Following on from advice from a flatmate, I’ve decided I’ll be writing summaries of books I’ve just read — hopefully to flush out some of their ideas within my head.
Kai-Fu Lee is someone everyone should know more about. As a PhD student at Carnegie Mellon, Lee created the first speaker-independent continuous speech recognition system. As the second AI winter was settling in, Lee worked at companies such as Microsoft, Apple and even headed up Google China, witnessing the rise of Eastern and Western tech companies. After this, he headed up one of China’s most successful VC funds, Sinovation Ventures. Thus, when it comes to understanding both Silicon Valley and China’s own tech scene, fewer people have a greater understanding. Below I’ve listed my favourite points from the book.
China’s AI Craze
Opening with a comparison to the Cold War and the last superpower race, Lee recalls the significance of Deepmind’s AlphaGo victory over the reigning Go Champion. Lee argues, that whilst most of the West was impressed by Deepmind’s AI for winning at the board-game, it had very little impact on their lives. In comparison, Go is revered within Chinese culture, considered one of the four arts of a scholar, with star players having huge culture followings (think of Bobby Fisher but in the age of Twitter). Of the 100 million viewers for the Lee Sodol vs AlphaGo match, a staggering 60 million were Chinese.
Watching the world champion get beaten (4–1), woke up an entire nation to the idea that AI is finally intelligent enough to compete at human level task. Lee calls this ‘China’s Sputnik Moment’, which just as the Soviet Satellite galvanised America to the moon, will inspire China to be the next AI superpower. Since the monumental game, AI has captured the minds of thousands of Chinese citizens, from aspiring engineers to businessmen and entrepreneurs. The central party has implemented a series of AI innovation programmes including targeting businesses, innovation and research, with a national discourse on how AI can improve our lives. In comparison within the West, discourse has focused mostly on job destruction and privacy, whilst these are important factors, they do not lend themselves to quick implementations.
Age of Implementation
In looking at which countries will be leading AI technology within the coming years, Lee considers what expertise will be required. Lee argues that whilst discovery is important, it is invention which has the lasting impact on people’s lives. Take for example the light bulb, telephone or washing machine, whilst these technologies were only possible by Thomas Edison’s discovery of electricity, it is the 19th century European entrepreneurs which brought them to mass. And so, whilst Canadian, British and American researchers have laid fantastic groundwork with seminal discoveries for neural networks, it is the entrepreneurs of tomorrow that will incorporate them into our daily lives.
Discoveries are also less secretive than before. Unlike the last arms race (for nuclear fission), where research was expensive and difficult to replicate, Computer Science has always encouraged an open-source culture in which technology and discoveries should be available to all. Lee declares that the ‘Age of Discovery’ is over, whilst the ‘Age of Implementation’ is where the AI superpower of tomorrow is chosen. Whilst the West most certainly leads in terms of quality of AI researchers, this by no means ensures its position as the leading AI superpower
The Coliseum and the Copycats
Having established the importance of the entrepreneur in the age of implementation, Lee now works to compare the shiny knights of Silicon Valley against the gladiators created within China’s own ecosystem. During the first decade of this century, many of the West (myself included) disregarded China’s tech start ups as simply ‘copycats’ of their Western counterparts. Lee argues, we should not have been considering the products themselves but the people behind them. China’s perception of intellectual property vastly differs from the West, and copying a novel idea is acceptable. This forces companies to differentiate themselves by less defensible aspects (e.g delivery or efficiency). Take for example Groupon, which in China created a spin off of other 100+ copycats, dubbed ‘the war of a thousand groupons’. Competitors created negative campaigns, undercut, filed lawsuits and leveraged other technologies to stay ahead of the game. Chinese entrepreneurs work significantly more hours than their Silicon Valley counterparts, and are always contactable. Andrew Ng recalls being able to call for meetings on a Saturdays and everyone being at the office less than an hour later.
The now dominant tech giants, such as Tencent, AliBaba and Baidu, have earned their current market dominance in this coliseum of startups, and as such, have proven their capability of fighting tooth and neck to win market dominance. This is already showing in how Eastern and Western tech giants decide to challenge foreign markets. When considering the expansion of Ride Hail apps, Western countries such as Uber, have always seen foreign markets as cash cows, assuming that if they win within America, the rest of the world will follow. You can see this in their approach of a single application which works anywhere in the world — the same can be said for the Google homepage, which Lee (as head of Google China) remembers was not allowed to differ from the American site, although the UI made it ‘un-interesting’ to the Chinese market. In comparison, eastern companies such as Didi, have chosen to invest heavily in the local competitor (such as Lyft) and share AI expertise, instead of entering said markets themselves. Time will show which approach is more effective, however initial results from Google and Uber’s expansion into China suggest that the advantage lies with these homegrown companies.
The New World Order
Independent of whether China or America becomes the AI superpower of tomorrow, Lee sees the rest of the world will as losing out. Unlike previous technology breakthroughs, machine learning is only as good as the data provided, and thus lends itself to monopolies. Whichever company brings about the first of these technology advances will subsequently win the data of future users, allowing improvements to its service and capturing further customers. Whilst this will predominantly jeopardise the chances of companies from developed countries, such as the UK, France and Canada, the new world order is even less favourable on economically developing countries. Advancements within Robotics and Machine Learning will affect jobs at both the top and the bottom, in particular, traditional manufacturing and production. This effectively erodes the only advantage developing countries had — cheap labour. Without the traditional method for moving forward, it seems unlikely that developing countries will be able to sustain current populations, nevermind progressing to later stages of development. Lee’s worldview is clear from this, AI will consolidate power.
Conclusion
I’ve only included a handful of the points within the book, and there are significantly more, including a fantastic chapter on how being diagnosed with cancer made Lee re-evaluate human value. I strongly recommend the book to anyone interested in AI research or policy. Equally, whilst Lee sets a clear dichotomy for China vs the USA, some argue that actually framing the problem in this zero sum fashion is dangerous(see here).