GPTs are GPTs: Labor Market Impact of Large Language Models
Generative Pre-trained Transformer (GPT) models are identified as general-purpose technologies (GPTs) with the potential to significantly impact a vast majority of the U.S. labor market. Research suggests that approximately 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of these models, while roughly 19% of workers may see at least 50% of their tasks impacted.
Labor Market Exposure Metrics
Large Language Models (LLMs) exhibit a broad reach across various occupations, with the degree of impact measured by the percentage of tasks within a job that can be performed or assisted by GPT capabilities.
- Broad Impact: 80% of the U.S. workforce is expected to have at least 10% of their tasks affected.
- High Impact: 19% of workers may experience an impact on 50% or more of their work tasks.
- Income Correlation: The influence of GPTs spans all wage levels, though higher-income jobs potentially face greater exposure to these technologies.
Methodology for Assessing Occupational Impact
To determine the potential labor market implications, researchers used a new rubric to assess occupations based on their correspondence with GPT capabilities. This assessment incorporated two primary sources of data:
- Human Expertise: Professional judgment and human analysis of job roles.
- GPT-4 Classifications: The use of GPT-4 itself to help classify and analyze the correspondence between model capabilities and occupational tasks.
Economic and Social Implications
Because Generative Pre-trained Transformers function as general-purpose technologies, their impact is not limited to industries that have already seen high productivity growth. This characteristic suggests that LLMs will have wide-ranging economic, social, and policy implications as they are integrated into the workforce across diverse sectors.
Key Findings Summary
| Metric | Impact Level |
|---|---|
| Workforce with >10% tasks affected | ~80% |
| Workforce with >50% tasks affected | ~19% |
| Most Exposed Groups | Higher-income jobs |