Artificial intelligence (AI), the counterpart of human intelligence, emerged as a concept in the 1950s. Since then, humanity’s quest to develop AI has undergone several stages. At first, AI development adopted a top-down design (e.g., early symbolism and expert systems), which prioritized feeding machines with as much knowledge as possible in advance. Over the past two decades, the prevailing paradigm of AI development gradually shifted to a bottom-up approach, which endows machines with the ability to learn and considers intelligence as a capability that exhibits adaptability to the environment through learning. Launched in 2022 as a large language model (LLM), ChatGPT marks a milestone in the development of AI neural networks to acquire human-like learning capabilities, and has sparked an AI boom across the world.

Introduced in the 2024 Government Work Report, China’s “AI Plus” initiative[1] aims to foster the fundamental integration of AI technologies into various industries, which is consistent with global trends in AI development and closely related to the country’s industrial upgrading. As a symbol of the government’s tremendous attention to AI, the “AI Plus” initiative lays out the policy framework to propel China’s transition from the “internet age” to the “AI age”. This is a crucial aspect of the development of new quality productive forces, in our view. 

While new breakthroughs in AI development impact the economy and all of society, it is important to note that technological progress results from economic activities. The prospect of AI development hinges on the evolution of the socio-economic environment, including public policies. To explore characteristics of the latest advancements in AI as a productive force and examine its impacts on relations of production from an economic perspective, the CICC Global Institute (CGI) and CICC Research collaborated to produce this in-depth report to offer a systematic analysis of AI’s macro implications, industrial impacts, and governance challenges.

As AI is a general-purpose technology, a prominent feature of the ongoing AI advancement is the scaling law[2], which implies competitive advantages for large countries from a static perspective, and for first-mover countries from a dynamic perspective. While the US has gained first-mover advantages in the R&D of large-scale AI models, China may leverage its huge markets and enormous population to catch up at an accelerating pace. In particular, China is likely to become the cradle of pioneering innovations in model applications, which may provide a fresh impetus to the country’s economic growth. We estimate that AI may lift China’s 2035 GDP by 9.8% from our baseline scenario, which is equivalent to an addition of 0.8 percentage point to the country’s annualized GDP growth over 2024–2035.

A technological revolution not only boosts productive forces but also reshapes relations of production. With human-like attributes, AI may have profound and far-reaching implications for digital governance, market competition, social ethics, and international relations. History shows that the boost from technological advancement to economic growth comes along with the exacerbation of income disparity. Therefore, improvements to the social security system are essential to sustainable development. Such improvements are made possible by supply growth in the economy thanks to technological progress. In this context, China needs to take precautions in AI governance and endeavor to optimize its social security system so as to ensure both efficiency and equity, i.e., the sharing of benefits from technological progress among all people. In view of insufficient aggregate demand in China’s economy, adopting expansionary fiscal policies can not only boost economic growth but also help the country to accelerate AI development and catch up with frontrunners.

[1] Source: https://www.gov.cn/yaowen/liebiao/202403/content_6939153.htm

[2] Kaplan, J., McCandlish, S., Henighan, T., Brown, T.B., Chess, B., Child, R., Gray, S., Radford, A., Wu, J. and Amodei, D., 2020. Scaling laws for neural language models. arXiv preprint arXiv:2001.08361.