Taiwan-born Kai-Fu Lee knows Silicon Valley well. He worked at Apple, Microsoft and Google after graduating from Carnegie Mellon with a PhD in computer science. He headed Google's China operations before setting up a Beijing-headquartered venture capital company, now called Sinovation Ventures, in 2009.
Lee's upcoming new book "AI Superpowers: The U.S., China and the New World Order" carries blunt advice to U.S. companies about the rising challenges that America faces from China in the all-important field of artificial intelligence. More broadly, Lee also said in a recent interview in Shanghai, China benefits from the competitive fervor of its business culture. China's entrepreneurs, he said, "are hardworking, hungry, never satisfied with the present success, always seeking more, and also incredibly resourceful to the degree of ruthlessness. That is very much lacking in the relatively gentleman way in which American companies compete. And it turns out that if you have the ambition to build a very strong company and empire and build a mote around your empire, that is very different from the Silicon Valley way."
Relatively speaking, U.S. Internet companies also tend not to adapt their products once successful, Lee believes. "In the U.S., once you reach success, the product actually stagnates. If you look at YouTube, or Instagram or Snapchat, they look the same as they did five years ago. But if you go to a Chinese Meituan or a Didi or a WeChat, they look completely different from five years ago."
Lee's analysis should be a wake-up call to Americans used to thinking of the U.S. as the unrivaled No. 1 in the technology world and the center of innovative entrepreneurism. Excerpts from my conversation with 56-year-old Lee follow.
Q. What was your motivation in writing the new book?
A. The primary motivation is there are a lot of misperceptions about AI. People either think of AI as futuristic and not here yet, or as something like science fiction. That’s not correct. There are also people who think that the U.S. will define the future of AI. That’s also not the case, because China and the U.S. will co-define AI.
A lot of people focus on issues with AI – that’s the No. 1 focus, especially due to the Facebook and Cambridge Analytica issues. But I personally think the biggest challenge is the job displacements and how we should rethink the meaning of our lives, not in terms of how hard we work but in terms of why is it that people exist, because the definition of life as "work as hard as you can and make lots of money" isn’t relevant anymore. AI can do most jobs that we do today. These are the types of misconceptions that I wanted to correct.
Q. How should we be viewing AI?
A. There is narrow AI and there is general AI. Let’s talk about narrow AI, because general AI isn’t here -- we don’t know to do it. General AI is like building up another human.
Narrow AI is within a narrow domain: to collect a huge amount of data and to be able to make decisions, classifications, predictions at an amazing level of accuracy that exceeds human capabilities. We see that in Alpha Go, in determining loan default rates, investment returns, customer services and factory inspections. The AI capabilities are phenomenal, but it’s built upon one domain, not multi-domain, and upon lots of data. If you can’t get data or if you have to (work) cross-domain and have to deal with common-sense reasoning, then AI is no good for that. However, most human jobs are single-domain jobs. People are not very good at doing those jobs compared to what machines can do. That’s what leads to the job issue.
Q. What are some of the other misperceptions that you want to correct?
A. One is that AI is like building another human. It has nothing to do with building another human. It is a purely statistical algorithm that learns from a massive number of examples – way more than humans can absorb. So it’s really smart in the sense that they make very accurate decisions, but it’s kind of dumb in the sense that it has no common sense, so to speak. It just makes decisions. A human says, "Minimize the default rate on a loan, or maximize customer satisfaction, or maximize the face recognition rate or the speech recognition rate." It’s a merely humanly defined, numerical objective function for one domain in which AI optimizes. That is what it is good for. It’s not good for general human thinking. It’s not good for empathy. It’s not good for common sense. It’s not good for self awareness. It doesn’t do any of that. That’s the big misperception.
Q. What will the U.S. role be vis-a-vis China?
A. Well, the U.S. will be fine in the sense that the U.S. will be one of the two superpowers. My new book is called "AI Superpowers." One interpretation is that there will be a duopoly – U.S. and China. The U.S. for obvious reasons is the strongest in research and technology, has a powerful ecosystem, has Silicon Valley and the deepest bench in terms of research talent. But China comes in as a surprising black horse that’s ready to challenge the throne of the AI emperor, if you will.
What gives China that opportunity? China is behind in those things that the U.S. is ahead in, but China is ahead in three main things: China has a huge amount of data. Because AI is about optimizing on the huge amount of data, China having more data – whether in face recognition, more people with mobile usage, or 50 times more mobile payment transactions – all of that is a treasure trove of data that is like rocket fuel to make China’s AI better. That’s reason number one. And I say that is the priority.
Reason number two is that China is full of dedicated, hardworking, fast-decision making entrepreneurs who have learned over the last 10 years how to listen to the market, iterate the product and maximize the product capabilities. That’s what brought about WeChat, Weibo and Toutiao and the other great Chinese products. These entrepreneurs are hardworking, hungry, never satisfied with the present success, always seeking more, and also incredibly resourceful to the degree of ruthlessness.
That is very much lacking in the relatively gentleman way in which American companies compete. And it turns out that if you have the ambition to build a very strong company and empire and build a moat around your empire, that is very different from the Silicon Valley way. In Silicon Valley, Pinterest looks at Instagram and says, "You’ve done the Instagram thing and I’ll do the Pinterest thing. We’re going to respect each other and cooperate and compete." But in China, it will be like, "We’re both in the same space. Either you kill me or I’ll going to kill you."
Companies in the U.S. are more satisfied with accomplishments. In China, there’s always striving to build a very strong moat. So while in the U.S. a company like Groupon or a company like Yelp or Instagram are happy building a very shallow layer of technology and then let other people do the heavy lifting, people in China will dive into doing tough, ugly and unappreciated work but ultimately for that 10% better in margin. It’s like Meituan versus Yelp: Meituan is hiring tens of thousands people to deliver food, while Yelp is saying, "We’re just a review site." The examples go on.
This kind of entrepreneurial push forward is going to cause to China to have a great opportunity with AI, because it’s all about launching, collecting data, making your product better, having a larger usage, raising more money, combining more machines, and getting more scientists. Then, the loop continues. In the U.S., once you reach success, the product actually stagnates. If you look at YouTube, or Instagram or Snapchat, they look the same as they did five years ago. But if you go to a Chinese Meituan or a Didi or a WeChat, they look completely different from five years ago.
The third reason is the government’s techno-utilitarian approach. A lot of American rhetoric is that "the Chinese government is pumping money, that’s unfair." But it’s not at all about the money. The Chinese government support is strongest in two forms. The first is in infrastructure building. For example, China has built high-speed rail network and it has built a very strong 4G network. And today, China is building infrastructure that will support AI technologies like the autonomous vehicle, for example. Xiongan will be a new city supporting autonomous vehicles. The U.S. probably wouldn’t think of building a new city to support autonomous vehicles. Zhejiang Province is repaving highways to make them safer for autonomous vehicles. Compared to the U.S., where truck driver unions are asking to slowdown the autonomous truck testing on highways, China is going the opposite way.
Q Do you see American companies responding to the changed environment more quickly in the future?
A. I live in China so I’m not in a position to speculate about how fast the U.S. will respond. But I can say that it’s understandable that most of America takes its leadership for granted. Because if you look back, 20 years ago, Wintel was the strongest duopoly in the whole world. No country had any chance of building up a PC that wasn’t on the Wintel ecosystem.
Over time, that American self-entitled leadership eroded as we went into the age of lithium, the age of mobile, and the age of social. The American self-entitled global leadership has been shrinking. It's in the mobile age that China came of age. China is roughly entitled to 50% of the world’s market capitalization of the world’s mobile companies. If you look at the values of public companies as well as private companies -- Xiaomi, Meituan, Didi and others and Ant Financial, China is worth roughly as much as the American companies. But the game hasn’t finished because there is still the global expansion that China has yet to try to accomplish.
Q. So part of the big conversation today is about the U.S. tariffs. How will this play out in Chinese policy and investment decisions?
A. I’m not an expert on policy so I can’t tell you that, but I can tell you that the Chinese companies are expanding globally and in most cases, in areas where the U.S. has no control. For example, if you look at where will the future of mobile internet be – let’s not even go into AI, you will see that American companies have a very strong stronghold in the U.S., English-speaking countries and Western Europe.
Chinese companies are going successfully – slowly but successfully -- into Southeast Asia and Islamic countries, and also probably Africa in due time. Unknown territories will be Eastern Europe and South America. But that is pretty much the world.
The Chinese companies approach is worth observing. America is used to technological hegemony – that is, dominating by default and without competition. Google insists on one platform. Facebook insists on one platform. Uber insists on one platform. But the Chinese companies, having been a technologically colonized country, are sympathetic to other countries that want their indigenous companies. So most of the Chinese companies, if you look at Tencent, Alibaba and Didi, aren’t expanding their brand to other countries. They are partnering with other countries in investing in the local competitor and helping them with mobile and AI interface and software. If you look at Didi’s creation of the anti-Uber alliance, it’s a very clever approach. It’s unreasonable to think that any one company can understand the whole world, every country’s policies. Every countries has their own laws and policies. So Didi is investing in companies in Indonesia, Brazil, Singapore, Malaysia and so on, taking a stake, giving them help in technology and AI and software, and letting them build a local brand.
I think that approach has more legs than the American hegemonic approach. I would predict a world in which China and the U.S. will both have large shares of the world’s users. That’s not going to be affected by any tariffs because I think Chinese companies by and large aren’t going to be that successful in the U.S. and vice-versa. So the skirmishes will be in fought in other countries, not in the U.S. or China.