2025年生成式人工智能与就业报告（英）_76页_3mb

Report Summary: Generative AI and Jobs - A Refined Global Index of Occupational Exposure

Introduction

The International Labour Organization (ILO) updated its 2023 Global Index of Occupational Exposure to Generative AI (GenAI), incorporating technological advances in GenAI tools and user familiarity. The study uses a representative sample of tasks from Poland's 6-digit occupational classification system (29,753 tasks) and surveys of employed individuals, combining human assessments with AI predictions and expert validation to refine global employment estimates.

Methodology

• Refined Assessment: Built on 2023 methodology by expanding to Poland's detailed classification, including 29,753 tasks.
• Data Collection: Surveyed 1,640 employed individuals in Poland for 52,558 data points on task automation potential. Used Large Language Models (LLMs) like GPT-4, GPT-4o, and Gemini for initial predictions, followed by expert surveys and score adjustments.
• Exposure Framework: Developed four progressive gradients instead of two categories, accounting for task variability and automation potential.

Key Findings

• Task & Occupational Exposure:
  • Clerical occupations remain highly exposed, with slight score reductions due to reassessment.
  • Other occupations, such as financial analysts and web developers, show increased exposure due to GenAI's expanding capabilities.
  • Globally, 24% of employment falls into exposure gradients; High-income countries have higher exposure (34%).
• Gender and Income Disparities:
  • Female employment shows higher exposure (9.6% in highest gradient) compared to males (3.5%).
  • Exposure increases with income level, e.g., 11% in LICs vs. 34% in HICs.
• Global Estimates:
  • One in four workers is in an exposed occupation; 3.3% in high-exposure tasks.
  • Distributions vary by region and gender, with significant differences noted.

Conclusion

The study shows GenAI poses challenges for job transformation rather than full automation. Recommendations include ensuring human-centered technology adoption, supporting workforce adaptation, and using the index for targeted policy responses and social dialogue while acknowledging exposure does not imply immediate job loss.