Scan barcode
A review by jwsg
AI Superpowers: China, Silicon Valley, and the New World Order by Kai-Fu Lee
4.0
In this brilliant, accessibly written book, Lee Kai Fu systematically makes the argument that China will pull ahead of the US in the AI stakes. Assumptions that in the realm of AI, China will remain a copycat and a laggard reflect a "fundamental misunderstanding on what is driving the ongoing AI revolution", as we move from "the age of discovery, to the age of implementation, and from the age of expertise to the age of data".
Lee argues that we are now in the "age of implementation", where the breakthroughs we read about stem from the application of deep learning and related technologies to new problems, rather than fundamentally new technological leaps. In the age of implementation, it is less about having a strong core of elite researchers, and more about having talented entrepreneurs, strong engineers and product managers to apply the technologies in novel ways to create value. Which China has in abundance.
And in the age of implementation, having data to train the accuracy of AI algorithms is critical. Which again, China surpasses the US in because of the way its eco-system has been set up, with WeChat and other apps creating, capturing and consolidating "oceans" of new data about the real world, by digitising millions of offline transactions - not only food delivery and transport as it has in the West, but also getting manicures, booking medical appointments, getting massages, pet care. As Lee notes: "Silicon Valley juggernauts are amassing data from your activity on their platforms, but that data concentrates heavily in your online behaviour, such as searches made, photos uploaded, YouTube videos watched, and posts "liked". Chinese companies are instead gathering data from the real world: the what, when and where of physical purchases, meals, makeovers, and transportation. Deep learning can only optimise what it can "see" by way of data, and China's physically grounded technology ecosystem gives these algorithms many more eyes into the content of our daily lives".
Lee explains the various dimensions of AI clearly and accessibly. The part I found particularly illuminating was in Chapter 5 on The Four Waves of AI - Internet AI (centred on using AI algorithms as recommendation engines); Business AI (the mining of an organisation's data to train algorithms to outperform human decision makers e.g. fraud detection, loan applications, medical diagnoses); Perception AI (based on taking in information about our lived environment via sensors and smart devices - sounds, objects, temperature etc - and having algorithms make sense of this info in much the same way that our brain does); and Autonomous AI (the integration and culmination of the three preceding waves of AI to yield machines that don't just understand the world around them but can shape it). He points out how the newer waves of AI will dramatically reshape our understanding of what is online and offline. Lee asks: when you order a meal just by speaking a sentence from your couch, are you online or offline? When your refrigerator at home tells your shopping cart at the store that you're out of milk, are you moving through a physical world or a digital one?
Looking across these four waves of AI, Lee assesses where the balance of power currently lies between the US and China and where it is likely to go. On Internet AI, the US and China are pretty evenly matched but Lee predicts that 5 years from now, China will have a slight edge given its much larger pool of internet users and how deeply embedded digital services are in their lives. For business AI, Lee notes that the US is miles ahead of China given that its companies already collect large amounts of data and store it in structured formats and use enterprise software for various corporate functions. This makes it easy to apply business AI solutions to maximise profits and reduce costs. Chinese companies, by contrast, do not use enterprise software and have relatively poor data. Lee believes that China will close the gap somewhat in 5 years, particularly in the areas of public services and industries with a potential to leapfrog outdated systems (e.g. financial services) but the US will still have an edge overall. On perception AI, Lee notes that China has a slight edge over the US but in 5 years, China will pull far ahead given the massive amounts of data created by Chinese consumers and Shenzhen's position as a manufacturing hub for intelligent hardware. Finally, on autonomous AI, Lee notes that the US is miles ahead of China currently but predicts that they will be evenly matched in 5 years. He argues that "predicting which country takes the lead in autonomous AI largely comes down to one main question: will the primary bottleneck to full deployment be one of technology or policy"; if it is a question of technology, the US (Google's Waymo) is far ahead of the competition. If it is an issue of policy adaptation, then China is at an advantage.
The twist in AI Superpowers is that while it starts off centred in tech, it ends off with a call to reconnect with what makes us human and gives meaning to life - loving and being loved by family and close friends, creating meaning through the relationships and communities we are a part of. A cancer scare forces Lee to reevaluate his priorities and he argues that to harness AI's potential to generate prosperity for humankind, and to mitigate its downsides (crushing inequality), we need to create a new blueprint for development, one that "embraces our essential humanity". Lee argues that the traditional policy proposals to retrain workers, or redistribute income are helpful but insufficient given the speed of disruption posed by AI (and that this disruption is continual, which means people might need to continually retrain to remain employable) while redistribution is essentially a form of "sedation" or "numbing" - of the pain felt by those displaced and of the guilt felt by technologists for displacing others. He instead advocates the creation of a "social investment stipend", a sort of salary paid to those who invest their time and energy in "those activities that promote a kind, compassionate and creative society" e.g. through community service, education, care giving activities. It's an intriguing proposal, although I wondered how this might lead to unintended consequences as warned by Michael Sandel in The Moral Limits of Markets, where he talks about money corrupting the things that are priced. Might paying people to engage in community service corrupt that activity? or is this fear unfounded, given that there are many individuals who are paid to do similar roles e.g. social workers, teachers, and who are driven by a deep sense of mission and purpose?
A fascinating and thought provoking book.
Lee argues that we are now in the "age of implementation", where the breakthroughs we read about stem from the application of deep learning and related technologies to new problems, rather than fundamentally new technological leaps. In the age of implementation, it is less about having a strong core of elite researchers, and more about having talented entrepreneurs, strong engineers and product managers to apply the technologies in novel ways to create value. Which China has in abundance.
And in the age of implementation, having data to train the accuracy of AI algorithms is critical. Which again, China surpasses the US in because of the way its eco-system has been set up, with WeChat and other apps creating, capturing and consolidating "oceans" of new data about the real world, by digitising millions of offline transactions - not only food delivery and transport as it has in the West, but also getting manicures, booking medical appointments, getting massages, pet care. As Lee notes: "Silicon Valley juggernauts are amassing data from your activity on their platforms, but that data concentrates heavily in your online behaviour, such as searches made, photos uploaded, YouTube videos watched, and posts "liked". Chinese companies are instead gathering data from the real world: the what, when and where of physical purchases, meals, makeovers, and transportation. Deep learning can only optimise what it can "see" by way of data, and China's physically grounded technology ecosystem gives these algorithms many more eyes into the content of our daily lives".
Lee explains the various dimensions of AI clearly and accessibly. The part I found particularly illuminating was in Chapter 5 on The Four Waves of AI - Internet AI (centred on using AI algorithms as recommendation engines); Business AI (the mining of an organisation's data to train algorithms to outperform human decision makers e.g. fraud detection, loan applications, medical diagnoses); Perception AI (based on taking in information about our lived environment via sensors and smart devices - sounds, objects, temperature etc - and having algorithms make sense of this info in much the same way that our brain does); and Autonomous AI (the integration and culmination of the three preceding waves of AI to yield machines that don't just understand the world around them but can shape it). He points out how the newer waves of AI will dramatically reshape our understanding of what is online and offline. Lee asks: when you order a meal just by speaking a sentence from your couch, are you online or offline? When your refrigerator at home tells your shopping cart at the store that you're out of milk, are you moving through a physical world or a digital one?
Looking across these four waves of AI, Lee assesses where the balance of power currently lies between the US and China and where it is likely to go. On Internet AI, the US and China are pretty evenly matched but Lee predicts that 5 years from now, China will have a slight edge given its much larger pool of internet users and how deeply embedded digital services are in their lives. For business AI, Lee notes that the US is miles ahead of China given that its companies already collect large amounts of data and store it in structured formats and use enterprise software for various corporate functions. This makes it easy to apply business AI solutions to maximise profits and reduce costs. Chinese companies, by contrast, do not use enterprise software and have relatively poor data. Lee believes that China will close the gap somewhat in 5 years, particularly in the areas of public services and industries with a potential to leapfrog outdated systems (e.g. financial services) but the US will still have an edge overall. On perception AI, Lee notes that China has a slight edge over the US but in 5 years, China will pull far ahead given the massive amounts of data created by Chinese consumers and Shenzhen's position as a manufacturing hub for intelligent hardware. Finally, on autonomous AI, Lee notes that the US is miles ahead of China currently but predicts that they will be evenly matched in 5 years. He argues that "predicting which country takes the lead in autonomous AI largely comes down to one main question: will the primary bottleneck to full deployment be one of technology or policy"; if it is a question of technology, the US (Google's Waymo) is far ahead of the competition. If it is an issue of policy adaptation, then China is at an advantage.
The twist in AI Superpowers is that while it starts off centred in tech, it ends off with a call to reconnect with what makes us human and gives meaning to life - loving and being loved by family and close friends, creating meaning through the relationships and communities we are a part of. A cancer scare forces Lee to reevaluate his priorities and he argues that to harness AI's potential to generate prosperity for humankind, and to mitigate its downsides (crushing inequality), we need to create a new blueprint for development, one that "embraces our essential humanity". Lee argues that the traditional policy proposals to retrain workers, or redistribute income are helpful but insufficient given the speed of disruption posed by AI (and that this disruption is continual, which means people might need to continually retrain to remain employable) while redistribution is essentially a form of "sedation" or "numbing" - of the pain felt by those displaced and of the guilt felt by technologists for displacing others. He instead advocates the creation of a "social investment stipend", a sort of salary paid to those who invest their time and energy in "those activities that promote a kind, compassionate and creative society" e.g. through community service, education, care giving activities. It's an intriguing proposal, although I wondered how this might lead to unintended consequences as warned by Michael Sandel in The Moral Limits of Markets, where he talks about money corrupting the things that are priced. Might paying people to engage in community service corrupt that activity? or is this fear unfounded, given that there are many individuals who are paid to do similar roles e.g. social workers, teachers, and who are driven by a deep sense of mission and purpose?
A fascinating and thought provoking book.