The U.S. has long been seen as the global leader in innovation, including in the field of artificial intelligence (AI). China, in contrast, has been viewed as a technology copycat. This, however, may not be the case anymore. China may soon take the lead in AI, according to Kai-Fu Lee, former president of Google China and an AI expert. He said China’s national focus on AI, its large data pool and massive market, as well as the presence of hard-working and ambitious entrepreneurs could help it overtake the U.S.
Lee is the CEO of Sinovation Ventures, an early stage venture capital firm that seeks to develop the next generation of high-tech Chinese companies. He is the author of AI Superpowers: China, Silicon Valley, and the New World Order. Lee was a recent guest on the Knowledge at Wharton radio show on Sirius XM, where he discussed these and other issues. (Listen to the podcast using the player above.) Next month, Lee is speaking at the AI Frontiers conference in San Jose, Calif., where Knowledge at Wharton is a media partner.
An edited transcript of the conversation follows.
Knowledge at Wharton: Where are we in the race for AI technology between the United States and China?
Kai-Fu Lee: Well, it’s not really a race. It’s two parallel universes, each making progress. U.S. is still way ahead in the core technologies from research labs and universities. But China is now taking the lead in implementation and creating value by using AI in all kinds of applications and industries.
Knowledge at Wharton: You write in your book that the skillful application of AI will be China’s greatest opportunity to catch up with and maybe surpass the U.S. But more importantly, it’ll help people rediscover what it means to be human. Can you explain what you mean?
Lee: I think the whole job market will change. We currently see narrow AI — not the science fiction, AI human intelligence — but specific AI engines that solve one problem at a time. For instance, [we see] engines that can make loan decisions for banks, customer service for large companies, simple robotics like fruit-picking and dish-washing. These vertical applications of AI are exceeding human capabilities. This means that routine jobs will be replaced by AI. But AI is also very good at producing tools for the creatives and professionals. I can see scientists, CEOs, writers, columnists, using AI as a tool.
All this will result in a significant job transformation. People in creative and strategic jobs will get their capabilities amplified and people in routine jobs will need to transform and transition to new jobs. Probably the only job category that’s large enough to accommodate that many people in routine work are service jobs. These are jobs that require a human touch, compassion and empathy, so it will be a very difficult transformation. But when it’s done, it will lead to many our population engaging in empathetic and compassionate jobs like that of nannies, teachers, doctors and caretakers for the elderly. This will create a positive energy and help us rediscover our humanity.
Knowledge at Wharton: We’re already starting to see that a little.
Lee: Yes. We are now seeing the beginnings of job displacements in the routine jobs. We see tellers, cashiers, starting to disappear, even without AI. With AI, more of these jobs will be gone. We’re also seeing a larger number of openings in jobs like elderly care. They’re currently not filled, perhaps because they do not pay enough or are not well-known as a job category. There’s not yet significant recognition in society that service jobs deserve respect and [good] pay. But hopefully, over time, we’ll see the need and then the pay and social status will increase to create an equilibrium.
Knowledge at Wharton: What is happening at present in China regarding AI that is different from what we are seeing here in the United States, to potentially put China in the lead in the next few years?
Lee: A couple of things are unique about China. First, Chinese entrepreneurs are much hungrier, they work much harder, and they are also much more tenacious. They are looking for all kinds of business models in which AI can help. AI in retail. AI in education. They are also working out operational excellence in applying AI to changing the way people eat, disrupting autonomous stores and autonomous fast food restaurants. So it’s displacing traditional industries faster.
“Chinese entrepreneurs are much hungrier, they work much harder, and they are also much more tenacious.”
Imagine convenience stores without people [manning the store]. Imagine fast food restaurants without people. AI is also being used in a lot of white-collar job displacement, which will impact the U.S. and China equally. I think China is moving faster because entrepreneurs are emboldened by the national priority on AI, funded by larger amounts of money. They see this as the hottest area.
The second reason, I think, is that the use of AI is no longer such a mystery. We think of AI as very advanced technologies that very few people possess. But actually, that is not true. AI is now open-source. New grads from college in a year’s time can start using AI in engineering and building these products. China has an army of new graduates who are all hungry to jump into AI as the new hot area.
Also, China has more data than anybody — and AI gets better with data. If you train an AI for, let’s say, an advertising engine or an ads-targeting engine, or a bank using AI for determining loans, the more data you have, the more accurate AI becomes. China has more users and more usage per user, because the use of digital services is pervasive. For example, China has almost no credit cards and no cash. Everyone’s using mobile pay. That’s fuel to make rocket fuel for AI to work better.
And finally, the Chinese government is very supportive of AI. Last July, it declared AI to be one of the most important areas to focus on. Provincial and city governments are building out cities the size of Chicago with autonomous vehicles in mind. There are two layers of roads. One layer is for pedestrians and the second is for cars, thereby limiting the possible accidents and casualties to the pedestrians. Highways are adding sensors to enable autonomous vehicles. These high-spend infrastructure projects are just what the AI industry needs, because private companies can’t possibly afford to build cities and highways.
Knowledge at Wharton: We talk a lot about the startup culture here in the United States, and the role that Silicon Valley has played in that. What does the startup culture look like in China?
Lee: The culture in China is different from Silicon Valley in a couple of ways. I think Silicon Valley tends to be more creative, innovative, wanting to be out of box, invent something no one has seen before. It frowns upon copycats, and it likes lightweight technologies. Instagram with 11 engineers gets acquired for a billion dollars. That’s the kind of story that Silicon Valley celebrates.
China is into incredibly hard work. Companies work 9 a.m. to 9 p.m. six or seven days a week without exception. Entrepreneurs are usually very strong, top down. A single person makes all the decisions. It’s data driven — so the decisions are very fast. There isn’t too much of consensus-building. It’s all about moving on and executing.
Chinese companies are better at raising large amounts of money because there’s a large market that can test ideas and scale them. Chinese companies are also willing to go heavy. That is, you build something that is incredibly messy and ugly and complex. But once you build it, it becomes a moat around your business.
For example, in the U.S., we have Yelp and Groupon, very lightweight companies. In China, there is Meituan, which has built a 600,000-person delivery engine, riding electrical mopeds with batteries that run out pretty quickly and have to be replaced. And yet, they run it to enable every Chinese consumer to order food on their way home and have it delivered to them by the time they reach their homes. The consumers don’t have to wait. The delivery time is 30 minutes and it costs about 70 cents. It’s the hard work that is shaving away a few cents a month, eventually getting to 70 cents per order. Then, they can break even. It is taking a large leap and a large bet and a large risk, because if they don’t succeed at 25 million orders a day, there’s a huge loss.
“We think of AI as very advanced technologies that very few people possess. But actually, that is not true.”
So it is a winner-take-all, gladiatorial, no-holds-barred kind of environment. It’s especially suitable for building powerful companies, or even monopolies. This is particularly so for AI because as you build a large customer base, you have a large amount of data, which gives you tremendous advantage.
Knowledge at Wharton: With the changes you expect to happen because of artificial intelligence, how is the economy going to be different in the United States in, say, 30 or 40 years? How is China adapting already to some of these changes?
Lee: The big benefits will be that AI will make companies more efficient and lower-cost. Existing processes running through AI can be made more profitable. By plugging in AI, Amazon gets more ad dollars. Google gets more revenue. Facebook gets more revenue. Microsoft gets more revenue and sales. When that starts to happen with banks, insurance companies and hospitals and so on, basically anyone adopting AI will see their P&L (profit and loss) improve. In some cases, AI will displace people and save on costs. In others, it will increase efficiency or deliver at higher margin.
PwC and McKinsey both estimate that by 2030, the world GDP will increase by about $12 trillion to $17 trillion, purely as net additional GDP, because of AI. This will make the U.S. and China wealthier. The wealth will go into the hands of a smaller number of people, those who take advantage of AI, so the wealth inequality will increase. One issue that’s raised is how does that redistribution of income happen? And, does it need to happen? This is because many people will be displaced from jobs. That’s one big question. The U.S. may need to look at ultra-high tax for ultra-wealthy people or companies. Whether that’s likely to pass through the system remains to be seen. China will face the same issue. But I think China will find it relatively easier to increase taxes.
The second big issue is how will new jobs be created? I think over a longer period of time — perhaps over 30 or 50 years — AI will create a lot of jobs and we may also be working less. We may be working three or four days a week. Some people may not need to work at all. So a lot of things could change. But in the meantime, people expect to work, and they need to be paid. How can we create those jobs so that the unemployment rate doesn’t suddenly increase? Unemployment rates are at an all-time low right now. That’s primarily because AI hasn’t yet started the displacement process. We will see that happen in the next two to five years.
Knowledge at Wharton: You mentioned the alternate universe that China is working on, especially with their internet. What was it that drove them towards this?
Lee: In developing a different internet ecosystem?
Knowledge at Wharton: Correct.
Lee: I think just entrepreneurship. In the beginning, a lot of American companies didn’t go to China either due to regulations that they didn’t want to accept, or because they felt it was too tough a market. So the Chinese entrepreneurs started copying the American ideas. This was not IP violation, but just copying the general idea of a search engine, a portal, an e-commerce site, and so on.
Over time, because of their consumer base and their entrepreneurship, they started to innovate. In the last three to five years, we’ve seen a lot of Chinese innovations that aren’t seen in the U.S. For example, for the young people in China, social media is dominated by a video-oriented social media system very different from Snapchat, Instagram, or Facebook in the U.S. And the payments system in China has grown to take over cash and credit cards.
“China has more data than anybody — and AI gets better with data.”
Imagine a parallel universe in which everything is paid for by giant software companies, and young people are in video social networks. The rest of the apps plug into a very different large piece of the puzzle in China. Think of China as one puzzle, with little pieces plugged in and the U.S. as another big puzzle. You can’t just take a piece from one and plug it into the other. That’s what I mean by the parallel universe.
Knowledge at Wharton: You can throw in WeChat as well, which has developed incredibly in the last few years.
Lee: Absolutely. WeChat is a giant Swiss Army app. It does everything. Think of this as Facebook plus WhatsApp plus Visa plus Mastercard plus everything. All the services you have, paying bills, and Uber and Airbnb — all these are part of this ecosystem. In the U.S., it would probably be subject to antitrust issues. But in China, it’s allowed to run. Half of my day is spent on WeChat. And I think for many people, like my wife, it’s even more than half.
Knowledge at Wharton: What lessons can we learn from the strategy of WeChat’s owner Tencent?
Lee: U.S. companies tend to focus and do one thing really well. Tencent strategically decided to build an empire for world domination. I think that is the difference. It had the ambition of Microsoft before the Department of Justice reined it in and said, “You can’t do that.”
Most of the practices are standard. Build a strong platform, add on top of that. Make smart investments in areas where you don’t have the competency. Keep building out and make big bets. They’ve spent billions getting their payments accepted. And, I think, a refusal to accept [defeat]. Four years ago, it seemed as if Alipay had won the payment wars in China. There were credit card companies and then there was Alipay, sort of like Visa/Mastercard in the U.S., and then Paypal.
But Tencent, as the Facebook of China, decided they were going to win in payments. They threw billions and billions at it. They subsidized people and created opportunities where people felt it was fun to connect their social network to their bank card. This tenacity, and never feeling that you are in X-industry so you can’t go into Y-industry, helped Tencent to totally disrupt the payment market. From zero market share it now has half the payments market.
Knowledge at Wharton: Are we at a point now where companies in the United States could learn from what is going on in China?
Lee: I think China is definitely worth learning from. Most of Silicon Valley still frowns upon China as merely a copycat. That’s a terrible mistake. Every Chinese entrepreneur is learning from China and from the U.S. They religiously read all the tech media — Wired, TechCrunch, and everything. If American entrepreneurs only learn from the U.S. but not China, they’re missing out on half of the opportunities, lessons and case studies.
Knowledge at Wharton: Based on some of the insights you have into AI and deep learning that’s going on in both Silicon Valley and China, which companies do you think — either in the United States or in China — are most advanced in their ability to transform business through the power of AI and data analytics?
Lee: Google, or Alphabet, is clearly by far the most advanced. If there is a disruption that completely changes everything that I stated in my book, it would probably come from Google. They have a phenomenal system from the hardware chips up to the platform level, and they apply it to many, many areas. I think they’re by far the most ahead in the core technologies.
“Most of Silicon Valley still frowns upon China as merely a copycat. That’s a terrible mistake.”
In a very clever implementation, with maybe some Chinese spirit, is Amazon. I think their technology team is elementary compared to Google’s technology team. Yet they are able to find the applications, and they are willing to make big bets. I think these two companies are leading in the U.S. Facebook is very good, but they need to recover. They have a strong AI team, but we don’t really see the benefit as yet. Theoretically, AI should help them fix a lot of the newsfeed problems and the PR issues they face.
In China, I think Tencent is by far the most powerful company. Their use of AI has been modest. I guess one could see that as a potential upside. Alibaba is applying AI much more rapidly, because they’ve been in payments and commerce, and they can see money coming out of AI. They’re probably leading in that. Baidu is the Google of China. They probably have the most AI scientists in China, but they haven’t done as much to create value. So that remains to be seen.
Knowledge at Wharton: What are the biggest breakthroughs that you see on the horizon for AI?
Lee: Actually, I don’t. I think AI is like electricity. Based on what has been invented plus the incremental improvements, we’re going to see amazing things, including autonomous vehicles, which I don’t view as requiring a lot of new technologies. It’s just a matter of gluing everything we know, and incrementally applying it to the application. We are in the midst of AI application, taking what is known and creating value in things like autonomous vehicles, autonomous airplanes, and smart robots. I think that will happen without any fundamental breakthrough.
Some of the big issues that are ahead are, can AI learn from a few examples? Can AI learn to have common sense and to learn multi-domain? Can AI learn by itself? And can AI start to have common sense? Another big question is, can an average engineer learn to use AI with just hours of training? I think these are interesting problems that we may or may not have solutions for in a couple of years.
