries to do that. And also because of the phenomenal creativity from the company.
46:11 Speaker 02 And their most important part, Waymo, what, cloud?
46:14 Speaker 03 I think Waymo's most interesting to me. But I think a lot of the new healthcare initiatives are interesting as well.
46:22 Speaker 02 Okay.
46:23 Speaker 03 The Chinese company would be a tie between Alibaba and Tencent. I think Alibaba has demonstrated that a company can grow so big and still top-level people feel empowered, like they own the company. And that kind of cultural strength will probably give it legs to go to the next level. I don't see that in other companies. Alphabet obviously is trying to do that with the different segments. But Alibaba, if you really go in, each of the 25 CEOs really feels like a CEO. And they're still working like startups. So they maintain a culture and that's expandable, that doesn't fall apart with size.
47:07 Speaker 03 I also respect Tencent a great deal because they're one of the very few companies that could build a product to disrupt itself. They had QQ, which was the dominant messenger. And then they allowed WeChat to be built and to disrupt QQ. But QQ didn't die. QQ kind of became the Snapchat of China, appealing to the younger generation. So a company that can tolerate two camps to compete and both actually continue to be successful, that I think is really rare to see.
47:39 Speaker 02 What about startup that you find?
47:43 Speaker 03 There are many great startups. We fund VIP Kid, which is an aspirational education company that connects American English teachers to Chinese students. It's essentially the Uber for education. And when everyone thinks education is not going to create unicorns, VIP Kid has really proved them wrong. And now they're using technology and AI. They're also using pro bono to give the English teachers a chance to give some hours to teach poorer kids in a group. And I think it's a company that thinks far ahead and more likely to disrupt education
48:24 Speaker 03 than the, some of the American education companies that have a big aspiration, but it's too difficult to implement.
48:31 Speaker 02 Absolutely. That's really interesting. All right. This has been fascinating. Kai-Fu, thank you so much for talking.
48:36 Speaker 03 Thank you.
48:37 Speaker 02 Kai-Fu Li's book is called AI Superpowers: China, Silicon Valley, and the New World Order, and it's available now. If you enjoyed this interview as much as I did, be sure to subscribe and leave us a review on Apple Podcasts. You can also find more episodes of Recode Decode on Spotify, Google Podcasts, or wherever you listen to your podcasts. If you didn't like this interview, I, again, do not care. You can just say, if you just want to say hi, tweet at me. I'm at Kara Swisher on Twitter. Now that you're done with this, go check out our latest episode of Recode Media. You can find that show wherever you found this one. Thanks for listening to this episode of Recode Decode. Thanks to our editor Joel Robbie and our producer Eric Johnson. I'll be back here on Wednesday. Tune in then.
49:18 Speaker 07 20th Century Studios presents Whale Fall, the most incredible survival thriller of the year.
49:26 Speaker 07 In theaters Friday. To survive the unthinkable, do the impossible. You're going to become so dangerous that the predators can't help but let you go. Austin Abrams. I'm going home. Josh Brogram. Survivor, survivor. Whale Fall, rated PG-13, maybe inappropriate for children under 13. In theaters Friday. Get tickets now.
Previous Recode Decode: Nicole Wong, former deputy CTO of the United States Next Recode Decode: 'Swiped' director Nancy Jo Sales Page updated October 26, 2026 · transcript and summary are machine-generated