This book is an eye opener for those of us unfamiliar with the wide ranging capabilities and imminent impact of AI. Lee tells us about the development, design and future of AI and associated web and mobile technology. He contrasts Chinese work in AI with that in the US. While Chinese AI is based on technologies developed in the US, Chinese companies are now taking their own direction. Lee makes a strong case that AI will have profound consequences for society and determine the relative power of nations. He means this in terms of economic power. He does not discuss military applications.
Kai-Fu Lee is CEO of a Chinese venture capital firm investing in AI. A former president of Google China with a PhD from Carnegie Mellon University he has thirty years' experience in AI research. From just the application of recently developed technology called "deep learning", Lee predicts society will be cast into worldwide upheaval. For example automated factories will eliminate the need for low cost labor devastating developing countries, allowing factories to be built near consumers. The economic demise of those displaced by AI in developed countries will exacerbate income inequality. Successful AI entrepreneurs and their companies will gain an even larger share of national income than their internet and social media predecessors eclipsing people such as Microsoft founder Bill Gates and Amazon founder Jeff Bezos. It's a scary picture but Lee evaluates ways to address the problem and offers his own solutions.
Lee believes China will lead in implementing AI with the US second. China secures an advantage because it is investing increasingly more in AI than the US. Chinese AI and other software startups are offered rent subsidies and tax discounts to move into designated zones to create technology incubators in the image of Silicon Valley. Local governments have started "guiding funds" which have grown exponentially in a few years to $27 billion in 2015 for investment in venture capital firms focused on AI and supporting technologies. China is also investing heavily in AI related education. The Chinese government is prioritizing AI because it sees AI leadership as a way to increase its world power and spread its culture just as the West has with past technology leadership. Google's Eric Schmidt in 2017 said that in five years China would equal the US in AI.
Americans often blame Chinese government interference for limiting their success in China; however Lee notes that the domestic success of Chinese internet companies has been due more to their customization for the Chinese market than government controls. Lee attributes Chinese sensitivity to its users to its flexibility "Unencumbered by lofty mission statements or 'core values', they had no problem in following trends in user activity wherever it took their companies." Lee believes a market where copying is problematic such as the US allows the originator to maintain a significant competitive advantage making them less responsive to customers.
Lee describes competition between Chinese AI companies as cutthroat; anything goes. He describes the work ethic as "maniacal", well beyond the intensity of Silicon Valley. American companies eschew copying from each other. Chinese companies are happy to take technology, business models, employees, whatever they can get from their domestic or international competitors making them more agile. In China they do not face lawsuits or antimonopoly investigations for their actions only the counterpunch of their competitors which can include smear campaigns and even charging competitors with crimes. Lee sees Silicon Valley entrepreneurs as "mission driven". Take an original idea and achieve an idealistic goal. He sees their Chinese counterparts as "market driven". It doesn't matter where an idea comes from, only that it makes a profit. In Lee's words "The core motivation for Chinese market-driven entrepreneurs is not fame, glory or changing the world. These things are all nice side benefits, but the grand prize is getting rich, and it doesn't matter how you got there."
Lee does note that if one of the big seven in AI research (four US, three Chinese - Google, Facebook, Amazon, Microsoft, Baidu, Alibaba and Tencent) makes a game changing breakthrough, it would keep that to itself, garnering a huge competitive advantage and rendering the implementation of present AI technology short-lived. Among the seven, Lee gives Google the best shot at such a breakthrough. It has over half of the top 100 AI researchers in the world and spends twice as much on math and computer systems research as the US government. However he still considers such a breakthrough more likely to occur in academia which openly shares its results. And the record of such significant breakthroughs shows they do not occur frequently.
If a major proprietary breakthrough in AI technology in the near term is unlikely, ability to implement becomes the key issue rather expertise in research, favoring China. The amount of data available to feed current state of the art deep learning AI systems determines how well they operate. This too favors China. We are not yet in the age of generalized AI systems. Deep learning AI systems need large amounts of data to learn a specific task, say controlling an autonomous vehicle or facial recognition. One platform, WeChat, can provide everything one needs to know about Chinese habits. WeChat owned by Tencent is a mobile super-app. WeChat allows other developers to incorporate apps within it. With WeChat users can exchange messages, connect with friends, transfer money, make investments, shop, make reservations, order meals, make a doctor's appointment, unlock a city bike, get a taxi, have groceries or prescriptions delivered, get movie tickets, pay traffic fines, and on and on. Mobile payments in China were $17 trillion in 2017. They are quickly replacing credit cards and cash. Some beggars hang QR codes around there neck to accept mobile payments. WeChat apps collect data on what you buy, who you send money to, the food you eat, the doctor you see, the medicines you use, where you take your bike share or ride share and much, much more. In the US this data is scattered among many platforms and Americans are pushing back on the invasion of their privacy. As I was writing this I read that Apple CEO Tim Cook attributed declining iPhone sales in China in part to WeChat. WeChat does everything and it does it just as well on a cheap Android as it does on an iPhone.
Lee sees four waves of AI development. The first wave is already with us. Algorithms are deciding the ads we see, the videos we are offered and the news we read. The key is data. The more that an AI engine sees of your clicks, pictures and videos viewed, articles and tweets read, the better it can give you what you want leading to a more addictive experience. That addictive experience is the goal since more clicks mean more profit for the website. A Chinese site, Toutiao, called ByteDance in English, is a purveyor of trending news like Buzzfeed. Except rather than use reporters it uses AI to find content that you want even creating its own headlines based on your preferences. The company is experimenting with the algorithm writing its own articles. It can summarize a sports event and post it two seconds after the event ends. These algorithms can both create and detect fake news. Toutiao pits two such algorithms against each other to hone their abilities.
The second wave of AI development is also already with us, in the US more than China. It's about business. Insurance companies use it to determine risk, banks to determine credit, hedge funds to trade stocks, and pharmaceutical companies to design drugs. The reason the US leads in this area is the same reason China leads in consumer applications - data. US companies have long used data bases to store massive amounts of information in formats readily accessible to AI engines. An emerging application is disease diagnosis. AI engines are thorough taking in and weighing far more detail than a human. AI diagnoses can be used as guides letting the doctor make the final decision. AI can help level the quality of care for underserved poor and rural areas. In China AI is being tested for court systems helping judges decide guilt or innocence and an appropriate sentience. This can help eliminate bias.
The third wave of AI development is just emerging. This is perception AI, AI that can recognize objects, voices and faces. Already a test KFC in China is charging customers based on facial recognition alone. The sensors and algorithm first check that the subject is alive to prevent being fooled by a picture. The WeChat wallet is updated immediately. The customer does nothing except to place their order and take their food. Perception AI requires extensive use of sensors. This is already common in China allowing urban traffic flows to be controlled. In the home devices like Amazon's Echo interfacing with smart devices allow people to control their environment. Combining home with store AI is next. For example your grocery cart could receive your home shopping list augmented by data from your refrigerator and your purchase history. And of course it would ring up and pay for your order just by placing items in the cart. Education is ripe for development with AI determining a student's needs, producing appropriate work assignments and grading them. AI could identify students requiring tutoring or those with exceptional abilities. The human teacher would still lecture and assist students.
The fourth wave of AI is autonomy. We are familiar with the tests of autonomous vehicles. The development of autonomous cars relies mostly on collecting more real world data which Tesla and the real technology leader, Google, are actively doing. Chinese companies are also developing autonomous cars but most experts are already with US companies which started working on this complicated application years ago. While Chinese companies are behind US companies they face fewer restrictions and have more government support than US companies. A California startup has devised a machine that can see and delicately pick only ripe strawberries. Now it is pulled by a conventional tractor. In the future it will be autonomous. Drones for everything from firefighting, search and rescue, product delivery and much more are coming. In Amazon's warehouses the shelves come to warehousemen who stay in one place filling boxes for delivery. The next step seems obvious. As for the home environment Lee does not believe current AI is ready for useful robots. The tasks just to clean a house are too diverse for today's technology.
While Lee sees artificial general intelligence (AGI) decades away including its ominous possibilities, he does see potentially dire consequences from the full implementation of current AI technology. These consequences are widespread loss of jobs, devaluation of labor, increased income inequality between nations and within nations. The impact will disrupt society and the balance of power in the world. Poor countries relying on cheap labor will be hit the hardest. As factories automate labor costs decline allowing factories to minimize transportation costs and improve service by moving near their markets in developed countries. Data driven monopolies will have significant advantages displacing smaller competitors. The AI superpowers, China and the US, will forge ahead while the developing countries will fall further behind. Within China and the US the divide between the rich and poor will be even greater.
PricewaterhouseCoopers estimates AI will increase worldwide GDP by $15.7 trillion dollars by 2030 with 70% of that going to the US and China. Since AI consists of algorithms that are easily installed where needed and maintained remotely, this revolution can take place at unprecedented speed. Jobs that don't require social interaction or advanced robotics are most at risk. White collar as well as blue collar jobs are at risk. Some examples: Radiologists are at risk, Psychiatrists are not. Loan underwriters are at risk, PR directors are not. Dishwashers are at risk, dog trainers are not. Fast food preparers are at risk, Physical therapists are not. As I write this the Brookings Institute issued a report that 36 million US jobs are at high risk to be replaced by automation in possibly the next few years although it could take up to two decades. Companies typically invest in labor saving technologies during downturns so a recession would speed things up. Lee shares estimates of how many jobs will actually be displaced in the US where almost all studies have been conducted. These vary from a low of 9% all the way up to 48%. Lee believes 38% is the number of jobs at risk but this could be ameliorated by social policies cutting actual displacement in half and some new jobs will be created. But the value of labor will also drop meaning lower wages for those still employed. The impact on the self-worth of most workers will exacerbate social and political divides.
Lee evaluates prescriptions for dealing with an AI world with too many people for too few traditional jobs. He considers retraining, reducing work hours and guaranteed income. Retraining should be done but will have limited impact. Reduced hours may help some but will reach its limits as will income redistribution. If people are paid for doing little or nothing their sense of self-worth can be undermined creating an underclass. It would lead to a divided society. Are you one of those who work to provide for the world or one who we support Lee believes there needs to be a new social contract in which government and industry invest in human oriented service jobs. AI will do the thinking but it has no emotion. Lee wants to use the tremendous gains from AI productivity to help fund jobs such as elderly caregivers, patient facing health workers, student supporting educators - jobs in which compassion and a human touch can improve people's lives. It would be a remarkable shift to divert corporate profits to create this caring society. In part Lee's ideas stem from his life changing experience being diagnosed with stage four lymphoma. He wants AI to do the diagnoses but a human not a bot to deliver the news and explain it.
Lee recovered with treatment but it made him question what is really important in life. The last chapters of his book reflect this seismic shift in his outlook. He went from seeing his value as revolutionizing the world with game changing technology to valuing personal relationships; being there for family and friends and helping people lead better lives. The dichotomy is reflected in the future we face as AI intrudes into every aspect of our lives. Will it be utopia, dystopia or just a muddle.