The useful thing about this book is a perspective. Almost everything written about AI in English describes a world centred on a handful of American labs, with China appearing as a competitor in an abstract sense or as an example in a paragraph about surveillance. This is written by someone who ran a major American company's operation in China, then spent years as an investor there, and who describes that ecosystem from inside. Whatever the flaws, that vantage point is not available elsewhere in the popular literature.
The central argument is the strongest part and it has aged reasonably well. He distinguishes an age of discovery, in which value comes from research breakthroughs, from an age of implementation, in which the fundamental techniques are widely available and value comes from applying them to real markets with real data and real operations. If that transition is happening, then the advantages that matter change. Not the number of researchers you have, but the speed of your product iteration, the amount of data your services generate, and your willingness to build businesses that involve messy physical operations.
He argues that the Chinese ecosystem is better positioned on all three. The description of the competitive culture is the most interesting material and the part most likely to surprise a Western reader. The account of an entrepreneurial environment where copying is a starting point rather than a disgrace, where competition is fast and brutal, and where companies fight through subsidy wars over unglamorous businesses like food delivery and bike hire, is specific and detailed and it is not how the Chinese technology industry is normally portrayed. Whether you find it admirable or alarming, the description is informative.
The personal chapters are a genuine turn in the book. He was diagnosed with a serious cancer while writing, and the experience visibly changed the argument. The final section, on what human beings are for in an economy where machines do more of the work, and on care and connection as the things that cannot be automated, is more thoughtful than the technology sections and reads as sincere rather than constructed. It is also the part that has the least to do with the book's thesis.
Now the problems. The author is an investor. He runs a venture firm with positions in the Chinese AI sector, and the book describes that sector's prospects in glowing terms. That is not a hidden conflict, it is stated, and it is present in every assessment in the book.
A fund manager's account of why the market they are invested in will win should be read the way you would read any other prospectus. The 2018 publication date has hurt this book more than most. Since then the regulatory environment in China has changed dramatically, with major interventions in the technology sector, and the export control regime on semiconductors has become the central fact of the competition. Neither is in the book.
The confident assessment of who was positioned to win was made before the two developments that have most shaped the outcome. Surveillance gets a pass that I find hard to accept. Data abundance is presented consistently as a competitive advantage, and the conditions producing that abundance, including the extent of monitoring and the absence of the constraints that exist elsewhere, are treated as a matter of different cultural attitudes to privacy rather than as something to examine. For a book praising an ecosystem partly because of what it can collect, that is a substantial evasion.
And the employment predictions were specific, confident and have not aged well. Particular categories of work were named as being displaced on particular timelines. Some of that has happened in a partial way, some has not happened at all, and the actual pattern of disruption since has landed on jobs the book did not flag. This is the standard fate of technology forecasting and it is worth noting given how firmly the claims were made.
My three point three is for a genuinely valuable inside account of an ecosystem most readers know only in caricature, with a good central observation about implementation beating discovery, marked down for an author whose financial interests run right through the argument, for a date that puts it before the events that mattered most, and for a comfortable silence about what the data advantage actually rests on. Read it for the reporting. Do not read it for the forecast.