China will lead the AI revolution

PwC estimates AI will add $15tr to global GPD by 2030 (45% for China, 23% for the US, and 11% for Europe)

Economic impact of AI driven by

  • Productivity gains from businesses automating processes, and augmenting their existing labour force with AI technologies (first phase of $6tr)
  • Increased consumer demand resulting from the availability of personalised and/or higher-quality AI-enhanced products and services (second phase of $9tr)

Distribution

  • US faster than China in first phase (increasing productivity gains), slower in second phase (increasing consumer demand for AI-enhanced products and services)
  • Europe has top AI talent but lacks VC ecosystem and large user bases to generate data

Deep learning powers two transitions that advantage China by minimizing its weaknesses and amplifying its strengths

From the age of discovery to the age of implementation

  • Minimizes the need for outside-the-box approaches to research questions
  • Amplifies the need for effective entrepreneurs

From the age of expertise to the age of data

  • Minimizes the need for elite researchers
  • Amplifies the value of massive amounts of data

Harnessing the power of AI requires four inputs; China leads in three out of four

Hungry entrepreneurs: China leads

  • China’s internet ecosystem is a coliseum where hundreds of copycat galdiators fought to the death: in a market where copying was the norm, they were forced to work harder and execute better than opponents
  • Silicon Valley culture
    • Innovation and elegance are the central values; there is a stigma around copying and this often leads to complacency as the first mover is ceded a market because others don’t want to be seen as unoriginal
    • Geekie-hippie hybrid ideology, techno-optimism: belief that every person and company can truly change the world through innovative thinking
    • Entrepreneurs are often the children of successful professionals, they grow up hearing “you can change the world”
    • This leads to mission-driven startups
  • Chinese culture
    • Cultural acceptance of imitation: rote memorization formed the core of Chinese education for millenia, rigorous copying of perfection was seen as the route to true mastery
    • Scarcity mentality: China was grinding poor very recently; people also have a sense of urgency to get rich, before the communist government changes the rules
    • Willingness to dive into any promising new industry: core motivation is getting rich (not fame or changing the world)
    • Entrepreneurs are often only children, products of the “One Child Policy”, they carry the expectation of two parents and four grandparents who invested all their hopes for a better life in this child; they grew up hearing about survival and a responsibility to earn money so they can take care of their parents
    • This leads to market-driven startups, perfect embodiment of the lean ideas
  • Recent history of Chinese entrepreneurship
    • Copycat era until around 2013: Chinese entrepreneurs copied successful American companies; this allowed them to learn the basics and served as a good starting point for a Lean Startup methodology
    • Then they tailored products and business models to local needs, and created new and better products
      • Alibaba (Jack Ma) started by copying eBay but then he instored free listings while eBay ridiculed him (“free is not a business model”), leading to success in a low-trust Chinese society
      • Americans treat search engines like the Yellow Pages (tool for finding a specific piece of information) while Chinese treat search engines like a shopping mall (a place to explore); Google took too long to adapt because they did not want to fork the code
      • Toutiao (ByteDance), a news platform powered by AI personalized recommendations, does not have its equivalent in the US
      • WeChat, the digital swiss-army knife, does not have its equivalent in the US
    • Chinese entrepreneurs are willing to get their hands dirty with the logistics of the real world to differentiate themselves (through economies of scale, subsidies, and efficiency at the grunt work), while Silicon valley companies like to stick to building clean digital platforms
      • Dianping (inspired by Yelp) pilots food delivery systems
      • Tujia (inspired by Airbnb) manages a large chunk of rental properties
      • Didi (inspired by Uber) buys gas stations and auto repair shops to service its fleet
  • Why Silicon valley companies failed in China
    • Not because of government controls
    • They marketed their existing products to Chinese users, while they should have tailored their products or created new ones to meet market demands; this is the one-size-fits-all approach they often take on foreign markets
    • They lost out on top talent: the most ambitious and talented young people join or start local companies (they know they would be considered “local hires” and their career perspectives would be limited)

Abundant data: China leads

  • China has a larger user base
  • Real-world data (the what, when, and where of physical purchases, meals, and transportation) available in China is much richer than online data (clicks, likes, or online purchases) available in the US
  • This data was generated by the willingness of Chinese entrepreneurs to get their hands dirty with the logistics of the real world, and mobile payments (WeChat, Alipay)

AI scientists/engineers: US leads but it’s changing

  • The US will likely continue to lead in research for the coming decade because of its capacity to attract and retain global top talent, although China is catching up
  • Two defining traits of the AI research community (openness and speed) make it easier for China to catch up
    • Openness stems from common goal of advancing the field and from the desire for objective metrics in competitions
    • The speed of improvements also drives researchers to instantly plant a flag on Arxiv
  • Chinese students have an incomparable work ethic
  • China can train a large number of top tier AI engineers

AI-friendly policy: China leads

  • For the past 30 years, Chinese leaders have practiced techno-utilitarianism: leverage technological upgrades to maximize broader social good while accepting that there will be downsides for certain individuals or industries; this allows China to make big bets on game-changing technologies and encourage faster adoption
  • In 2014 China encouraged mass entrepreneurship and innovation; it was the first endorsement of internet entrepreneurship by the government and it changed the country’s cultural zeitgeist
  • The plan sets objective metrics and implementation is left to ambitious local officials competing with each other and experimenting with different policies
    • Created the spaces and ecosystem: technology incubators, entrepreneurship zones, and rent rebates in neighborhoods reserved to entrepreneurs
    • Got the money: government-backed “guiding funds” to attract greater private VC
    • Local governments experimented with different policies and the best policies were adopted by all
  • This process was both somewhat inefficient and highly effective: brute-force a fundamental shift of the Chinese growth from manufacturing to innovation; short term overpaying (e.g incubators unoccupied) but greatly accelerated long term monumental upside
  • After AlphaGo in 2016 (China’s Sputnik moment), the Chinese central government announced its AI plan for 2030: become the world leader in AI, leading in theory, technology, and application; the plan again sets objective metrics and lets local officials implement

Four waves of AI

Internet AI: recommendation online, dominated by major internet companies (Google, YouTube, Baidu, Amazon, Alibaba, Facebook, Toutiao etc.) — ratio China to US of 5:5 in 2018, predicted 6:4 in 2023

Business AI: decision-making from structured data (banking, insurance, healthcare) — 1:9 in 2018, predicted 3:7 in 2023

  • US corporations already collect large amounts of data and store it in well-structured formats while Chinese companies don’t use enterprise software or standardized data storage as much and spend far less money on third-party consulting
  • China’s immature financial system and imbalanced healthcare system give it strong incentives to rethink consumer credit and medical care from scratch; not having lobbies from established players like in the US can be an advantage

Perception AI: simple applications of perception (online-merge-offline) powered by hardware (sensors) — 6:4 in 2018, predicted 8:2 in 2023

  • Online-merge-offline: bring together convenience of online world and rich sensory reality of offline world in retail, transportation, and at home
  • China’s data collection in public spaces and openness to experimentation give it a massive head start on implementation
  • China has the best manufacturing capabilities in the world in Shenzen: skilled industrial engineers who prototype quickly and build at scale at low cost

Autonomous AI: hard applications of perception (robots, drones, self-driving cars) requiring elite expertise and legislation — 1:9 in 2018, predicted 5:5 in 2023

  • If the primary bottleneck to deployment is technology, the US has the lead; if it’s policy, China has the lead

Chinese and US tech companies take different approaches to global markets: US companies enter themselves while China funds local startups

The AI revolution will lead to a job and inequality crisis

Why there will be a job and inequality crisis

Four recent general purpose technologies (GPT) increase productivity and fundamentally alter economic processes and social organizations

  • The steam engine and electricity
    • Powered first and second Industrial Revolutions (1760-1830 and 1870-1914)
    • Deskilling: factories took tasks that required high-skilled workers (e.g handcrafting textiles) and broke the work into simpler tasks that could be done by low-skilled workers (operating a steam-driven power loom); gave low-skilled workers a productive role in the industrial economy
  • Information and communication technology (computers and the internet)
    • Skill biased in favor of high-skilled workers
    • Along with globalization and the decline of labor unions, it contributed to rising inequality in the US; most of the benefits from the productivity gain were absorbed by the top 1%
  • AI
    • Skill biased in favor of high-skilled workers

The last transition in date greatly widened inequalities and AI will widen inequalities even more because of its winner-take-all economics

  • More data leads to better products, attracting more users, leading to more data
  • Corporate profits will explode: how profitable would Uber be without drivers? Apple without factory workers? Walmart without cashier, warehouse employees, and truck drivers?
  • The combination of data and cash attracts the top AI talent to the top companies

The transition to an AI-driven economy will be faster than the previous transitions because of three catalysts

  • AI products are digital: infinitely replicable and instantly distribuable
  • The VC industry that barely existed before the 1970s made the ICT revolution much faster than the steam engine and electricity revolutions
  • AI will be the first GPT in which China stands shoulder to shoulder with the West

AI can replace jobs that are asocial and optimization-based (low in creative and strategic thinking for cognitive labor, and low in dexterity in a structured environment for physical labor): 40-50% of jobs in the US

Physical automation of the 20th century mainly hit blue-collar workers, but AI automation will hit white-collar workers first: business AI is far easier than autonomous AI

AI will revolutionize manufacturing, removing the cheap labor ladder out of poverty for developing countries

A path to a solution

China will not be part of the solution in the short term

  • China’s tech elite do not fear a job and inequality crisis, they have witnessed technology improve the lives of all Chinese for the past forty years
  • Even among Chinese entrepreneurs who foresee a negative AI impact, they believe the government will take care of all the displaced workers
  • There is little discussion of the crisis in China and even less of proposed solutions

The long term solution is education and retraining, but redistribution is necessary in the short term

Universal Basic Income (UBI)

  • The Silicon Valley elite like the idea of a Universal Basic Income (UBI) because it is a simple, technical solution to an enormous and complex social problem of their own making
  • The unconditional nature of the transfer fits the highly individualistic, live-and-let-live libertarianism that undergirds much of Silicon Valley
  • UBI acts as a pain killer for the population hurt by AI and the tech elite feeling guilty about it
  • UBI does not provide meaning to displaced workers

Kai-Fu proposes a stipend conditional on participation in (currently undervalued) care, service, and educational activities