2025年生成式人工智能与就业报告

Updates: The report refines the ILO’s 2023 Global Index, incorporating recent advancements in Generative AI (GenAI). It uses tasks from Poland's 6-digit occupational classification (≈30,000 tasks) and surveys 1,640 respondents per ISCO-08 1-digit group to assess automation potential for 2,861 tasks.

Methodology: Tasks were scored using a blend of Large Language Models (LLMs) and human experts through iterative surveys. Scores were adjusted by aggregating survey, expert, and AI-arbitrated inputs, leveraging two independent AI predictors to ensure consistency across tasks.

Exposure Gradients: Occupations are classified into four gradients based on GenAI exposure (scores 0–1), task variability, and standard deviation. Gradient 4 (highest exposure) includes roles like clerical workers (data entry, accounting clerks) and some financial operators.

Global Estimates: One in four global workers (≈24%) is in an exposed occupation, rising to 34% in High-Income Countries (HICs). Female employment is more affected (up from 7.8% in 2023). The score reduction for clerical tasks reflects updated GenAI capabilities.

Key Findings: While automation risks exist, jobs consist of bundles of tasks requiring human input. Social dialogue and adapting policies are essential to manage transformations. The study highlights a need for human-centered technology integration and skills development.