Review of AI Superpowers: China, Silicon Valley, and the New World Order by Kai-Fu Lee

In "AI Superpowers: China, Silicon Valley, and the New World Order," Dr. Kai-Fu Lee, with his rich background in the tech world (including significant academic contributions to machine learning and executive experiences at Apple, SGI, Microsoft, and Google), delves into the complexities of AI and its impact on the global stage. He offers insights on how AI might automate jobs, and stresses the importance of data, comparing its value to oil in the digital age.

Lee also contrasts AI development and adoption strategies between China and the U.S., drawing on his experiences to highlight how cultural nuances influence technological success. His book prompts readers to consider the broader implications of AI beyond the workforce, suggesting a shift towards valuing human creativity and compassion.

...and if we don't, then we face the consequences of even higher wage inequality, at a scale and speed that we haven't seen before.

Data as the New Oil

Data is often hailed as the new oil, and its quality is crucial for advancing AI performance. When comparing Chinese and American societies, several factors influence their ability to collect valuable data:

Population Size: China has a significantly larger population, with widespread smartphone usage. This results in a vast volume of data, providing a substantial advantage for AI development.
Data Privacy Sensitivity: Dr. Kai-Fu Lee notes China's edge in AI due to its population's lesser concern for data privacy, prioritizing convenience. While Americans also value convenience, there's a greater apprehension about extensive government surveillance.
Quality of Data: The quality of data collected is vital. Lee observes that China's Business AI development is hindered by the lack of high-quality data, attributing this to Chinese companies' reluctance to adopt enterprise software and standardize data storage. Their unique, often non-scalable systems complicate data integration and cleaning processes.

Localization of Products

Part of what drove the success of Chinese internet startups was their copycat strategy of tailoring American products to fit the Chinese market. Dr. Kai-Fu Lee, drawing from his tenure at Google China, has seen how reluctance to modify core products for different cultural contexts can impede a company's growth in foreign markets.

Lee points out a common strategy among American tech companies: deploying large sales forces to penetrate new markets without adequately adapting their products to meet local needs. This oversight may stem from a lack of cultural awareness among tech leaders, who often overlook the profound impact of cultural nuances on product design and user experience.

It may have something to do with the individual backgrounds of tech leaders and founders - engineers aren’t necessarily aware of the cultural context and biases they have been raised in.
As someone with an Anthropology degree, differences in culture and the biases we inherit is all we ever talked about, and their implications on products can be huge - imagine the differences in UX you’d have to acknowledge if you knew that your users read in completely opposite directions than Americans are used to, or if the market operated mostly on a cash payment system because of limitations in their local banking industry.

It is understandable that up until this point, it would have been difficult for American tech companies to justify forking their codebase just to serve a new market.

Software engineers aren’t cheap, and it’s expensive to maintain a complex codebase, much less entertain the thought of forking to accommodate the needs of a fraction of your existing user base.
Historically, the high costs associated with software development and maintenance may have deterred American companies from significantly altering their products for new markets.

However, advancements in AI, like Claude 3 and GitHub Copilot, are changing the landscape. These tools not only assist in code development and error correction but also present an opportunity for more efficient and cost-effective product localization.

I wonder if the resistance to forking a codebase will start to diminish. Claude 3 has proven to be a high performance AI coder, and could be used to quickly develop working prototypes that can be tested on a local market to gain insights and refine the product experience.
Software engineers have already started using AI to find and correct errors in their code (in place of asking other developers questions on forums like Stack Overflow), streamlining and speeding up the process of maintaining code or deploying new features.

It’s possible these tools will make it possible for American tech companies to consider greater localization efforts. With AI's ability to streamline the adaptation process, American tech companies might find it increasingly feasible to cater to the unique demands of international markets, potentially reshaping their global expansion strategies.

Vertical Integration and AI Development

The distinct approach of Chinese startups, characterized by their comprehensive vertical integration, contrasts sharply with the more software-centric strategy of many American tech companies (excluding notable exceptions like Amazon and Apple).

This 'heavy' involvement in every aspect of their business—from managing delivery teams to controlling payment processes—allows these companies to collect a diverse array of user data, from shopping habits to location and transportation patterns.
Such data richness, structured and well-labeled, is invaluable for training sophisticated AI models.

Dr. Kai-Fu Lee highlights a significant advantage of AI over human cognition: the capacity to consider thousands of minor, or "weak," features in decision-making processes that would overwhelm the human brain.

For example, the speed at which someone types or their phone's battery life might seem irrelevant to a loan officer but can provide predictive insights when analyzed by AI.
This approach has enabled AI-powered finance startups like Smart Finance to outperform traditional banks significantly, leveraging such weak features to inform their lending decisions with impressive accuracy.

This ability to utilize vast amounts of varied data and recognize the importance of seemingly trivial details underscores a potential for vertically integrated businesses to advance AI development significantly. By capturing a broader spectrum of user behaviors and preferences, these companies might not only refine their AI models but also enhance their competitive edge in a rapidly evolving digital economy.

Automation and Job Displacement

There have been many estimates on how many jobs are at risk to AI replacement - some ranging from 9% and others closer to 40%. Dr. Kai-Fu Lee estimates that 40-50% of US jobs could be automated by the 2030s, a projection I find quite realistic. The significant impact will likely stem from the comprehensive transformation of entire industries by new technologies, rather than the automation of individual tasks within existing jobs.

After all, when was the last time you called a travel agent to book a hotel or a flight?

There are still some areas that AI will be slower to adapt to - Moravec's Paradox states that while AI excels at complex cognitive tasks, it struggles with basic perceptual and motor skills that come naturally to humans, including infants. AI is more like a clumsy savant.

However, recent demos like Figure’s robot demonstrate rapid progress in these areas.

It's particularly noteworthy how Chinese companies might advance in robotic AI, driven by their quick adoption of perceptual AI technologies (the ability for AI to see and interpret the world around them), and the hardware mecca of Shenzhen.

Thanks to manufacturing outsourcing, Shenzhen has become an ecosystem that allows Chinese startups to develop rapid, inexpensive hardware prototypes (while creating barriers to foreign companies due to the usual challenges of tax implications, language barriers, visa issues and distance from headquarters).
Both the large volume of data collected for perceptual AI and the rapid hardware prototyping environment in Shenzhen could increase the rate of learning for Chinese robotic startups.

This region’s capacity for swift, cost-effective hardware prototyping, combined with the extensive data generated from perceptual AI, could potentially favor China for pioneering developments in robotic AI.

Human Value Beyond Economic Contribution

The tail end of Lee's book focuses on the implications on society and evaluating the popular Universal Basic Income position of technocrats. He's doubtful that UBI alone is enough to address the deeper philosophical issue of the social contract - the one in which our human worth is judged by our economic contributions.

It's fair to say that there is a tendency for society to label us by our jobs, but I do think there's been recent shifts in this mindset:

Gen Z vocalizing online that they 'do not dream of labor,' and embracing concepts like "quiet quitting" to seek a healthier work-life balance.
"Revenge bedtime procrastination" has emerged as a way for people to reclaim leisure time, staying up late to enjoy personal moments after a day filled with work and responsibilities. It's a signal for a desire for more unproductive time and enjoyment.
The FIRE movement (Financial Independence, Retire Early) gains traction among those striving for a life that values personal time and experiences over traditional employment's economic rewards.

These trends illustrate a broader movement towards finding value and meaning beyond the confines of work, challenging the notion that our worth is tied solely to our economic contributions. These voices may provide greater support for UBI and ease concerns over its negative impacts on our mental health and well being.

So, what now?

Ultimately, what does it mean for us? Well, as Alan Kay said, "The best way to predict the future is to invent it." I personally could see a world in which AI can automate away the job of a Product Manager, or at least the position within a company, and possibly allowing just a few individuals at the top of the organization to make a majority of the decisions that they're too busy to make today.

For that matter, once a Product Manager role can be automated, most of us knowledge workers may be out of a job anyway. I'd like for an AI to try and predict all the tasks and situations I find myself dealing with on a regular basis - each day and week is a usually a new challenge, given the ever changing business environment and multitude of factors that affect the role of the PM.

But, given how quickly AI has been improving, it will probably happen one day (exactly when is up to debate). At my individual level, what I can do is to learn about this space and try to be prepared for the change rather than be surprised and supplanted by it.

Should we be worried or excited about the future?

Even after reading this book, I don't think anyone knows for sure - not even tech leaders like Dr. Lee. But there are variables that are within our control - will we have discussions about preparing the future workforce for AI, or try to avoid them (because it's unpleasant to think about)? Will we allow ourselves to understand new technology rather than ignore it?

I think the first step for many of us is just being willing to engage with the discussion.

Like this post? Let's stay in touch!

Learn with me as I dive into AI and Product Leadership, and how to build and grow impactful products from 0 to 1 and beyond.

Follow or connect with me on LinkedIn: Muxin Li