A Comparative Study on the Diffusion Speed and Economic Effects of Artificial Intelligence and Traditional General-Purpose Technologies

In recent years, artificial intelligence (AI) and large language models have rapidly transformed production processes and social structures, whereas traditional general-purpose technologies (GPTs) such as steam power, electricity, and computers have historically had profound effects on productivity and employment. However, current studies provide limited comparisons of AI and traditional GPTs in diffusion, productivity, and employment, and systematic analysis of diffusion bottlenecks, labor adaptation, skill needs, and social costs remains scarce. This study investigates the features and mechanisms of AI and traditional GPTs in diffusion, productivity, and employment, thus examining differences in spatial and industry diffusion, long-term productivity effects, and substitution, compensation, as well as creation impacts on employment. The results reveal that traditional GPTs diffuse slowly with notable regional and industry differences, while AI achieves faster, more synchronized diffusion. Consequently, by driving a shift from scale- to quality-driven productivity growth, GPTs combine substitution, compensation, and creation effects to keep overall employment relatively stable in the short term.