A few weeks back, I came across the book AI SUPERPOWERS by Chinese author KAI-FU LEE.  I think it’s about 250 pages that anyone, who works in the field of data analytics should read (or at least think about it). It’s one of those books that are best when you read them yourself. Therefore, I will try to keep my review a reasonable balance between teasing and the feeling that you already know all what the book should tell you.

AI SUPERPOWERS offers many points to think about. I personally counted at least 20 (!) thoughts that I realized that I haven’t thought about yet. However, before  outlining some of them, we should explain who the author is. Kai-Fu Lee is a Taiwanese man who worked for 35 years in the field of artificial intelligence. He started voice analytics for Apple, set up a Microsoft research center in Asia and, as CEO of Google of China faced the dilemma of establishing Google in a country that does not necessary celebrate its existence. He also manages venture capital funds to develop AI solutions. Kai-Fu Lee is a rare combination of experience with state-of-the-art AI approaches from Silicon Valley and a typical Asian “cautious overview” that does not accept simplifications as well as does not need adhere to cult of America is Great. He praises where he sees real mastery and pinpoints hardly pretending and unwarranted stereotypes.

The reason, I think you should read this book by yourself, is that among official lines of the text you will likely find your own inspirations (as it was with me). The book is a busy tree where everyone can choose “how much they sit” on each branch. However, in essence the book is a cocktail of three complementary streams (some of which you would not expect to appear judging by the very book title) :

The first stream (most in line with the name of the book) describes developments in the field of artificial intelligence. It contrasts how diverse the paths to the sophisticated analytics were for the US and China. Taiwan and Hong Kong Hong Kong have a bond with China, but their relationship is not, ehm, optimal. (I have a colleague from Hong Kong who narrates often about it in detail.) So, Kai-fu Lee’s position is not a pink ode on the Chinese model. Quite the contrary, it offers a very balanced view of where China stands in AI area and where lags behind US. As he had experienced both environments, his comparison is a valuable counterweight to the general propaganda both for and against China.

The second line is the author’s personal account of how (thanks to the cancer he managed to overcome) he changed his view on the direction in which artificial intelligence should go. The story of a seriously ill man, who at stair to possible end of his life completely alters way of thinking, is almost a cliché in our culture. But if you manage to be stay less cynical and you will close your eyes on this emotional aspect of the story and focus in reading this section rather on his conclusions, it becomes an inspirational reading.

The third stream of the book was a bit of a surprise to me. But a pleasant one. Being tricked by the title of the book, I did not expect author to try to extrapolate AI trends and describe what awaits us. The focus of this last part is similar to (for me the hilarious) SuperIntelligence book, so very inspiring read. However, as AI SUPERPOWERS came out later, it is already looking at some future aspects of  AI richer, with result of the first experiments (eg. with UBI) and, hence, in more specific narrative.

However, in order not to just scratch the surface of this masterpiece, let me offer you few specific inspirational ideas that this book has brought to me. I believe they might be the right “teaser” to actually read the whole book:

Copy-cat China. The book clearly and detail depicts that China has reached its peak in economic significance by copying foreign products. Therefore, author bluntly admits that in industrial production and design of material things, China is certainly not a ruling world power, but rather an “embarrassing copier”. However, the development of online services, AI and data analytics has gone through completely different story. As a result, the latest Chinese advancements in AI and digital services are a still bit in shade of “copy-cat sticker” of the past. But book clearly explains that it would foolish and outright dangerous for the external world to keep perception China in this mental illusion.

From cash directly to App-pay. In some parts of Africa, they have long lagged behind in building a fixed line network, so many areas have been cut off from the world. However, suddenly, with the advent of mobile networks, they could skip the landline stage and get access (to internet) directly via mobile network. A similar episode took place in China in the area of payments. In China, credit cards have never  properly settled as a form of payment. And when e-commerce was launched, the market has jumped directly on in-app payments like We-chat or Alibaba.

4 AI development forces. Like any effort, even the development of artificial intelligence has its own factors that can accelerate or hinder it. In the case of AI, the following 4 dimensions seem to be relevant: a) Computing power in the form of HW, b) sufficient human talent, c) volume and quality of data you have for training AI, d) business underpinnings for implementing developed solutions. At the same time, the extent to which these 4 factors are fulfilled by a particular country predetermines what role that country should take in applying the AI. I also used this knowledge to prepare an AI strategy for Slovakia, which construction of I had the honor to participate in.

The status of your phone’s battery. When predicting phenomena, you should use all the available inputs and be aware of whether you are not limiting the possibilities of AI by your very own prejudices. The book gives some great examples on this subject, most of which I liked how the usual state of your mobile battery is related to your discipline of paying off financial obligations. My regular readers know I’m a strong promoter of Feature engineering and data riddles, so I really enjoyed this part.

Probably this much, Your Honor. In many areas, AI will serve as a human counselor. Medicine is often discussed, but Justice is more of the Taboo so far. Artificial intelligence can also be helpful in this sensitive area without machines deciding on us. There are already systems that search within historical court records to detect false testimonies of witnesses, contrasting the information given in previous legal litigation. Moreover, AI can provide inputs to calibrate the severity of penalties for the same criminal acts (using scatter-plots between particular aggravating/attenuating circumstances and the length of the sentence periods to see if the proposed sentence is too strict or moderate).

Autonomous car(t)s. The discussion of self-propelled vehicles is primarily zooming on autonomous cars. However, there are much simpler implementations that are both not as dangerous and show much more of immediate, mass-use potential. These include shopping carts, for example. They could be programmed to follow you (and stop whenever you turn to fetch something) or even set themselves the fastest route through supermarket, depending on where your shopping lists items are located in store.

Hold on, I’m sending you a drone. The second implementation of self-propelled vehicles, which is simpler than cars, are flying things. No, this is not hype about drones, but really there is more space in the air and lower chance of collisions than on the road. We might not realize it, but the planes were equipped with autopilots sooner than cars. We also have pilot-less attack aircraft, but not unmanned tanks or warships. Therefore, one of the nearest ways of AI use will be unmanned rescue units that will be able to extinguish fire or rescue people even in the exposed terrain, without compromising the lives of the helicopter or aircraft crew.

O2O, the key to platform success. Online-To-Offline (O2O) is a concept where you start a service in the online environment but at its end there is material fulfillment in the physical world. Examples of such services are E-commerce, Uber or Markets that offer o2o products are more tangible to people than pure virtual services (self-learning courses, online software business). People feel the physical dimension of such a service. Therefore, we are also more willing to pay for such a service (such as pizza delivery), while services such as online tax advice are only slowly collecting their enthusiasts.

What is different this time? Past industrial revolutions are often used as an example of how mankind has dealt with the harsh changes in the labor market. Thus, the optimists say that even AI will not be a disaster for jobs (by the way, a few words why Kai-Fu Lee is not so optimistic on this note). This book brings one interesting twist to this issue, the Deskilling paradigm. When attentively reading the history, we find out that the jobs, that sprung after the industrial revolution, required of workers shallower knowledge of the matter then their alternatives before revolution (weaver vs. weaving mill operator, mathematician vs. man with a calculator, …). This phenomenon is called Deskilling. The important question remains whether we are ready to admit such a development for healthcare professionals or teachers. To put in one sentence: In the AI Industrial Revolution, the vocations at stake are those where the credibility of the profession is linked to the human factor.

Bigger surveillance, not weaker one. Due to the accumulation of data and some other factors, AI services have a greater tendency towards  monopoly than in other economy sectors. (anybody, Google?) It is, thus, important that the AI industries are subject to stronger rather than the weaker (antitrust) regulation than traditional industries. However, the states are lagging behind both in legislation as well as competency to steer them. There is no clear idea how to regulate services such as Facebook and public authority, at the same time, lack educated employees to supervise them first place. It feels almost like if there was no one with medical education in Health Care Supervision Office.

If you are interested in any of the topics, I encourage you to read the entire book. It’s really worth it. If you’re still wondering if it’s a good (time) investment, check out the Kai-Fu Lee’s video, where he talks about some parts of this book.

Publikované dňa 27. 4. 2019.