(Yicai) July 18 -- As a new-generation general-purpose technology, artificial intelligence is prompting economists to rethink the fundamental logic behind growth models, labor markets and wealth distribution.

At the center of the debate is a divide between technology optimists, who anticipate “exponential growth,” and labor economists, who warn of the potential “devaluation of human capital.”

Academics are engaged in a theoretical debate over whether AI can enable the macroeconomy to break away from its long-standing path of linear growth, with competing views based on different underlying microeconomic assumptions.

Traditional semi-endogenous growth models suggest that while global investment in research and development has continued to rise, diminishing marginal returns have kept the compound growth rate of per capita gross domestic product in developed economies at around 2 percent for a long time.

Growth Debate

Once AI’s reasoning and cognitive capabilities surpass a certain threshold, its role in economic models will shift from that of a passive auxiliary tool to a “synthetic R&D workforce” capable of autonomous iteration, said Anton Korinek, an economist at the University of Virginia. This could drive economic growth beyond its linear trajectory and accelerate.

By contrast, Stanford University economist Charles Jones challenges this view with his “weak-link” model, arguing that the ultimate ceiling for economic expansion will inevitably be determined by the most difficult-to-automate bottleneck tasks.

In labor markets, AI is expanding into cognitive and decision-making domains for the first time in a substantial way, raising concerns that the skill premiums enjoyed by purely cognitive workers could decline.

Nobel laureate economist Daron Acemoglu has proposed a “pro-worker AI” framework, distinguishing between technologies that simply automate existing tasks and those that create new tasks. The former extracts and commercializes the tacit knowledge of skilled professionals, potentially reducing labor’s share of income. The latter, however, can increase the economic value of expert judgment through human-machine collaboration.

For example, in precision engineering and medical diagnostics, AI can conduct initial screening of unstructured data, while professionals focus on higher-level reasoning and validation. Acemoglu argues that the current structural risks facing labor markets mainly stem from excessive capital investment in purely automation-driven technologies.

Economic Distortion

Meanwhile, extremely low marginal costs could undermine the effectiveness of traditional economic indicators such as GDP. If AI significantly lowers the production cost of software development, professional services and certain physical goods, the amount of money circulating in markets may shrink.

This could create a disconnect in macroeconomic statistics. While consumers enjoy substantially greater consumer surplus and public services become far more accessible, traditional economic indicators could mistakenly portray the economy as experiencing a deflationary decline.

Economists point out that the current tax system in most developed economies imposes higher payroll taxes on hired labor while offering accelerated depreciation on corporate capital investment. This asymmetry creates opportunities for firms to favor automation over human employment. They argue that a combination of policy measures, including tax reforms, stronger antitrust enforcement and clearer data ownership rules, is needed to address these distortions.

(Yang Yanqing is director of the Center for Education, Innovation and Sustainable Development at ShanghaiTech University and An Xu is an AI analyst.)

Editor: Kim Taylor