Why China Will Overtake the U.S. in AI
IEEE Spectrum: Why do you believe that China will soon match or even overtake the United States in developing and deploying AI?
Kai-Fu Lee: The first and foremost reason is that we’ve transitioned out of an era of discovery—when the person who makes the discovery has a huge edge—and into an era of implementation. The algorithms for AI are pretty well known to many practitioners. What matters now is speed, execution, capital, and access to a large amount of data. In each of these areas, China has an edge.
That’s why I began the book by talking about China’s entrepreneurism. It’s not like Silicon Valley, which is built on iPhone breakthroughs and SpaceX innovations, it’s built on incredibly hard work. Chinese entrepreneurs find areas where there’s enough data and a commercially viable application of AI, and then they work really hard to make the application work. It’s often very hard, dirty, ugly work. The data isn’t handed to you on a silver platter.
Spectrum: You say that Chinese tech giants like Tencent have a clear advantage in terms of access to data that’s needed to train AI. Do they really have more data than companies like Google?
Lee: There are a few ways to look at the data advantage. The first is how many users you have. Google probably has more users than Tencent, because it’s international. The second question is: How homogenous is your data set? Google’s data from Estonia may not help its work in India. It may be better to have rich data from one set of people who have the same language, culture, preferences, usage patterns, payment methods, and so on.
The third way to measure is how much data you have about each person. Tencent has a catch-all app, WeChat, that does basically everything. The average Chinese internet user spends half of his or her time online in WeChat. When you open WeChat, you have access to everything U.S. users get from Facebook, Twitter, iMessage, Uber, Expedia, Evite, Instagram, Skype, PayPal, GrubHub, LimeBike, WebMD, Fandango, YouTube, Amazon, and eBay.
Spectrum: You describe China’s startup ecosystem as a brutal “coliseum” where companies don’t win because they’re the most innovative, but rather because they’re the best at copying, using dirty tricks, and working insane schedules.
Lee: There is creativity, but it’s just one tool. Another is copying. Entrepreneurs do whatever it takes to win, to build value for the user, and to make money. If you look at WeChat, you can’t point to one moment when it shocked the world like an iPhone. WeChat today is an amazing innovation, but it didn’t come about because someone at Tencent dreamed it up and built it and shocked the world. They kept layering on features that users wanted, they iterated, they threw away the features that didn’t work, and at the end they had a product that was the most innovative social network. It’s so good that Facebook is now copying them.
“50 Percent of Jobs Are in Danger”
Spectrum: You write that the big AI question isn’t whether China or the United States will dominate. Instead it’s how we’ll deal with the “real AI crisis” of job losses, wealth inequality, and people’s sense of self-worth.
Lee: AI will take many single-task, single-domain jobs away. You can argue that humans have abilities that AI does not: We can conceptualize, strategize, create. Whereas today’s AI is just a really smart pattern recognizer that can take in data, optimize, and beat humans at a given task. But how many jobs in the world are simple repetitions of tasks that can be optimized? How many jobs require no creativity, strategizing, conceptualization? Most jobs are repetitive: truck-driving, telemarketing, dishwashing, fruit picking, assembly line work, and so on. I’m afraid that about 50 percent of jobs in the world are in danger.
Whether these jobs will disappear in 15 years or 20 or 30, that’s debatable. But it’s inevitable. Not only can AI do a better job, it can do the job for almost marginal cost. Once you get the system up and running you just pay for the server, electricity, bandwidth. To be competitive, companies will be forced to automate. And this shift will happen a lot faster than has ever happened before in the history of humanity.
Spectrum: Why do you think “techno-utopians” have it wrong when they say that AI will ultimately create entirely new categories of jobs, just like the industrial revolution?
Lee: People say that in the industrial revolution, more jobs were created than destroyed. They say it was the same with electricity, and that we shouldn’t worry, because the same thing will happen this time with AI. I would agree with that if we had enough time. Those earlier technological revolutions took a century or longer. Electricity has been around for over 100 years, and we still don’t have electric cars, we’re still working on the grid. That gave people time to grow, and develop, and invent new jobs. But we have basically one generation with AI, and that’s a lot less time.
Spectrum: You argue that even if governments figure out a way to distribute money to all these jobless people, it will still be a crisis.
Lee: Most people don’t think of their job just as a source of income. It brings meaning to their life, it’s their contribution to the world. That’s how we decided to structure our capitalistic society: There’s the idea that even by working routine jobs, they can make money and make better lives for their families. If we pull the rug out from under them and say, you have no job, but here’s some money from the government, I think that would lead to bad outcomes. Some would be happy and retire early. Some will learn a new skill and get a new job, but unfortunately many will learn the wrong job, and get displaced again. A large number of people will be depressed. They will feel that life has no meaning, and this can result in suicide, substance abuse, and so on.
The Inevitability of the AI Revolution
Spectrum: If this kind of economic and societal upheaval is an inevitable consequence of AI, is there any chance that we’ll decide to turn away from the technology, and decide not to use it?
Lee: Individual governments can make certain decisions to slow down the deployment of AI. But for humanity as a whole, it’s not possible. We’ve opened Pandora’s box. We did, as humans, control the proliferation of nuclear weapons, but that technology was secret and required a huge amount of capital investment. In AI, the algorithms are well known to many people, and it’s not possible to forbid people to use them. College students are using them to start companies.
Take autonomous trucks as an example. While China is building cities and highways to facilitate autonomous trucks, the American trucking union is appealing to President Trump to forbid testing on highways. If the U.S. is currently ahead on autonomous trucks, but chooses to slow down the development for fear of taking away jobs from truck-drivers, the only outcome is that China will catch up. Chinese companies will test the trucks, get the data, that data will make the AI better, at some point the technology will be so good that China will export it to the rest of the world. At that point, the U.S. will still have to give in to automation.
Spectrum: You call this a grim picture, but then go on to say that there’s hope, and that the potential for human flourishing has never been greater. You had your own grim experience that led to your own flourishing. Can you talk about your cancer experience?
Lee: I had been a workaholic for my whole life, and I always put work as my top priority. It was only when I faced cancer and possible death that I realized that no amount of money, success, or fame could substitute for the love I have from other people. And I felt great regret for not giving back what they gave me. That was a wake-up call.
After I became better and my cancer was in remission, I changed my life. It’s not that I don’t work hard anymore, I still do. But I prioritize differently. I prioritize family, I found a much better balance. I realized that the optimization that I used to do at work brought me money, success, and fame, but those are not the things that really matter to me… although I once thought they were.
A Blueprint for Coexistence
Spectrum: So that experience led to your idea for what you call a blueprint for coexistence with AI, in which the use of AI gives people more time to love each other. You write that we must “forge a synergy between artificial intelligence and the human heart.” Can you give me an example of how this synergy might manifest in the job market?
Lee: One of my favorite examples is in medicine. Imagine a future clinic, in which the room is enhanced with all sorts of sensors that take readings of your body and give the human doctor lots of information. The doctor will help tease out information like family history and specific symptoms.
AI can make a great diagnosis and suggest the treatment, the prescription, and so on. In early days, AI will give statistics and the doctor will make the final choice. But in time, it will be very rare that the doctor overrides the system. So the AI will do the diagnosis, then the doctor will deliver the message in a way that feels caring and warm. The doctor will also let the patient tell his or her story. In many countries, each person gets only five minutes of doctor’s time—but while the doctor may only need five minutes, the patient needs much more to feel heard, to ask questions, and to be reassured.
If each doctor spends more time with each patient, there will need to be more doctors. Maybe they don’t need to be full MDs with 10 years of education and internship, since they no longer need to memorize all the symptoms and treatments. Instead they could get four years of education to become caring, compassionate caregivers, similar to nurse practitioners. Then the cost of health care will come way down, people will get better care, and the number of caregivers will go way up.