Labor market impacts of AI: A new measure and early evidence (www.anthropic.com)

A recent study introduces a novel measure of AI displacement risk, termed "observed exposure," which assesses the actual risk of job displacement by evaluating the theoretical capabilities of large language models (LLMs) alongside their real-world usage. This measure weighs automated use more heavily and is designed to capture the degree to which AI can automate tasks within different occupations. The research highlights that, while AI's potential is significantly vast, actual implementation in the workforce currently covers only a fraction of those capabilities. As a result, occupations with higher observed exposure to AI tools are projected by the Bureau of Labor Statistics to experience weaker growth by 2034. The findings are significant for the AI and labor market communities as they pave the way for a more nuanced understanding of AI’s future impacts on employment. Initial evidence suggests that, despite some professions being highly exposed to AI, there has not been a notable rise in unemployment among these workers. However, hiring for younger employees in those exposed fields appears to be slowing. By establishing a framework for measuring the influence of AI on labor markets, this research aims to preemptively identify vulnerable jobs before displacement is evident, thereby laying the groundwork for more accurate future analyses.