人心猛于AI

One Human fear of AI will likely remain a major topic for a long time. Recently, the US research institution Citrini Research released a report titled "Global Intelligence Crisis 2028," which caused a stir online, plunging many into panic about AI and even triggering significant stock price declines for AI-related companies. What exactly did this report say? It emphasized that this was merely a "thought experiment" depicting an economic collapse triggered by AI development being "too successful." For example: AI efficiency is so high that intellectual capital is no longer scarce, making economic behaviors built on this foundation unsustainable; white-collar workers are replaced on a large scale, leading to loss of income, mortgage defaults, and financial market collapse; Corporate profits grow due to AI efficiency gains, but since a large number of people are unemployed and have no consumption, it creates a "Ghost GDP" phenomenon where "output grows but the consumption engine stalls"; Original business models also face huge challenges; some business models based on human laziness, information asymmetry, and brand dependence will collapse due to AI intervention, such as software services, intermediary platforms (food delivery, travel booking), payment processing (credit card interchange fees), and private credit, etc. Just days after this report was released, the US fintech company Block announced layoffs of approximately 4,000 people, reducing its workforce from 10,000 to 6,000. The reason given was that the company developed its own AI tool named "Goose," which significantly increased efficiency and replaced a substantial portion of employees. Previously, when companies carried out mass layoffs, they were almost always facing severe survival crises. However, Block's gross profit for the year reached $10.36 billion, up 17% year-over-year, representing a "profitable layoff." Wall Street expressed great appreciation for this move, and Block's stock price rose by over 24% at one point. Media outlets all exclaimed, "AI has begun to replace humans on a large scale!" Workers further fell into anxiety, believing there is not much time left for humans. In fact, as we wrote in our previous article ["The Truth Behind 'AI Causes 100,000 Silicon Valley Jobs Lost'"](https://mp.weixin.qq.com/mp/wappoc_appmsgcaptcha?poc_token=HIH6p2mj_Wv8Iqj4BRMybF-5Ggsz-oLBV7ZZ16RN&target_url=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzkwMzIxNDIyMA%3D%3D%26mid%3D2247493873%26idx%3D1%26sn%3Dad5d5c248c766b7a8a9d920d5bfc6c40%26scene%3D21#wechat_redirect), while tech giants' mass layoffs under the banner of embracing AI may indeed involve some replacements by AI, a significant portion cannot be blamed on AI. For instance, during 2020-2022, tech companies underwent massive expansion. During the pandemic, online demand surged, and the Federal Reserve lowered interest rates to near zero levels, drastically reducing corporate financing costs and directly prompting numerous tech companies to launch large-scale expansions. Consequently, Amazon's employee count doubled, Google and Microsoft both expanded by nearly 70,000 people each. At that time, domestic companies like ByteDance, Meituan, and Tencent also broke through the 100,000 mark, and Block expanded from nearly 4,000 to 12,000 people during this period. The short-term absorption of massive personnel was the main reason why tech companies launched waves of layoffs in the following years. According to data from the US employment information website Layoffs.fyi, tech companies globally laid off about 160,000 people in 2022, about 260,000 in 2023, close to 150,000 in 2024, and last year saw news like "100,000 Silicon Valley jobs cut." "Global Intelligence Crisis 2028" is more like a "feel-good story" that goes smoothly from start to finish with rapid upgrades, ignoring the complexity of technology's expansion in reality, as well as the squeezing and balancing effects of population, culture, economy, society, policy, and many other factors on technology expansion. Currently, the AI industry has not yet stepped out of the "Solow Paradox"—in 1987, computers had gradually become popular, but economist Robert Solow discovered that "you can see everywhere that we have entered the computer age, but you don't see any change in productivity statistics." That is, although computers have been widely applied, they have not brought significant improvements in productivity. It was not until the late 20th century, with changes such as Walmart improving inventory turnover by 40% through its "Retail Link" system and Dell revolutionizing its build-to-order model, that computer technology truly unleashed productivity. This shows that relying solely on new technologies is not enough to drive productivity growth; it also depends on organizational变革, business model innovation, policy environment, and a series of other factors. OECD estimates that in the next decade, labor productivity driven by AI will only increase by 0.4%-0.9%. Therefore, the power of "employment doomsday" will be constantly diluted by the lag caused by the "Solow Paradox" and will not arrive in the short term. However, an undeniable fact is that the impact and destruction of AI on employment will be unprecedented. Where is the way out for human employment? Two Recently, I read Economist Cai Fang's book "New Trends in China's Employment: How Artificial Intelligence Reshapes the Labor Market," which comprehensively and systematically analyzes this topic. From it, we can roughly summarize three possible paths for human employment in the AI era. First, the most ideal scenario is achieving "human-machine collaboration." Artificial intelligence is just a technology platform; it may destroy jobs through automation or create higher-productivity jobs by changing production processes. Which path AI takes depends on policy choices and institutional arrangements. Human capabilities include cognitive and non-cognitive abilities. The former includes things like arithmetic reasoning, vocabulary ability, text content understanding, mathematical ability, and coding speed, which are easy to measure and easily imitated by artificial intelligence. The latter includes motivation, self-control, adaptability, social skills, empathy, and compassion; these are tacit knowledge unique to humans who can do but not necessarily say. Therefore, humans and machines have complementary capabilities. Artificial intelligence does not need to replace employment as a necessary motive; machines can be used to enhance human capabilities. Humans tell machines what to do, thereby increasing service creativity and experience at a higher level, improving consumption quality and consumer surplus. If robots simply replace humans to batch-produce what humans produced before, they are just replicating industrial-era products faster, causing severe oversupply and ignoring changes in the consumer market and product innovation. Second, shift to positions with characteristics of "Baumol's Cost Disease." What is "Baumol's Cost Disease"? There are industries in the economy where productivity growth is particularly slow, and operating costs tend to rise in the long run, possessing high income elasticity of demand. Economist William Baumol originally studied the performing arts industry as an example, later expanding to many fields, such as healthcare, education, social work, sports and culture, entertainment, public administration, social security, and social organizations. Productivity in these fields is difficult to improve significantly like in digital industries or manufacturing; in fact, many rely on subsidies to survive. However, as society advances, demand in these areas continues to grow. For example, in recent years, theater fever and concert fever have emerged. Regarding healthcare, lives are at stake, and no one dares to say they will fully let AI take over; demand will only continue to rise. Education in the AI era is not competition between humans, but between humans and machines, requiring personalized teaching, holding considerable development potential. Generally, the higher the level of social productivity, the more it can support a larger number of low-productivity-growth "Baumol's Cost Disease" positions, absorbing a significant number of employment opportunities. Third, positions with characteristics of "Reverse Kuznetsization." There is a concept in economics called the "Kuznets Process," where labor transfers from low to high labor productivity, thereby driving industrial structure upgrading. "Reverse Kuznetsization" is the opposite: falling from high productivity to low, such as white-collar workers becoming unemployed and then driving ride-sharing cars or doing deliveries. If there exist a large number of "Reverse Kuznetsization" positions, it means the response to technological employment shocks has failed. A large number of unemployed people can only compete for low-productivity, low-wage positions. The productivity gains brought by technological progress are offset overall, which is also an important reason why the "Solow Paradox" appeared earlier. Another very important point is the productivity sharing mechanism in the AI era. For example, introducing an "AI Tax" as funding for retraining compensation, skills training, etc., for workers affected by AI. Expanding the supply of public goods is also an important means of income redistribution. Once public goods have a higher proportion than private goods in total social supply and demand, it will change the market price of labor factors. For example, housing, education, and medical care, which cost the public a lot, if they all become public goods, then even if one switches to a job with lower income than before, living standards will not decrease but rather improve. With the tremendous improvement in social productivity, implementing universal welfare may become possible. Current social security often distributes benefits only to those "qualifying" groups, providing only basic living guarantees, with significant differences between urban and rural areas. Universal welfare should be distributed without distinction, with higher guarantee levels that rise with productivity. Because in the AI era, the dilemma of worker employment does not stem from lack of effort or wrong choices; there is no need to distinguish distribution of social welfare anymore. Another very noteworthy point is that Cai Fang also mentioned in his book that facing the onslaught of AI, governments should establish a bottom line of public interest protecting workers and employment positions. They should not view themselves merely as third-party institutions playing a checking role, nor as neutral arbiters, because points of interest balance do not naturally exist. When major changes come, workers and employment positions are naturally the weaker side; applying tilted protection is consistent with social interests. We analyzed in our article ["Why Hard Work Alone Doesn't Lead to Salary Increases?"](https://mp.weixin.qq.com/mp/wappoc_appmsgcaptcha?poc_token=HM36p2mj7ma3PPPtj5-fgAEXG-LwJ9lZQd0cxdek&target_url=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzkwMzIxNDIyMA%3D%3D%26mid%3D2247494366%26idx%3D1%26sn%3D2a1c44d043c04f3f57669f4ced836501%26scene%3D21#wechat_redirect) how American workers gradually lost their bargaining chips for salary increases. During major upheavals such as the oil crisis, globalization, and technological progress, capital owners often use efficiency improvements, market changes, and natural results of technology as excuses to systematically transfer economic risks and uncertainties to ordinary workers, while capital owners gain more bargaining chips. This time, regarding the AI technological revolution, companies like Block attribute mass layoffs entirely to technological change. What they are开启 is not the AI era, but the "AI Doomsday." Therefore, "Human Hearts Are More Ferocious Than AI." Whether the negative effects of new technologies manifest or are curbed ultimately depends on human action.