The deployment of large AI models across various industries is poised to drive economic growth with potentially far-reaching effects. Photo: TUCHONG

---

New breakthroughs in artificial intelligence (AI) are influencing socioeconomic development. Technological advancements are themselves a product of human economic activity. Therefore, AI’s future trajectory will in turn be shaped by the evolving socio-economic environment—including public policy.

General purpose technology & AGI

The scaling law, a hallmark of this round of AI progress, implies that, in static terms, larger nations hold inherent advantages, while dynamically, early movers benefit more. Historical experience shows that while technological progress fuels economic growth, it also tends to exacerbate income disparities. Consequently, strengthening social security has a material foundation and is essential for sustainable development. 

AI qualifies as a general purpose technology (GPT) in economics. [A GPT refers to a transformative innovation that shapes production methods and spurs invention sufficiently to have  a protracted aggregate impact.] AI has the potential to rival electricity, acting as a GPT that drives human development.

In terms of technological feasibility, recent breakthroughs in large AI models signal that the first inflection point of the S-curve has been surpassed. As technology iterates and application scales expand, the deployment of large AI models across various industries is poised to drive economic growth with potentially far-reaching effects. In the long term, the ultimate socio-economic impact of AI hinges on when and where the second inflection point on the S-curve emerges. A key question here is when will artificial general intelligence (AGI) be achieved. The scale effect is a key concept in analyzing the logic of both AI’s development and its economic implications.  

Scaling law & economies of scale 

The scaling law, sharing certain similarities with economies of scale, refers to systematic improvements in performance as large AI models are scaled in application. Though closely intertwined, the scaling law addresses technological feasibility, while economies of scale pertain to economic viability. 

The adoption of large AI models is not merely a technological issue; even when the marginal output decreases with expanded data scale, the economic feasibility remains intact if the benefits outweigh the costs. Economies of scale are a key factor in AI applications and industrial development, and can be both internal and external. Internally, AI’s economies of scale are evident when individual firms gain efficiency improvements from increasing the scale of their operations through large AI models. This non-linear effect means that developing large models requires a threshold level of resource investment. Coupled with economies of scope in application, large tech companies are better positioned to leverage internal economies of scale. External economies of scale of AI manifest in three aspects. First, the advancement of large AI models drives increased investment in new algorithmic frameworks, databases, and computing infrastructure, which helps lower the average costs of algorithms, data, and computing power. Second, external economies of scale are evident in the interactions and synergies between developers and users of AI models. Third, as AI technology spreads beyond the tech sector, businesses in other industries can integrate AI into their operations, creating industrial chains and ecosystems that foster collaboration both within and across sectors.

However, technological progress is endogenous, and even if the second inflection point is still distant, scaling AI may face constraints. From a broader socioeconomic perspective, one macro constraint is the need to address climate change. As energy consumption and carbon emissions become increasingly prominent in discussions surrounding AI’s development, the green transformation should be accelerated and the transition to green energy solutions to reduce dependence on fossil fuels should be promoted. In this regard, China can make significant contributions to global green transformation. 

Great Convergence & Great Divergence

Technological progress plays a crucial role in shaping the economic competitiveness and development of nations or regions. During the Industrial Revolution, Western economies experienced rapid growth, while Eastern countries, particularly China, gradually fell behind, resulting in profound shifts in the global economic and political landscape. This phenomenon is known as the Great Divergence in economic history. Following World War II, a few economies—mainly in East Asia—successfully caught up with developed economies during their industrialization, and China has achieved rapid economic growth since the reform and opening up. These phenomena are known as the Great Convergence. 

Will the current wave of AI development lead to further divergence or foster convergence? The key to this question lies in our understanding of technological progress and economies of scale. The scale effect indicates that the ongoing advancements in AI will benefit the two largest global economies (China and the United States).   

Humanoid robotics & replacing humans

Replacing and empowering labor are two primary roles of AI, but the nature and intensity of these effects differ from previous technological advancements. The replacement of physical and mental labor is not mutually exclusive; one significant potential of the current wave of AI progress is the integration of both, as seen in the development of embodied AI. Embodied AI represents a deep convergence of AI and robotics, with humanoid robots being a key application. Humanoid robots, which resemble humans in both appearance and function, are capable of performing certain tasks traditionally carried out by humans. The widespread adoption of humanoid robots faces not only technological feasibility challenges but also economic considerations. With design optimization and large-scale production driven by innovation, the production costs of humanoid robots are expected to decline over time. By combining the economies of scale in manufacturing with those in digital technologies, China holds unique advantages in developing humanoid robots. 

However, the development of humanoid robots is a global phenomenon. On one hand, humanoid robots offer new solutions to labor shortages caused by an aging population. On the other hand, the potential for widespread machine replacement raises concerns about large-scale unemployment, making technological unemployment a significant topic of discussion. In this context, the economic concept of Baumol’s Disease offers a useful framework for analysis. The theory suggests that resources, including labor, shift from sectors experiencing rapid efficiency improvements (with an oversupply of labor) to sectors where efficiency gains are slower and demand exceeds supply. Surplus labor from industries with fast technological advancements is eventually absorbed by industries with slower progress and unmet demand. This transition requires careful public policy guidance and may increase energy consumption, carbon emissions, and pollution, thereby intensifying the pressures on global green transformation efforts.

Efficiency & equity

While AI enhances production efficiency, it may also induce shifts in economic and employment structures. Economies of scale amplify the efficiency-enhancing effects of technological progress, but they could also widen income disparities and exacerbate the issue of insufficient aggregate demand. From the supply side, AI’s integration with various industries is expected to boost the potential economic growth rate. This round of technological progress has endowed AI with stronger generality, raising the total factor productivity  by both replacing and empowering labor. Given that AI technology’s penetration and widespread application take time, the improvements in production efficiency are likely to follow a pattern of slow initial gains, followed by more significant progress. 

At the macroeconomic level, a key prerequisite for AI to enhance economic efficiency is a substantial increase in capital investment, which also offers a perspective for identifying new economic growth drivers. In the long run, compared to supply-side improvements, the widespread use of AI technology will have a relatively smaller effect on boosting aggregate demand, potentially leading to a macro pattern of oversupply. 

In contrast to past technological advances, AI, as a “human-like” technology, may pose greater challenges to laborers in terms of income distribution. As a product of research and development, AI technology yields high returns, largely benefiting innovators and venture capitalists. The commercial application of AI, particularly the widespread adoption of humanoid robots, requires substantial investment. Capital deepening inevitably leads to a decline in the share of labor income within GDP. To address the inequality arising from AI-driven progress, fiscal policy can play a crucial role, ensuring that the benefits of technological advancement are shared across society. 

The economic boost generated by technological progress lays the foundation for improving the social security system. But how can the fiscal expenditure associated with enhancing social security be offset? At present, fiscal policy can raise funds through increased national bond issuance, rather than relying on taxes, to support improvements in social security and public services. In response to insufficient aggregate demand, expansionary fiscal policies can stimulate economic growth, providing a robust demand base and a relatively loose macroeconomic environment for technological innovation. This in turn will support China’s efforts to accelerate its progress in the AI sector. 

In the long run, technological progress and capital deepening will also necessitate reforms in the tax system. Increasing direct taxes, especially those related to property, while reducing indirect taxes, will be vital for  promoting equity.

Peng Wensheng is the dean of the China International Capital Corporation Limited (CICC) Global Institute.

Edited by CHEN MIRONG