With the growing popularity of ChatGPT and GPT-4the questions about whether they can put us out of workThey did not wait. And while these statements tend to be quite biased, OpenAI is studying the possible impact of artificial intelligence on the job market. The results obtained so far are very interesting.
In the last few hours, researchers from the startup Directed by Sam Altman they published a document analyzing the potential implications that GPT models could have on US employment. To do this, they looked at the consistency of AI use across a wide variety of jobs, as well as human experience and recently earned GPT-4 academic qualifications.
While OpenAI’s research is extremely technical, it allows us to extract some relevant details. For example, that professions that require scientific and critical thinking skills, are less likely to be impacted by developments like GPT-4. While those jobs that require writing and programming skills are the most exposed. What does the latter mean? Which are the jobs most likely to be influenced by large language models.
The works most exposed to language models such as GPT-4
Another interesting point is that researchers have presented their data talking about “exposure percentage” to models like GPT-4. This is because they have studied how affected a job is by studying the possible implementation of artificial intelligence on it. Specifically, How much would it help to reduce the completion time that a human requires to accomplish at least 50% of the task?while maintaining consistent quality.
In this way, they have managed to identify three levels:
- No exposure: that reducing the completion time of a task is not a requirement —not even a minimum— to maintain an equivalent quality. Or that any attempt to improve the time it takes leads to a drop in the quality of the result obtained.
- direct exposure: when just using a language model like GPT-4 on platforms like ChatGPT is enough to halve the time needed to complete a task or activity.
- LLM+ exhibition: when a great language model (LLM) alone is not enough to complete the task 50% faster. However, it could be used as a basis to develop software that does.
Thus, OpenAI has shared a table that shows the jobs with the greatest exposure to GPT models, or software based on them. It is worth noting that both “human evaluators” and models based on GPT-4 have been used. In the case of the former, they identified the following professions as 100% exposed to language models or software based on them:
- mathematicians
- accountants
- financial analysts
- Writers and authors
- Web and digital interface designers
On the other hand, the lowest exposure percentages were registered in Public Relations specialists (66.7%), animal scientists (66.7%) and poets, lyricists and creative writers (68.8%).
While the models based on GPT-4, identified the following jobs with 100% exposure:
- mathematicians
- Accountants and auditors
- News analysts, reporters and journalists
- Legal secretaries and administrative assistants
- Clinical data supervisors
- Climate change policy analysts
Other professions that also have a high exposure to language models such as GPT-4 —between 90 and 97%— are: blockchain engineers, court reporters, proofreaders and even those in charge of delivering correspondence within a company.
OpenAI also highlights that the most impacted jobs are those that require more education:
“Our analysis suggests that people with bachelor’s, master’s, and professional degrees are more exposed to GPTs and GPT-powered software than those without formal educational credentials. Interestingly, we also found that people with some college education, but no degree, exhibit a high level of exposure to GPT and GPT-powered software.”
The other side of the coin
The other side of the coin, as OpenAI explains, is that while GPT-4 can help reduce the time required to complete tasks in the most exposed jobs by up to 50%, does not mean that technology is capable of automating all its functions. This means that, even if the exposure of a profession is 100%, AI is still far from replacing it. At least in the immediate future.
The impact of language models like GPT-4 and applications based on it, like ChatGPT, is evident. In fact, it is reported that 80% of the US workforce has at least 10% of their tasks affected by GPT models. While 19% of workers would see at least 50% of their tasks impacted. Is this the same as saying that artificial intelligence will take your job? Not at all. In fact, OpenAI raises it from the point of view of labor efficiency, and not of replacement of the human worker.
«[…] While the technical ability of GPT models to make human labor more efficient seems self-evident, it is important to recognize that social, economic, regulatory, and other factors can influence actual labor productivity outcomes. As capabilities continue to evolve, the impact of GPTs on the economy is likely to persist and increase, posing challenges for policymakers in predicting and regulating their trajectory. More research is needed to explore the broader implications of GPT advances, including their potential to augment or displace human labor and their impact on job quality, inequality, and skills development, among many other consequences.”
Today there is an increasingly strong tendency to say —often with few or no arguments— that AI is coming to keep our work. And reality is usually quite far from that drastic position. It is clear that it is necessary to study the impact of tools such as GPT-4 on employment, since everything indicates that they will be increasingly powerful and capable, and that in one way or another they will be integrated into the labor dynamics. However, blindly promoting the idea that jobs will disappear overnight because of artificial intelligence is, to say the least, irresponsible.