AI Superpowers: A Short(ish) Review

A couple of weeks ago, I attended a talk by the Chinese technology executive and venture capitalist Kai-Fu Lee, where Lee was promoting his book AI Superpowers. Formerly the world’s leading researcher on AI speech recognition, Lee went on to work at Apple, Microsoft and Google — where he led Google China — before founding the VC fund Sinovation Ventures.

AI Superpowers is a fascinating read. It begins by offering a brief history of the field of artificial intelligence to support Lee’s claim that we are not living through an age of discovery in AI. To explain this, Lee highlights the distinction between discovery and implementation and claims that many of the milestones being reached in AI today (for example, Google’s AlphaGo beating the best human Go player) are in fact the application of past decades’ breakthroughs to new problems. So while the recent AI achievements are impressive, they do not signal a breakthrough in our ability to develop a general AI of the sort generally imagined by Hollywood.

The central message of Lee’s book is that there are only two credible AI superpowers in the world today, America and China. Lee’s view is that, of the two, China is best placed to dominate. To explain why, Lee describes what he sees as the requirements to becoming a true AI superpower in the 21st century:

1. Abundant data
2. Tenacious entrepreneurs
3. Well trained (but not necessarily elite) AI scientists
4. A supportive policy environment

If data is the new oil, China is the new Saudi Arabia. Partly this is simply down to scale. Remember that China has 100 cities with populations of greater than one million people and is generating more data than any other country on the planet. This gives Chinese companies the largest (and therefore the best) data sets to train their algorithms. There’s no data like more data.

Given that we are in the implementation phase rather than the discovery phase, Lee maintains that China is also better placed than the US to win the AI race as China has a larger supply of scrappy entrepreneurs with sharp instincts for building robust businesses. Lee uses the example of the Chinese Groupon copycats here and it is really instructive, both on what it takes to be a successful entrepreneur in China but also on the cultural background of Chinese entrepreneurs and how that informs the differences between Chinese and American startups.

For the final two requirements, Lee argues that in the deployment phase, quantity of solid AI engineers will be rewarded rather than the quality of a few elite engineers, given that no new landmark breakthroughs are required to deliver on most AI applications. He also argues that China’s government provides a supportive policy environment, one which Lee contends can be somewhat inefficient but also extremely effective.

If you believe that it’s important to understand the Chinese AI innovation engine and what it means for the pace and direction of AI development — and I firmly do — then AI Superpowers is a very worthwhile read.