GPT will affect 80% of US jobs, OpenAI research says

As large language models like OpenAI’s GPT-4 become more sophisticated and able to write, code, and perform mathematical calculations with greater precision and consistency, it won’t be long before artificial intelligence penetrates workplaces as a tool of daily use.

In fact, OpenAI itself is betting that at least part of the work of the vast majority of workers will be automated using GPT models.

Researchers from OpenAI and the University of Pennsylvania say that 80% of America’s workforce has at least 10% of their jobs affected by GPT implementation.

They also found that about 19% of workers would see it affect at least 50% of their tasks. The impact of GPT is greater for high-wage jobs but covers almost all industries. They argue that GPT models are general-purpose technologies, like the steam engine or the printing press.

Researchers used the O*NET database, which is the primary database of US occupations and contains 1,016 occupations with standardized descriptions, to identify the tasks to measure for each occupation.

They then combined human-written annotations and annotations generated with GPT-4 using a rubric to determine whether accessing GPT directly or a secondary GPT-based system would reduce the time required for a person to perform a certain task by at least 50%.

Occupations with the highest GPT exposure include mathematicians, tax advisors, writers, web designers, accountants, journalists, and legal secretaries. The professions least likely to be affected by GPT are graphic designers, developers of search marketing strategies and financial managers.

The researchers also list GPT’s overall anticipated impact on different industries, with the greatest impact being on data processing services, information services, publishing industries, and insurance carriers, while the least impact is on food manufacturing, wood product manufacturing, and support activities for agriculture and forestry.

The researchers acknowledge that their study has some limitations because the people who made the annotations were familiar with the models’ capabilities and did not belong to some of the occupations that were studied. Another limitation is that GPT-4 is sensitive to the wording and composition of questions, and can sometimes fabricate information, so its results are not necessarily the ultimate truth. Of course, it should be noted that OpenAI produced this work itself, and as a commercial AI modeling company, it has a strong incentive to portray its tools as disrupting industries and automating tasks that ultimately benefit employers.

However, the report shows how GPT models will soon become a mainstream tool. Google and Microsoft have already announced that they will introduce artificial intelligence into their office products, such as email, documents, and search engines. Startups are already using GPT-4 and its encoding capabilities to reduce human development costs.

“Our analysis indicates that the impacts of LLMs like GPT-4, are likely to be pervasive,” the researchers write. “While LLMs have consistently improved in capabilities over time, their growing economic effect is expected to persist and increase even if we halt the development of new capabilities today.”