On the topic of US/China relations, themes reminiscent of the Cold War era almost immediately spring to mind. Yes, one can look at the geopolitical, economic, and societal disparities between the world’s two leading nations exclusively through the narrow lens of domination, and thereby reach the conclusion that head-on conflict is unavoidable. Both fiction and non-fiction writers have explored that possibility at length, and technology always plays an essential part in these deliberations.
Taking into account the two countries’ vastly different historic and cultural backgrounds, how do their approaches to innovation, entrepreneurship, and government style stack up against each other? And how do these issues affect their respective potential to leverage and influence what’s arguably the single most transformative technology of our times: Artificial Intelligence?
Dr. Kai-Fu Lee, born in Taiwan and socialized in the US, has a background in AI research, has worked for Apple, led Google’s efforts in China, helped create Microsoft Research Asia, and these days runs his own venture capital fund which is focused on Chinese tech startups. Armed with this unique cultural and professional experience, in his 2018 book AI Superpowers he attempts to provide an informed, honest, and comprehensive assessment of the state of AI in both countries, as well as an outlook on how things are likely to play out in the future.
Taking stock
Undoubtedly, the US has led the world in AI research over the past 30 years or so. Groundbreaking innovations as deep learning, backpropagation, and transfer learning emerged there, and not in China or elsewhere, because of its culture’s unique appreciation of out-of-the-box-thinking and exceptionalism, supercharged with well-funded high quality research institutions. On that fertile ground, outstanding individuals such as Geoffrey Hinton, Demis Hassabis, and Yann LeCun were able to develop foundational technologies that made prominent applications such as Google’s AlphaGo and AlphaZero possible in the first place.
However, Kai-Fu Lee argues that concluding that the US therefore has an uncatchable advantage in the AI space would be short-sighted. Going forward, what will matter more than single, disruptive instances of breakthroughs in research will be an ability to apply and scale the existing technologies in the real world. And for that, a country doesn’t need to be home to the world’s top scientists, but rather to thousands of well-trained engineers and hungry entrepreneurs who can bring those innovations to life in practical applications.
The Chinese advantage
But how well is China positioned to do that? After all, many still associate the country first and foremost with cheap, low-quality imitations of physical products and not with innovative digital business models, disruptive ideas or an abundance of entrepreneurial spirit. The author however presents a solid case for why our intuitive assessment doesn’t hold up under close scrutiny.
While it’s true that even in the digital era, Chinese companies started out by copying proven American business models (and often imitating US software products nearly pixel-for-pixel), this copycat approach sparked an interesting dynamic among Chinese startups. At the height of the “War of a Thousand Groupons” for example, not just one company was trying to bring the idea of group-buying to Chinas one billion consumers, but thousands of them fought intensely over the attention of the public. This fiercely competitive environment, which the author refers to as “the coliseum” for the brutal practices that the contestants often applied, unsurprisingly forced countless companies out of business. But the survivors had learned valuable lessons on how to tailor business models to the unique needs of Chinese customers. These learnings then directly translated into unique and innovative approaches in the future. At the same time, Chinese engineers had gathered valuable experience in building highly scalable software solutions. The fact that most Chinese consumers never owned a computer, but instead became accustomed to digital services exclusively using smartphones, furthermore fostered mobile-first thinking and turned into a breeding-ground for online-to-offline models that took much longer to emerge in the west.
In the author’s view, battle-hardened entrepreneurs and engineers are only one piece of the puzzle of Chinas AI advantage, though. Equally important is the fact that Chinese companies have access to the most critical resource for AI applications: Data.
Consider for example the implications of China’s rapid adoption of mobile payments. The country simply skipped the credit- and debit-card frenzy that the US experienced in the 1970s and 80s and arrived directly at a point where it’s simple, convenient, and free to transfer money from one person to the other by scanning a QR-code with a smartphone. The two leading providers, WeChat Pay and AliPay, handle billions of transactions per day and have replaced cash payments almost entirely, thus providing the platforms with unprecedented quantities of highly detailed sales data.
Furthermore, the data available to some Chinese companies is much more comprehensive and extends deeply into the physical world. Apps like WeChat, which in addition to payments also serves as an ecosystem with plug-ins that cover everything from text messaging to ride-hailing, food-delivery, doctor’s appointments and restaurant reviews have access to a much more granular picture of how its users behave than anyone else, both online and offline.
Finally, one also has to acknowledge the huge role that the central Chinese government plays in all of this. When the “New Generation Artificial Intelligence Development Plan” was released in 2017, it didn’t cause much of a stir worldwide, but it triggered the reallocation of trillions of dollars within China. The city of Nanjing alone, for example, poured over $450 billion into its efforts to attract AI-based companies. Beijing remodeled the Zhongguancun area at great expense to bring VCs, incubators, and startups together in a single location. 60 miles south, the “Xiong’an New Area”, an entire city to house 2.5 million people, is currently being built from the ground up to accommodate self-driving vehicles, green energy, and other smart IoT technologies. As the example of high-speed rail visibly demonstrated to the world, when the Chinese government is determined to accomplish something big, there’s nothing that can stop it.
Scaling out
So far, Kai-Fu Lee has made his argument crystal clear: Experienced entrepreneurs, well-trained engineers with access to high-quality data, and billions and billions of dollars in public and private funding make up an ecosystem in which AI-based applications are bound to thrive. While all of these ingredients do exist in the US (and to a lesser extent in Europe) as well, the sheer scale of the Chinese operation is simply unprecedented.
Can we therefore expect that at some point, all of us will be chauffeured around by Chinese-manufactured self-driving cars which we’ll hail using Didi, while we’re reading the news on Toutiao, and texting our friends on WeChat? Again, the author warns, we shouldn’t jump to premature conclusions.
When the internet first took off in China, many US-based companies saw the country primarily as another market in which to position their existing offerings. Kai-Fu Lee reports from his own experience at Google China how reluctant the big Silicon Valley juggernauts were (and sometimes still are) to change even small aspects of their products to suit the specific needs of local customers. That experience, the author argues, has frustrated many Chinese IT executives and in turn shaped their own approach to global scaling of their now highly successful home-grown products. Instead of building a one-size-fits-all solution and using it to steamroller over the world, Chinese companies tend to follow a much more subtle and nuanced approach.
China’s ride-hailing giant Didi, for example, set a strategy of local partnerships and investment, rather than going straight into new markets itself. Tencent and Alibaba are both actively funding startups in India and other Southeast Asian countries instead of attempting to storm into these regions head-on. So, the author concludes, we can expect to see more Chinese influence in the future, an all-out take-over of Chinese businesses in locations away from their home turf is not that likely.
The future of AI
After the discussion of the strategic positions of the US and China with respect to AI, the book pivots pretty hard towards the general future of AI and humanity. The author’s own experience of surviving cancer made him rethink not just his approach to life, but what it means to be human in general, and how AI will go on to transform that notion.
Of course, no discussion of the future of AI would be complete without an excursus into Artificial General Intelligence (AGI), which Kai-Fu Lee thankfully keeps rather short. The more specific concerns about automation and projected lob losses due to the rise of narrow AI however get serious attention. Ultimately, the author argues, it doesn’t matter whether the US or China will dominate in the AI space, as long as we can find smart ways to utilize the disruptive potential of the technology in order to increase everyone’s quality of life and manage to distribute the spoils somewhat fairly. Whilst the author remains critical of quick fixes, such as a universal basic income scheme, he uses a fair portion of the book to advance some more nuanced ideas of his own. On can expect that his upcoming book AI 2041 will explore these topics further.
Summary & opinion
Other critics have pointed out that AI Superpowers exaggerates Chinas capabilities, and that it downplays the US’ involvement in future technological advances. That impression of sino-centrism is further exacerbated by the fact the Europe doesn’t even register on the author’s map of AI superpowers. However, I find that criticism not entirely justified: First of all, I picked up this book particularly because I wanted to get Kai-Fu Lees insight into the somewhat obscure world of Chinese startups and venture capital, and it fully lived up to my expectations on that front. Being able to contrast that with what we’re already familiar with in terms of US and European endeavors should be squarely within each readers own ability. The fact that, apart from DeepMind which originated in London, European companies don’t play a more prominent role in the book is definitely no fault on the part auf the author, but rather caused by the undeniable reality that European companies simply haven’t played that big of a role in AI to begin with.
The closing part of the book, over which the author tries to grapple with the overall consequences of AI on the future of humanity, didn’t do so much for me though. In my opinion, the text would have benefitted from more focus on the geopolitical, and maybe even military, impact we can expect to see from the increasing application of autonomous technologies by the two superpowers of our day and age.
If you’re looking for deeper discussions on the ethical, moral, and philosophical impact of both narrow and general AI, I can recommend a few books which are dedicated to these topics and therefore are able to provide a more nuanced picture. These range from classics such as The Second Machine Age by Erik Brynjolfsson and Andrew McAfee to Architects of Intelligence by Martin Ford, Künstliche Intelligenz und der Sinn des Lebens by the German philospher Richard David Precht, Life 3.0 by Max Tegmark, and even Conversations on Consciousness by Susan Blackmore. But if you’re in for unique insights into the Chinese AI community, as well as a first-hand account of the history of the countries’ rapidly developing VC and startup ecosystem, AI Superpowers is absolutely worth reading.