Book Review: Lee, K.-F., (2018), AI Superpowers, Houghton Mifflin Harcourt

This book focuses on China versus the US in the race to take the lead in AI. Why do the Chinese seem to be doing so well? And which particular strengths might Silicon Valley have to play on?

Few, if anyone, seem to be as well qualified to address these issues as author Dr. Kai-Fu Lee. As of present, he is Chairman/CEO of Sinovation Ventures, a leading technology-focused investment firm focused on developing the next generation of Chinese high-tech firms. Before this, Dr. Lee was the President of Google China. Earlier, he held executive positions at Microsoft and Apple. Dr. Lee received his Ph.D. from Carnegie Mellon University in 1986.

AI is now clearly at the forefront of our minds. Just a few years ago, AI was a much more narrow or limited field of interest, and largely confined to various research labs. But recent, major theoretical breakthrough have led to significant practical applications and huge consumer acceptance, consumption and attention. China, in particular, been particularly focused on AI. Dr. Lee points out that this does not seem to merely be an issue of hardware/software developments, but, perhaps even more importantly, how human beings, the Chinese versus US cultures, differ.

Dr. Lee points out that a clear outside threat to a nation’s superiority may be critical. The US faced this when the Soviets came up with advances in the space exploration field. This led to concerted US efforts, not only to take the lead, but also to reestablish dominance. Dr. Lee points out that perhaps a similar situation exists today when it comes to AI today for China – to catch up and take the lead! It is a national issue therefore, and a power issue.

It all seemingly started out with so-called deep learning. The US, backed by Canada and the UK, were more or less entirely alone in this “space”, then China eventually entered. Lee points out that Chinese internal entrepreneurs were attracted to areas of high competition, where they might mobilize speed, in particular. This led to a situation where Chinese entrepreneurs more or less constantly improve their products, building effective business models.

The Chinese government was highly supportive. There were no constraints from Peking standing in the way. For instance, the Chinese government did not impose any constraints due to potential job losses, say, within tracking, insurance, manufacturing or retail. And a “winner takes all” economic approach was adopted, even with the clear realization that much wealth would fall into the hands of some relatively few tycoons.

An abundance of data gave Chinese companies another clear advantage. In China there was no strict protection of personal data, such as what we find in most of the western world. While China had initially been seen as a copycat, this was no longer the case. Speed and data were now driving much of this country’s advances. The so-called copycat entrepreneurs were, of course, working hard, around the clock, and with speed.

As Dr. Lee sees it, Silicon Valley and China grew out of very different soils. In the former, most developers came from situations of wealth, seeing it as their mission to do better than the generation before them. In the latter, it was much more a matter of market-driven culture. For many years, the Chinese culture was based on three over-riding factors: to accept copycatting, to reorient into any promising new industry (i.e., extreme flexibility), and to advantageously build on China’s laissez faire inherent system. And it seemed to work!

Why did Silicon Valley giants fail in China? These companies mostly saw China as another export market. But the Chinese market needed more tailored products, specifically developed for its market. So, many large US firms failed (eBay, Google, Uber, Airbnb, LinkedIn, Amazon, …).

Chinese entrepreneurs were employing a wide array of “war”-tactics which Silicon Valley were not able/willing to adopt. They were faced with expensive investigations and distracting lawsuits.

As pointed out, Chinese startups were lean, highly adaptable, and fast. Their focus was to come up with a product in the market which could then be improved almost continuously, i.e., not wait, as in Silicon Valley, to release a more complete new product! Chinese entrepreneurs were more willing to “get their hands dirty” this way, and to build “clean” digital platforms was considered a luxury.

Another way to characterize the difference between the Chinese and Silicon Valley cultures, according to the author, is to see entrepreneurs in Silicon Valley as “going light”, in contrast to the Chinese “going heavy”. A light approach implies making relatively small improvements to aspects of AI platforms. Going heavy, in contrast, implies not just building AI platforms, but also being active in marketing, handling the goods that may be part of a given AI service, running the delivery team, taking charge of the payment, and so on, i.e., a much more heavily integrated value chain!

The author now discusses what he sees as the two remaining factors that distinguish China from Silicon Valley AI — expertise and governmental support (having already discussed the difference in speed, and the contrasting data abundance). When it comes to AI expertise, the author thinks that Google is way ahead of its US competitors. He points out that the essential issue of having superb data chips available in general seems to put the US ahead. For China to catch up here might be difficult - Silicon Valley remains a clear leader!

Now to the governmental policies. As already noted, the Chinese government assumes that there might be clear disruptions in some societal sectors. Apparently, the Chinese government sees this as inevitable, and not a dilemma!

There are four waves of AI, according to Dr. Lee. He labels the first as “internet AI”, using AI algorithms largely for recommendations about how to do things in different ways, hopefully better. The second wave is labeled “business AI”, where companies take advantage of AI to further improve the way they work. Now to the third wave, “perceptional AI”. This implies integrating AI with a proliferation of sensors and smart devices. More data needs to be gathered and analyzed for new optimal solutions to be achievable. China seems to be particularly strong when it comes to developing sensors and smart physical devices. The fourth wave is “autonomous AI”. This means the AI is used to support self-operating entities such as self-driving cars, self-functioning machines in factories, warehouses or on farms. We see, in this context, that Chinese automotive manufacturers have come a long way, relative to the overall global market leader Tesla.

Venturing to assess how the two entities, China and Silicon Valley, seem to compare today (2018) and in five years (2023), Dr. Lee sees China ahead in internet AI, remaining behind in business AI, well ahead in perceptional AI, and the two to be on par in autonomous AI. 

As we know, there are highly differing predictions about whether AI can eventually take off. Some, “utopians” according to Dr. Lee, see an excessively bright future ahead with AI leading to truly significant improvements in the way we live and work. Others, “dystopians”, see things differently. They make dizzying predictions about how AI could take over, leaving us humans on the sidelines. Dr. Lee comes down clearly on the side of utopians. But he proposes two so-called replacement graphs, one relating to cognitive labor, the other relating to physical labor, characterizing these spaces into a “safe zone”, a “danger zone”, “human veneer” and “slow creep”. Examples of safe jobs might be CEOs or elderly home caretakers. In contrast, assembly line inspectors or telemarketers seem to be in the danger zone. In general, firms may need to analyze how job losses could impact them when it comes to replacements. But there may also be entirely new job disruptions due to new AI empowered business models, entirely new ways of undertaking a given business. One might expect a high degree of turmoil in how these industries may be reshaped. And, for some, this could become intensely personal.

This leads us up to a blueprint for human co-existence with AI as Dr. Lee sees it. He sees it as critical that AI supports people in good ways, for instance, in making life easier for the elderly. To reduce loneliness might be particularly key. To proactively consider the benefits from AI is also important. If not, more inequality and even political instability may result. Lee sees the following issues of particular importance:

-       “Reduce”. This implies reducing working hours, and in such a way that no societal groups are left behind, taking account job losses in some work categories, discussed before.

-       “Retrain”. Workers that otherwise might be left behind must be retrained.

-       “Redistribution”. No groups come out as losers. Unemployed workers should be supported!

So, in summary, it is critically important to improve these three dysfunctionalities, and to guarantee a minimal basic universal income. Dr. Lee feels that broader social welfare should not be the anathema to the logic of private enterprises.

In general, the so-called AI race between China and the US might be dysfunctional. How might an AI future without an AI race become a reality? To rewrite our educational system is probably key. Drawing on the best from each country's culture might also be critical. Less government intervention could thus become a reality. And this could lead to a more natural way to adapt to the strengths of AI.

Dr.  Lee’s book is remarkable in several ways. Firstly, he pinpoints clear differences between China and the US— which are well done, thanks to his Chinese citizenship, and broad educational and work experiences in leading institutions and firms in the US. Lee has the lens to look critically at both nations. Importantly, Lee, also seems to recognize fundamental aspects of the social side of AI, including potential dysfunctionalities. He acknowledges that some types of social interventions may be needed— i.e., to stimulate education for those losing their jobs, guarantee minimum wages as “safety nets”, and establish certain types of regulation— in order to steer clear of an AI take-over threat. We humans cannot let that happen.