OECD（2021）和世界银行（2025）曾指出：AI对就业的影响还不明确、无法预估。从某种意义上讲，量化AI对就业的影响是一个悖论。它在操作上面临 着无法切割、难以界定和不可预判三个层层递进的难题。 一、三大不足 技术对就业影响的研究是一门显学。随着生成式人工智能蓬勃兴起，全球又掀起了一波新的研究浪潮。OECD、IMF、世界经济论坛、联合国贸发会议、 国际劳工组织、世界银行等国际组织，高盛、麦肯锡、皮尤研究中心等咨询机构，纷纷推出报告。如下表所示。 | 机构 | 发布时间 | 核心结论 | | --- | --- | --- | | | 2023.3 | 美国约三分之二的工作岗位面临一定程度的AI 自动化风险 | | 高盛 | | (two-thirds of US occupations are exposed to some degree of | | | | automation by AI ) | | | 2017.1 | 现有的半数工作活动将在 2055年被自动化(half of today's work activities could be automated by 2055 > | | 麦肯 ...

Core Viewpoint - The article discusses the impact of artificial intelligence (AI) on employment, highlighting the ongoing debate and confusion surrounding the quantification of AI's effects on jobs, as well as the limitations and challenges in measuring these impacts [3][5][11]. Group 1: Research Findings on AI and Employment - Various international organizations and consulting firms have published reports on AI's impact on jobs, with findings indicating that a significant portion of jobs are at risk of automation. For instance, the OECD states that 27% of jobs in its member countries are at high risk of automation, while the IMF estimates that nearly 40% of global employment is exposed to AI [4][5]. - The reports show a wide range of estimates regarding job exposure to AI, with figures varying from 0.4% to 67%, indicating a lack of comparability and consistency among studies [5][6]. - The concept of "AI Occupation Exposure" is often misunderstood, leading to unnecessary panic about job losses, as high exposure does not necessarily equate to job elimination [5][6]. Group 2: Challenges in Quantifying AI's Impact - The quantification of AI's impact on employment faces three main challenges: the inability to isolate AI as an independent factor, the difficulty in clearly defining the scope of AI, and the unpredictability of future technological developments [8][9][10]. - AI's influence on employment is intertwined with various macroeconomic factors, making it challenging to isolate its effects in a meaningful way [8]. - The dynamic nature of AI and its integration into various sectors complicates the ability to define its impact clearly, as AI is often embedded in existing technologies and applications [9]. Group 3: Limitations of Data in Employment Studies - Data used in employment studies can be influenced by subjective factors and may not always reflect objective reality, leading to potential biases in the findings [12]. - The pursuit of accurate data is often hindered by practical challenges, such as funding and sampling issues, which can result in distorted outcomes [12]. - The inherent limitations of data mean that predictions about the future labor market based solely on past data are often unreliable, as unforeseen changes can significantly alter employment landscapes [12].