https://www.nber.org/papers/w33777
https://futurism.com/time-workers-save-ai-jobs
A research working paper pointing to LLM not having a big impact. And not improving salary even if it helped with productivity of workers. It’s just 2 years in to the AI craze, might be too early to evaluate the true impact.
Summary of “Large Language Models, Small Labor Market Effects” Working Paper
Central Research Question
The paper investigates whether the adoption of large language models (LLMs), such as ChatGPT and similar AI chatbots, has significant impacts on labor markets. The central hypothesis is that while LLMs are widely adopted in workplaces, their effects on employment, wages, and job displacement are relatively modest, challenging assumptions that AI will rapidly disrupt labor markets.
Methodology
The authors conducted a large-scale survey of workers in 11 professions (e.g., journalism, software development, customer service) in Denmark, analyzing responses from over 16,000 participants. They combined self-reported data on AI chatbot usage, employer policies (e.g., whether companies encourage AI adoption), and perceived impacts on job tasks, productivity, and earnings. Statistical techniques such as difference-in-differences models and empirical Bayes shrinkage were used to isolate the effects of AI adoption from other variables. The study also compared outcomes between workers in firms that actively encouraged AI use and those in firms without such policies.
Key Findings
- Adoption Rates: Approximately 50–80% of workers in AI-relevant roles reported using AI chatbots for work, with higher adoption in occupations like software development and customer service. Employer encouragement significantly boosted adoption (doubling usage in some cases).
- Productivity and Benefits: Users reported modest productivity gains, such as time savings (15–60 minutes daily) and improved task quality. However, these benefits varied by occupation and were often offset by increased workloads (e.g., new tasks tied to AI oversight).
- Earnings and Employment: There was no significant evidence that AI adoption led to widespread job displacement or substantial wage changes. While some workers reported small earnings increases (under 15%), others noted no effect or even reduced income in specific roles.
- Task Transformation: AI adoption primarily led to task augmentation (e.g., automating routine tasks) rather than replacement. New tasks emerged, such as managing AI outputs or integrating tools into workflows, but these were often absorbed by existing roles.
Implications for the Labor Market
The study suggests that LLMs are not currently a major driver of large-scale job displacement or wage volatility. Instead, their impact is more nuanced, involving incremental productivity gains and task reshaping. Employers and policymakers should focus on reskilling workers for AI-augmented roles rather than anticipating mass unemployment. The findings also highlight the importance of workplace policies (e.g., training programs) in maximizing AI’s benefits while mitigating risks like skill gaps or over-reliance on technology.
Limitations and Caveats
- The study focuses on Denmark and specific professions, limiting generalizability to other regions or industries.
- Self-reported data may introduce bias, as workers might overestimate benefits or underreport challenges.
- The analysis captures short- to medium-term effects; long-term impacts (e.g., as AI tools evolve) remain uncertain.
- The study does not address indirect effects, such as broader economic shifts or non-quantified qualitative changes (e.g., job satisfaction or creativity).
Conclusion
While LLMs are transforming work, their labor market effects are currently “small” in magnitude compared to historical industrial shifts. The paper underscores the need for balanced policymaking that supports workforce adaptation without overestimating AI’s disruptive potential. Future research should monitor how these tools scale and integrate into diverse sectors.
Leave a Comment