22.03.2023 IT AI Farvest Decrypt Study

How large language models are to impact the labor market?

Writer Samira Joineau
large language models

What could be the impact of large language models, such as GPT? This is the question that researchers from OpenAI, OpenResearch, and the University of Pennsylvania investigated in a study report published yesterday. It revealed that 80% of the workforce could be affected.

The field of generative AI and large language models (LLMs) has experienced considerable progress over the last few years, months, and even weeks – notably with the introduction of GPT-4. As a reminder, LLMs can “process and produce various forms of sequential data, including assembly language, protein sequences and chess games, extending beyond natural language applications alone”, notes the report. 

As the researchers explain, their study aims to investigate the potential implications of generative Pre-trained Transformer (GPT) models and related technologies rather than “the progress of these models alone”. And the results are rather concerning for job positions. 

Both human and GPT-4 annotations suggest and indicate that, based on their task-level capabilities, GPTs have the potential to “significantly affect a diverse range of occupations within the U.S. economy, demonstrating a key attribute of general-purpose technologies”. On the one hand, the numbers show that 80% of the workforce has an occupation with at least one task exposed to GPTs; while, on the other hand, 19% of the workforce belong to an occupation where “over half of their tasks are labeled as exposed” – meaning LLMs could strongly affect them

When it comes to skills, the study’s findings show that science and critical thinking ones are less likely to be impacted by current LLMs compared to programming or writing skills; meaning that their chances to be influenced by LLMs are higher. 

Overall, job positions in the sector of finance, journalism, engineering, education, and also graphic design, are the most exposed.  

As the authors mentioned, the research and results present limitations in the sense that the study focuses on the United States; findings could indeed differ from a country to another, considering factors such as industrial organization, technological infrastructure, or regulatory frameworks, among others. Besides, they note that “there may be discrepancies between theoretical and practical performance, particularly in complex, open-ended, and domain-specific tasks”. 

On top of this, the study specifies that it would be rare to find LLMs or other AI tools that could fully take over an entire job position. “Further research is necessary to explore the broader implications of GPT advancements, including their potential to augment or displace human labor, their impact on job quality, impacts on inequality, skill development, and numerous other outcomes”, state the authors.  

On his side, OpenAI CEO Sam Altman admitted that AI, as it is growing unerring and more powerful, will likely replace some jobs in the near future. And this could happen sooner than expected.