I first heard of Dr. Kai-Fu Lee at the Forbes Under 30 event in Boston. He gave a keynote speech promoting his recently published book AI Superpowers. During the speech, I bought a copy of his book and soon after, AI Superpowers became one of my favorite reads in recent memory. Written by AI pioneer Dr. Kai-fu Lee, AI Superpowers examines the history of AI, the competitive overview between China and the US, and predicts AI’s future.

One reason I found this book to be so fascinating is Dr. Lee’s unique insight and qualifications on the subject of AI. Dr. Lee was an AI researcher at the elite Carnegie Mellon program in the ‘80s followed by various executive roles at technology companies including a stint as President of Google China. Lee also has the unique perspective of running Sinovation Ventures, a VC shop with a successful portfolio of US and Chinese based tech startups (including 5 Chinese unicorns).

His experiences give him expertise with AI technologies as well as a fascinating perspective on AI policies and cultural attitudes held by the US and China. Given the complexities in a field such as AI, AI Superpowers does a tremendous job in presenting a clear and accessible read that’s eye-opening for anyone not close to the domain.  

The Building Blocks of AI

At this point, it’s increasingly rare to find a person or industry where AI as a buzzword hasn’t already permeated.

AI has gone from an esoteric concept found exclusively in research labs and science fiction novels, to an ever-growing presence embedded in your smartphone, powering the decisions of your bank, and penetrating new industries at a rapid pace.

One of the building blocks of AI is deep learning (DL), a sub-discipline of machine learning (ML). A gross oversimplification of DL is that it involves creating algorithms and then feeding them large data sets to train a system to do something to optimize outcomes.

AI Superpowers discusses three dependencies for DL to be successful:

  1. Data: the new “oil”. DL requires huge datasets to optimize outcomes. Some of the largest datasets include those gathered by internet giants such as Google, Amazon, and Facebook.
  2. Computing power: crunching vast tomes of data requires significant computing power. Once limited to government and academia but with advancements in chips and cloud infrastructure, the barrier to entry has dramatically diminished.
  3. Technical Expertise: finally, technical expertise is still needed, although not necessarily at the caliber it once demanded. Dr. Lee makes the comparison of implementing AI today to electricity. After the initial breakthrough of electricity, the implementation was primarily relegated to talented (but not top talent) engineers to find and deliver specific use cases.

It’s worth noting that elite level of AI talent clusters with the 7 Internet Giants that are: Alphabet (Google’s parent company), Amazon, Facebook, Microsoft, Baidu, Alibaba, and Tencent.

Breakthrough level discoveries usually only happen every few decades, and there’s fierce speculation whether the next significant development will occur by a corporation or in academia. Depending on where it happens could have dramatic outcomes for how it’s implemented (open or closed source).

AI Superpowers Competitive Landscape  

After a brief overview of how AI works, and today’s major players, Dr. Lee explores the competitive landscape between China and the US.

This was my favorite section of the book that explores the competitive landscape between the current AI superpowers of China and the US. I found the Chinese perspective to be especially fascinating.

For example, there’s a common perception that Chinese researchers and entrepreneurs generally don’t have original ideas compared to their American counterparts. AI Superpowers takes this one step further by highlighting several examples of successful Chinese entrepreneurs who copied US platforms including Facebook and Groupon, but through fierce (even cutthroat) competition, ended up developing a superior product tailored to the Chinese market. In the US where it’s frowned upon to copy IP, and original ideas are held in high value, this becomes a competitive disadvantage.

While major breakthroughs are still more likely to come from the West (for the time being), the implementation of practical AI applications is exploding in China. The access to massive data, huge investments by central and provenvial Chinese governments, and a significantly laxer cultural attitude towards privacy expectations, all translate to giving China a competitive advantage over the US.  

For example:

  • The scale of China’s mobile payment adoption is significantly ahead of the US. Because of this, China has unique access to massive data sets that capture what Dr. Lee conceptually refers to as O2O (online to offline). China is well positioned to understand behaviors and preferences that occur in the physical world and not just what a user looks at or clicks online.
  • The central nature of China’s government results in massive investments and subsidies. While the West often views this as ineffective and wasteful, the Chinese perspective is that even though waste is sure to occur, there will be a handful of outcomes that justify the waste. In the US, there is a great deal of skepticism, and political risk that occur anytime the Feds get involved (e.g., TARP, the Solyndra “scandal”).
  • Western tech firms, especially as young startups, tend to be mission-driven. They want to change the world and have a meaningful impact on society. Chinese startups tend to be much more focused on making money. This results in breakneck speeds of iteration and an emphasis on localization while US firms tend to be more concerned with waiting until the product is “perfect” and avoiding the complexities and risks of managing multiple forked bases across user bases.
  • Chinese cities like Shenzhen have huge homegrown ecosystems for hardware capabilities. While there are still low-tech, high labor factories in China, the majority have undergone a high-tech transformation over the years. This translates into a competitive advantage that allows Chinese startups to rapidly prototype and iterate software and hardware while firms in the US are left to wait for physical components to be created and shipped.

Net-net: while the US has been the undisputed AI heavyweight over the decades, China is rapidly closing the gap.

One of Dr. Lee’s more sobering predictions is that China will surpass the US in AI capabilities by 2030.

4 Waves of AI

One of the ways AI Superpower frames AI is in its four distinct waves.

  1. Internet AI (exists today): this type of AI is already widely used today and responsible for recommending content based on behaviors such as clicking a link or liking a video. Currently, capabilities are relatively balanced between China and the US with Dr. Lee predicting a slight Chinese advantage by 2023.
  2. Business AI  (exists today): AI loan decisions, automation of accounting and capacity forecasting functions are just a few examples of business AI. At its core, AI’s 2nd wave leverages DL and other types of AI technologies to automate and optimize routine business operations. Given the maturity of US corporations and their vast data warehouses, the US is predicted to continue leading in this space for the foreseeable future.
  3. Perception AI (on the cusp of mass adoption): perception AI deals with extending AI capabilities to the physical world. Although it’s not quite ready for prime time, examples of perception AI include autonomous driving, smart cities, and even types of AR/VR powered by AI.  
  4. Autonomous AI (still a few years off): the final wave is a general purpose of autonomous AI. In this final phase, AI is let loose and able to intelligently optimize and respond to variables not explicitly accounted for in its programming. This is the stuff of nightmares or utopia depending on how it plays out.

What’s Next?

The final section of the book addresses Dr. Lee’s warnings and hopes for a future where AI compliments the best aspects of what it means to be human.

Who knows what will happen over the next few decades as AI becomes more capable and autonomous, and those outcomes will be directly impacted depending on which governments and corporations gain the competitive edge as breakthroughs occur. 

From Elon Musk’s dire warnings of AI destroying humanity to the Ray Kurzweil’s of the world who predict AI will usher in unimaginable wealth and utopia, there’s no clear-cut answer to how it will all play out. All we can do is buckle up and enjoy the ride.

You can learn more about the book and Dr. Lee here.