中国人工智能技术暴露度的测算及其对劳动需求的影响——基于大语言模型的新证据

Abstract

The Large Language Models (LLM) and generative Artificial Intelligence (AI) are developing rapidly. Using 1.25 million online job posts published between January 2018 and May 2024, this paper constructs the LLM-AI exposure index for each occupation based on the detailed descriptions of job tasks, and estimates the impact on labor demand. The study finds that the average AI exposure for new jobs in the Chinese labor market shows a decreasing trend during the sample period; occupations with higher exposure are mainly white-collar occupations with relatively higher education requirements and higher salaries, such as accountant, editor, salesperson and programmer. Regression results based on instrumental variables at the occupation level show a negative relationship between exposure and labor demand, with high exposure occupations experiencing lower salary increases, larger within-occupation salary gap, and higher education and work experience requirements. This paper highlights the need to strengthen the adaptability of China's labor market to new technologies, and suggests that the threshold of AI application should be lowered to encourage enterprises and workers to adopt new technologies.