Many white-collar jobs can be done much more efficiently with the help of artificial intelligence such as ChatGPT, according to a recent research paper. While the AI can help at least a little bit with most jobs, the impact would tend to be greater on higher-paid jobs that require higher education, the paper says.
Two of the authors work for OpenAI, the developer of ChatGPT.
The researchers looked at individual tasks associated with each job and marked “affected” tasks—those that could be completed at least twice as fast using a “Generative Pre-trained Transformer” such as the language processing AI ChatGPT. They gauged separately which tasks would be affected by a general-purpose GPT on its own and which would be affected by specialized GPT-based software.
They found that only 3 percent of U.S. workers would have at least half their tasks affected by the general GPT, but that the percentage would rise to nearly 50 percent if including anticipated GPT-based tools.
Higher-paid jobs had about twice the share of their tasks affected than lower-paid jobs. Also, jobs that require a Bachelor’s degree or higher had more than twice their tasks affected than jobs that require an associate’s degree or less.
The authors cautioned, though, that “there are numerous low-wage occupations with high exposure and high-wage occupations with low exposure” to the AI.
The most affected professions would be mathematicians, tax preparers, financial quantitative analysts, writers and authors, web and digital interface designers, interpreters and translators, survey researchers, public relations specialists, and animal scientists.
The most affected industries would be various market and financial services, real estate and insurance services, data analysis, media, and occupations revolving around writing, math, and coding, according to the paper.
Many occupations would be affected very little or not at all, particularly those focused on manual labor, such as construction workers, farm workers, etc. As far as brain work is concerned, tasks that require critical and scientific thinking, as well as active learning, would tend to remain unaffected.
Just because someone’s job would be highly affected doesn’t mean it would be automated away. The paper didn’t examine which jobs might disappear due to the AI. It showed that many tasks can be done faster with the help of an AI. That could mean that fewer workers would be needed in the affected occupations, but the authors note that “far-reaching consequences” of technologies with such generalized potential “unfold over decades” and “are difficult to anticipate, particularly in relation to labor demand.”
Much would also depend on how the GPT tech is implemented. The various GPT-based tools first need to be developed and workers need to learn how to use them effectively.
“Users of GPTs and GPT-powered systems are likely to become increasingly acquainted with the technology over time, particularly in terms of understanding when and how to trust its outputs,” the paper says.
Finally, the government could shake things up by stepping in to regulate the use of GPTs.
“Social, economic, regulatory, and other determinants imply that technical feasibility does not guarantee labor productivity or automation outcomes,” the authors said.
