Over time, AI is expected to penetrate more industries and in a more profound way. While AI is reshaping the global labor market, it is important to understand the purpose of this reshaping for humans. How should people respond to job displacement? Where does the value of human work lie while AI can churn out a vast amount of knowledge? NewsChina talked to Cai Fang, an academician at the Chinese Academy of Social Sciences who has just published a book titled New Employment Trends in China – How AI Is Reshaping the Labor Market. Cai is a prestigious scholar on population and labor economics. He points out that in competing with AI in the job market, the only way is for people to leverage their advantages and achieve differentiated complementarity with AI.
NewsChina: How is AI impacting the labor market? Which occupations are most affected? Will this job displacement intensify in the future, and how should we respond?
Cai Fang: The development of AI, particularly the emergence of generative AI, has already caused significant shocks to the job market. Surveys in the US show that the launch of ChatGPT has noticeably impacted certain white-collar positions, with a sharp decline in demand for new graduates. In the legal industry, for example, law graduates usually handle basic tasks such as drafting documents and researching case files – these are precisely the areas where AI demonstrates high productivity.
In fact, this wave of technological revolution driven by AI differs from previous ones. In the past, job displacement caused by new technologies primarily targeted repetitive manual labor or low-skilled, semi-skilled work. Generative AI, however, can replace technical roles that require higher educational backgrounds. Consequently, for many industries with technical barriers, AI has now reached a medium level of proficiency, making entry-level graduates particularly vulnerable to replacement. In the future, as AI evolves from large language models toward embodied intelligence, even more positions will risk being displaced.
The core advantage of generative AI lies in its strong cognitive capabilities. In this aspect, its level of intelligence has already surpassed that of humans. Therefore, the only path for humanity is to identify their comparative advantages and achieve differentiated complementarity with AI in the job market, rather than competing with it in areas where it excels. Ultimately, our goal is to realize human-AI collaboration and coexistence.
Computer scientist and roboticist Hans Moravec [a faculty member at Carnegie Mellon University] made a famous observation known as Moravec’s Paradox in the 1980s. He observed that tasks that are highly challenging for humans, such as becoming a world checkers champion, can be mastered with ease by AI, while activities that are easy for humans, such as navigating around furniture to hand someone a glass of water, remain difficult for AI. This insight reveals a fundamental non-overlap of capabilities between artificial and human intelligence. As unique living and intelligent beings, humans have accumulated a wealth of “tacit knowledge” or “practical wisdom” through evolution – capabilities that are inherently more “human.”
These abilities, which can only be understood intuitively, constitute our uniqueness, such as self-control, teamwork, social skills and artistic perception. Only humans shed tears over Shakespeare’s tragedies. In the era of AI, we need to delve deeper into these latent “innate powers” and amplify the human capital that complements AI. This is the key to human-machine coexistence in the future labor market.
In scenarios of human-machine coexistence, work should not be viewed merely as an abstract position. AI can deconstruct different tasks within traditional jobs, reshaping the entire production process and allowing human intelligence and AI to focus on their respective areas of expertise. In other words, human-machine collaboration holds the potential to create more possibilities for employment.
We need to understand that the impact of AI on employment is dualistic – it can both disrupt and create jobs. When addressing the employment challenges posed by AI, the core strategy should be to ensure that job creation outweighs job destruction. This underscores the importance of AI alignment, which refers to guiding the development of AI to “serve humanity,” aligning it with human goals, preferences and ethical principles. Achieving this requires not only technical adjustments at the model level but also consensus among technology developers, investors, entrepreneurs, users and so on, to establish common orientations and action priorities in developing AI.
NC: How should education adapt to achieve the complementarity between human intelligence and AI, from the angle of workforce development?
CF: The impact of AI on employment has increasingly exposed the limitations of the current human capital cultivation model focusing on extending years of education. Talent cultivation in the new era should not rely entirely on schooling. Instead, it must shift toward a sustainable, career-spanning approach that covers the entire life cycle of workers. This means placing greater emphasis on early childhood development, expanding compulsory education to include both preschool and high school education and strengthening mechanisms for lifelong learning.
Preschool education is particularly important. Newborns retain more human “instincts” than adults do, whereas the latter in some ways are more like AI. According to research by US economist James Heckman, the first three years of life are the most critical period for cultivating one’s non-cognitive abilities, where capabilities and skills vital for a lifetime are largely developed during childhood.
Therefore, the goal of shifting the educational focus to preschool educational in the age of AI is not letting children study more academic knowledge but enhancing their comprehensive capacity to perceive the world, teaching them to recognize objects, discern colors and more. These early experiences are crucial. The government’s next step should be to deeply integrate preschool education resources, as making human capital investments at earlier educational stages yields higher returns.
The core challenge in establishing such a lifelong human capital development system lies in effectively coordinating resources. For example, in China, the management and services for children aged 0-3 and those aged 3-6 fall under the health and education systems respectively, which complicates the integration of preschool education resources. In fact, achieving intrinsic connections and effective transitions between different educational stages has long been a weak point in China’s education system. To meet future demand for lifelong learning, the government may need to comprehensively adjust and update the existing institutional arrangements, teaching content, goals and evaluation methods across all educational stages.
In addition, due to demographic changes, the number of students from kindergartens and primary schools to middle schools and high schools will gradually decline in the long term, leading to huge amounts of educational resources becoming moribund. This, in turn, provides an opportunity for resource coordination and integration. The government could consider using some of these disused resources for vocational training or for enhancing skills for more older workers. To sum up, in the age of AI, we must rethink what kind of education we truly need and what capabilities the talents we cultivate should possess.
NC: Besides resource coordination, providing lifelong training for the workforce means more investment. China still faces big downward economic pressure. Where does the money come from? How should the government make policies to cope with the short-term shock of AI on the labor market?
CF: From the perspective of the short-term impact of AI on the labor market, job destruction always arrives before job creation. There is research indicating an obvious time lag between those who lose their jobs due to technological change and those who gain new ones for the same reason. The occupational structure shift sometimes takes an entire generation to finish the transition. Therefore, ahead of the impact, the government must establish an effective safety net mechanism for every individual whose job is displaced. This, too, is a form of “alignment.”
In industries experiencing large-scale AI-driven job replacement, an “AI tax” could be introduced at an appropriate time. The revenue generated from such a tax could serve as a crucial funding source, allocated as transition-related funding for workers affected by AI, including compensation for job shifts and skills training. This would help mitigate the shocks brought about by the displacement. In a sense, in the era of AI, labor market-related safety net systems should become more inclusive. In the future, the government should provide lifelong education and training services for workers of all ages and shoulder the needed investment.
The logic behind an AI tax is, if the application of AI can indeed significantly enhance labor productivity and generate sustained economic value for relevant enterprises, such productivity gains should not be monopolized by individual companies. The government should facilitate the sharing of these gains across society through redistribution, which requires meticulous institutional arrangements. A productivity-sharing mechanism, on the one hand, can help reduce monopolies and a “winner-takes-all” scenario, a situation already emerging in the US. On the other hand, it can help mitigate the damage to economic structural stability due to the polarization of the labor market.
In 1930, British economist John Maynard Keynes predicted that in 100 years, labor productivity would inevitably increase four to eightfold, so people would only need to work 15 hours a week, no longer wasting most of their lives on dull and tedious labor. Many economists’ calculations show that the productivity gains Keynes envisioned have already been achieved. Yet today, people still work around 40 hours per week, not 15. The deeper logic behind this “involution” (sometimes referred to as rat race or excessive competition) calls for deeper reflection. From the perspective of labor economics, I believe that as AI brings about a leap in labor productivity, we urgently need to redefine work and reconstruct the relationship between compensation and labor.
Looking further into the future, especially when artificial general intelligence (AGI) truly arrives, future professions should focus on promoting the comprehensive development of individuals. By then, advances in AI will have the potential to break resource constraints and decouple returns on human capital from labor productivity. Although this process may take a long time, AI comes closer than any previous technological revolution in achieving this breakthrough.
NC: How will the development of AI affect the structural employment mismatches of an aging of population? What do you advise to address this mismatch?
CF: For the time being, China’s labor market has gradually come out of the periodic unemployment it suffered during the pandemic, but the headache of structural unemployment persists. Structural employment mismatches manifest in the coexistence of both labor and job shortages.
The popularity of involution and “lying flat” (social withdrawal) among young people are both manifestations of this structural mismatch. The core cause of this issue is the lack of effective matching between workers’ skills and market demands. What’s worse, China’s current demographic structure actually weakens our capacity to address the employment challenges posed by AI.
Young people and older workers must be the focus when addressing employment imbalances. The core vulnerability of young workers in the job market stems from the signiffcant impact of AI on entry-level positions. Therefore, the key to policy design lies in how to strengthen training and use AI to empower young workers, enabling them to quickly overcome the experience barrier and upgrade their human capital to an intermediate level as rapidly as possible.
The employment challenges faced by the aging workforce are not only their outdated skills but also that they might be excluded from the entire intelligent work environment, which will worsen their disadvantaged position in the job market caused by the digital divide. As a result, employment support for the elderly should emphasize not only skills training but also ensuring that AI technology itself is supportive and inclusive for them. Preferential technical solutions could be designed for this purpose.
As China’s population aging intensifies, the proportion of older adults among the workforce is set to rise continuously, making it crucial to enhance their actual labor force participation rate. To achieve this goal, public training resources should be made more accessible for them. Through precise skills assessments, they can see the gaps between their existing skills and market demands, thereby getting targeted improvement. Therefore, the key to addressing the structural imbalances in the labor market lies in promoting deep integration of population policies, employment policies and social security policies.
Old Version
