AI Superpowers: China, Silicon Valley, and the New World Order

 

A snazzy red jacket, an evocative title, and glowing 5-star reviews.  (Let's not forget the cover blurbs by the CEOs of Microsoft and O'Reilly Media, amongst other prominent tech figures.)  I have to say, when I eagerly began reading AI Superpowers, I was a little nervous - would this book live up to its hype?

Capture medium
Kai-Fu Lee - by SheilaShang [CC BY-SA 4.0], from Wikimedia Commons

Kai-Fu Lee

A Taiwanese American, Kai-Fu Lee came to the U.S. as a child in the 1970s to follow his older brother and gain an American education.  Lee would eventually earn his doctorate in computer science, becoming well versed in machine learning and applying his expertise to speech recognition projects at Apple.  From the late 90s to the late 2000s, he then pivoted to leadership roles at Microsoft and Google, ultimately becoming the president of Google China in 2005.

In 2009, Lee left Google and founded his own investment firm, now called Sinovation Ventures.  The investment money goes towards Chinese start-ups, where much of China's AI development is happening.  Lee realized this momentum early on and was motivated to provide leadership and funding for these new companies.

AI & China: Explained

If you, like me, have little or no prior knowledge of AI, rest assured. Lee begins his book with two brief histories: AI and Chinese technology.

Lee describes the sophisticated AI of today: deep learning, also known as neural networks - a kind of digital impersonation of a human brain.  Deep learning leverages vast amounts of data to study and recognize patterns, and these patterns can be as intricate as human faces or dog breeds.  This concept originated in the 1950s, but it took decades of research, faster hardware, and more data before it really came to fruition in the mid-2000s (p. 9).

China's technological success has also arrived in very recent decades and through mixed means.  Lee spends the entire second chapter, "Copycats in the Coliseum," analyzing the cause and effect of Chinese companies' reputation for copying American products (e.g. knockoff iPhones), not to mention each other's.  From this "coliseum" arose a powerhouse of entrepreneurs - whom he calls "gladiators" - who survived the storm of competition and are now primed to take on similar challenges in AI.

I found this topic somewhat awkward to read (at times, Lee verges on sounding like an apologist for copyright infringement).  Nonetheless, if you push on, you'll find some fascinating stories, like how WeChat, the Chinese super-app, popularized mobile payments one Chinese New Year (2014) by offering a "digital red envelope" service for people to exchange gifts (p. 60).  It was a creative and direct hit to Alibaba's Alipay service.

China's Advantage in AI

Lee's current estimate of the two countries' AI capabilities shows the U.S. with a lead (p. 136).  However, the running theme of the book is that China will surpass us in just five years or so.  He cites a number of reasons, for example:
  • Data - Chinese users generate an enormous amount of data.
  • Speed - Chinese entrepreneurs work very hard and very fast, having weathered the dog-eat-dog battle of copying each other
  • Culture - China (both entrepreneurs and government) may use tactics which its U.S. counterparts would avoid.  For example, China is more likely to devote significant funding to AI, accepting the risk of failure or waste in favor of longterm benefits. 
  • Government - China's government has made AI a priority, and local governments are motivated to implement it just like any other central government policy.
The four areas of AI to watch: Internet, Business, Perception, and Autonomous.

Lee's Warning, and a Call for Action

What is the biggest threat which AI poses to both countries?

From Lee's perspective, the biggest danger isn't robots ruling the world, or the more mundane term for it: "general AI."  Rather, his concern is for the economies of both the U.S. and China. He forecasts that AI will, in the next decade, be capable of replacing many jobs - so many that the change in society will be monumental, like the discovery of electricity.  In chapter 6, he cites several studies and his own experiences, all of which make a pretty reasonable case for concern.

Here you will find the surprise in the book: a very personal story.  Lee, in the prime of his career, was diagnosed with stage IV lymphoma and thought he had only a 50% chance of survival.  This led to a time of deep reflection, and he began to question the choices he had made.  Without giving it away, I will say that I found this section of the book incredibly moving, and I admire Lee for inserting this vulnerable story from his life into a book of this genre.

It was this turning point in his life which gave him an idea.  In the last chapters of the book, Lee describes the mainstream theories of how we can address the dangers of AI.  He then provides his alternative idea, one in which humans are reconciled with technology and economic damage is minimized.  Again, I won't give it away.  What I can say: I see some significant problems with Lee's concept, yet I also find it very interesting.

Overall

AI Superpowers is a fantastic book - intriguing and highly readable.  (Perhaps a little too polished; Lee acknowledges a co-writer, Matt Sheehan, which may explain it.)  I felt it was detailed enough that I learned a lot, yet not so heavy that I was overwhelmed.  If he writes a "sequel" or follow-up book, I plan on reading it.  4.5 stars.

Comments

  1. Great review. The book sounds very interesting. It seems like it graples with some very big and very important issues. I would also recamend Nick Bostrom’s SuperIntelligence. That is more about Hard AI (what sounds like Lee is calling General AI) but it does deal with the comming possible revolution in jobs and society that this book addresses.

    ReplyDelete
    Replies
    1. Cool, I'll have to look into that one. In some ways, Lee's book is cursory, covering more breadth than depth, but it's certainly increased my curiosity in the topic!

      Delete
  2. Kai-Fu showed up a lot in that Google book I just read. Google had a LOT of problems trying to adapt to Chinese culture, and likewise -- its coders would often "borrow" code from other companies to make their own work faster, for instance, and its managers thought nothing of buying gifts like iphones to help them woo government officials or other businesses. Given how pervasive AI already is online -- shopping, youtube, facebook, both in curating content and in filtering it out -- this is definitely something of interest.

    ReplyDelete
    Replies
    1. Yes, he talked about localization, and how big U.S. companies were trying to put a one-size-fits-all product in front of users with different expectations. He seemed to suggest that was a bigger problem for foreign companies than, say, censorship.

      Delete
  3. i think i'm afraid to read this book... the future seems so ominous in so many ways that i'd rather not think about it... back to my Victorian closet...

    ReplyDelete
    Replies
    1. Mudpuddle, I hear ya! The ending of this book is pretty positive, but lingering in the back of my mind is the gut feeling that humans will goof this up (as we have with so many things).

      Delete

Post a Comment

Popular posts from this blog

My First Classics Club List

Slowing Down with Tolkien, Lectio Divina