Review of AI Superpowers: China, Silicon Valley, and the New World Order, by Kai Fu Lee

Rating: 7/10* (9/10 for the first six chapters)

The Review

In a nutshell: AI Superpowers is an China-optimistic perspective from a Chinese venture capitalist on the state of advancing competition between China and the rest of the world, which one might (understandably) believe was only Silicon Valley after having read this book. In the first several chapters, Lee sets up a framework to defend his optimism for Chinese firms potential to compete in artificial intelligence by contrasting Chinese entrepreneurial culture to Silicon Valley’s, and drawing stark contrasts between American and Chinese governments approaches to their respective technology industries. Some of Lee’s claims require scrutiny, including his bolstering for Chinese government, borderline-advocacy of ruthless business practice, and characterization that Silicon Valley is pitting elite researchers against Chinese good-enough engineers. However, the book is an invaluable resource for anyone seeking to gain better understanding of how entrepreneurship operates in China, and how the rising Chinese technology economy may overwhelm expectations as an unprecedented powerhouse in emerging technology areas.

Secondarily, Lee’s analysis is comprehensive in the middle of the book. Lee gives a careful analysis of the four application “waves” of Artificial Intelligence, in considering those that Chinese firms will be most competitive in. He weighs questions about “grid” versus “battery” approaches to artificial intelligence. Championing for the Chinese government’s inefficient, yet clear, approach to stimulating growth with massive incentives, and invasive data collection practices, Lee’s analyses are sometimes poetic, but always optimistic.

In the last half of the book, Lee changes topic. His close encounter with cancer drives him to climb a mountain, to talk to a monk, to celebrate a volunteer golf cart driver, and finally present his thesis on how AI tech has the capacity to displace whole societies. His last two chapters describe what we as a society might do to limit the negative consequences of what will be gradual but massive displacement and unemployment. He proposes a new culture code for society, around the principles of human love, service, and compassion, over present imperatives to find productivity. His transformation story, both for himself and for society land a bit clumsily: his personal transformation from human-turned-work-algorithm to friend and teacher read as an abrupt turn: an out of place single chapter in the book.

By the end of the book, I felt I had a plausible and detailed thesis for how Chinese entrepreneurs might compete with their American counterparts. I understood the advantages that Chinese businesses wield, and the role the Chinese culture and government may play in that transition. The latter half of the book read more clumsily. Lee’s arguments for how we may transform society to avoid dramatic social upheaval were well considered, though less insightful, and more castle-in-the-sky than his exposition on the Chinese technology story.

First half (6 chapters): information dense enough for a 9/10.

Second half: often off topic, with a cancer story, a thesis for positive societal change, unfortunately punctuated with emotionally disingenuous personal anecdotes. 5/10

Total: 7/10

Things I learned


1. The “Chinese Sputnik Moment”: AlphaGo’s victory over Go champion Ke Jie in 2017 can be seen as a turning point for Chinese perception of Artificial Intelligence.

2. AI and hardware are now to a sufficiently advanced state that entrepreneurship skill may matter more than research chops.

3. In China, if your only advantage is an idea, copyright be damned, your business will be copied and outcompeted.

4. China’s copycat technology scene allowed it to build a technology industry with competent engineers and competitive, if not original, entrepreneurs.

5. American companies are mission driven: matching the exact engineer or researcher to the perfect business. Chinese companies are market driven: any business model, any product, Chinese entrepreneurs look to make money.

6. “Growing up, their parents didn’t talk to them about changing the world. Rather, they talked about survival, about a responsibility to earn money so they can take care of their parents when their parents are too old to work in the fields.”

7. The Chinese internet ecosystem has three foundations: The copy culture, scarcity mentality, and willingness to pursue any business opportunity.

8. The Chinese government is trying to “brute-force” the culture of Silicon Valley by supplying huge government subsidies for “VC firms, startups, incubators, and service providers” in specific geographic areas. This stands in direct opposition to the Silicon Valley zeitgeist: innovation and original thinking can’t be built “merely using bricks and rent subsidies.”

9. Chinese online commerce is now populated with “O2O” services, online to offline businesses bringing everything from cars to haircuts to where users are.

10. The heavy-light contrast: American internet companies move bits, not atoms, to connect people to knowledge and services, and leave brick-and-mortar logistics to existing businesses until at giant scale. Chinese companies, with abundant access to cheap labor are much more comfortable moving whatever people or objects (eg. scooters) need moving, and doing logistics at any scale. Moving atoms is harder to copy, making this a serious advantage for Chinese entrepreneurs.

11. Because of this and weak Chinese privacy regulations, Chinese startups often have access to data (payment, location, etc.) that American internet companies rarely would.

12. With WeChat, Tencent has the richest data ecosystem of all technology giants.

13. (Particularly interesting) 4 Waves of AI model: Internet AI, business AI, perception AI, and Autonomous AI. China is in the best position to take advantage of perception and internet AI, and has the weakest position in business AI.

14. Internet AI is the oldest wave, and uses recommendation engines to learn preferences and serve content, search results, videos, social media posts, products, etc. The service will continue to grow in value for large content serving companies, and may play a role in spotting fake news.

15. Business AI takes advantage of data to make decision recommendations, for instance, loan data informing bankers, and medical data informing doctor decisions. Humans make decisions using strong features, though algorithms can consider weak features, allowing them to make far more complex analyses. Chinese businesses don’t buy into enterprise software or standardized data storage as much, making western companies a better fit for this model.

16. Perception AI digitizes the physical world to understand objects like faces, to see the world. Chinese cities have more cameras and interfaces to collect data about people, therefore, more data. More data allows Chinese firms to more seamlessly merge online and offline interfaces, what Lee calls Online Merge Offline (OMO). He gives the example of the supermarket making recommendations, and teachers providing tailored curriculum. Perception AI is “hardware heavy”, and will continue to grow as the IoT/5G trend allows more perception devices to exist in our periphery, collecting data about us. Hardware development in particular is a Chinese specialty.

17. Autonomous AI integrates algorithms with machines acting autonomously: self driving cars, warehouse robots, and swarms of delivery drones. This wave of AI is the most visible: there’s nothing like seeing a swarm of autonomous robots. As mentioned above, Chinese companies have an advantage in hardware manufacturing and iteration. This market will be dominated entirely by large corporations who can iteratively manufacture hardware, and design well trained, and safe algorithms. American companies are more likely to look like Google: creating the most advanced algorithms, than Tesla (which Lee analogizes to the Chinese model): collecting the most data. In addition, the Chinese government is willing to experiment with changing everything down to the roads (see Xiong’an) to improve the navigation capacities for autonomous vehicles.

18. Model of General Purpose Technologies (GPT) and the analogy to the internet, electricity, and the steam engine: upending traditional modes by deskilling tasks (craftsmen work becomes pull-lever, push button tasks).

19. Electricity and the Steam engine empowered large numbers of low skilled workers to take on repetitive, yet productive jobs, and increased overall societal standard of living.

20. The internet grew worker productivity, but not median income or employment rates. Societal standard of living has stayed constant through the internet revolution (Brynjolfsson and McAfee).

21. “One reason why ICT [internet connective technology] may differ from the steam engine and electrification is because of its “skill bias.” While the two other GPTs ramped up productivity by deskilling the production of goods, ICT is instead often—though not always—skill biased in favor of high-skilled workers.”

22. Moravec’s Paradox: it is easy for AI to mimic high intelligence/computational abilities, and hard for AI to move robotic limbs much better than a toddler.


1. Two transitions will allow China to thrive in AI: from discovery (research) to implementation (engineering), and from expertise to data.

2. Within 15 years, AI will be able to (able to, not will) replace 40-50 percent of American jobs

3. AI runs on large amounts of data, data is a winner take all problem market, Winner take all economies create wealth concentration in several giant market winners.

4. Mass psychological loss of purpose is a threat of AI wealth aggregation.

5. Reducing returns to talent on the margin: After a certain point, talent is decreasingly relevant, and China is at the good enough point.

6. Four Pillars to AI Superpower-dom: “abundant data, tenacious entrepreneurs, well trained AI scientists, and a supportive policy environment.”

7. AI is moving from breakthrough, by a “handful of elite researchers” to application, by an “army of tinkerers.” AI will be diffused throughout society, like electricity, and the internet.

8. The “grid” approach to AI (commoditize and provide as a service) will benefit the “Seven Giants” of AI research: Google, Facebook, Amazon, Microsoft, Baidu (search and software), Alibaba, and Tencent (Wechat). If one makes a key breakthrough, as with deep learning, it may pull ahead of the others significantly.

9. The “battery” approach to AI (using application-specific algorithms trained on particular data-sets) is the startup approach: become the endpoint provider for all things in industry or service X.

10. AI will more closely resemble internet technology than electricity or steam engines. It will not take advanced tasks to a low skill level. It will create high value jobs at the top end, but carve out many more at the middle and low skill ends: “it will simply take over the execution of tasks that meet two criteria: they can be optimized using data, and they do not require social interaction.”

11. AI adoption will be accelerated by three catalysts: algorithms can be instantly distributed around the world. Second, venture capital firms will be first in line to capture value to be unlocked by a burgeoning AI revolution. Last, Silicon Valley has not historically had competition, outside the valley. Now it does, in the form of Chinese entrepreneurs and technology companies.

12. “While AI-rich countries rake in astounding profits, countries that haven’t crossed a certain technological and economic threshold will find themselves slipping backward and falling farther behind.“ Low wage factory labor will be less and less an opportunity for developing economies to grow their industries.


1. PwC estimates AI deployment will add $15.7 trillion to global GDP by 2030. China will take around $7T, North America will take around $3.7T. Economic balances may continually shift in Chinese favor.

2. Chinese smartphone doubled between 2009 and 2013, from 233 million to 500 million.

3. In 2014, new Chinese subsidies created 6,600 new startup incubators more than quadrupling the overall total.

4. Chinese local government guiding funds quadrupled from $7 billion in 2013 to 27 billion in 2015.

5. “U.S. federal funding for math and computer science research amounts to less than half of Google’s own R& D budget.” Of the top one hundred AI researchers and engineers in the world, half of them work for Google.

6. That elite group in the United States has roughly doubled its share of national income between 1980 and 2016. By 2017, the top 1 percent of Americans possessed almost twice as much wealth as the bottom 90 percent combined.

7. We’re likely in for societal upheaval, creating the Harari “useless class”: “Rates of depression triple among those unemployed for six months, and people looking for work are twice as likely to commit suicide as the gainfully employed. Alcohol abuse and opioid overdoses both rise alongside unemployment rates, with some scholars attributing rising mortality rates among uneducated white Americans to declining economic outcomes, a phenomenon they call “deaths of despair.” ”


1. “To those steeped in the innovation mythology of Silicon Valley, the mini-iPhones were the perfect metaphor for Chinese technology during the copycat era: a shiny exterior that had been copied from America but a hollow shell that held nothing innovative or even functional.“

2. Hall of the Ancestors of the Ming Dynasty, and the copycat clocks from Jesuit missionaries from Europe, copied and perfected by Chinese craftsmen

3. “The moment Kaixin001 became a threat, the owner of Renren simply bought the original URL from its owner. He then recreated an exact copy of Kaixin001’ s user interface, changing only the color, and brazenly dubbed it “The Real Kaixin Net.” Suddenly, many users trying to sign up for the popular new social network found themselves unwittingly ensnared in Renren’s net.” Brutal.

4. “Mass electrification exemplified this process. Following Thomas Edison’s harnessing of electricity, the field rapidly shifted from invention to implementation. Thousands of engineers began tinkering with electricity, using it to power new devices and reorganize industrial processes. Those tinkerers didn’t have to break new ground like Edison. They just had to know enough about how electricity worked to turn its power into useful and profitable machines.”

5. The Solyndra story: Obama’s stimulus program guaranteed loans on renewable energy products, to stimulate growth in a renewable energy economy. The Solyndra bankruptcy crisis in 2011 became fuel for political attack ads against Obama as a high-profile failure by the American government, stemming future initiatives in technology.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s