Chatbots and other AI technologies have already found their way into many jobs. This is changing the world of work. Around 80 percent of workers now come into contact with AI in their tasks, as a recent survey shows. The technology helps them to complete at least part of their work much faster. According to the study, well-paid jobs and those that require long training periods are particularly affected.
Large language models such as ChatGPT and others, which are based on artificial intelligence (AI), are already part of everyday work in many professions. In the future, these Large Language Models (LLMs) and similar technologies could also completely take over some areas of responsibility and therefore threaten some jobs, according to the common narrative. But how great is the risk of losing your job to AI? Which professions are particularly affected? And how are GPTs changing the job market?
Hardly any jobs without GPT influence
A team led by Tyna Eloundou from the AI company OpenAI has now investigated these questions. To do so, they evaluated information on 923 jobs and their respective tasks stored in the O*NET 27.2 database. The researchers tested themselves and with the help of GPT-4 whether the tasks described therein can be completed at least twice as quickly if the human workers use an LLM. They assumed that the quality of the work performed would remain the same or increase as a result of the technical support.
The analysis showed that around 80 percent of employees work in positions and professions where tasks can be accelerated by at least ten percent with an LLM. For 18.5 percent of employees, tasks can be completed 50 percent faster with the help of an LLM. However, less than two percent of all the work tasks described can be carried out completely without human intervention, the team reports. The positions that use LLM the most or would theoretically benefit from it are predominantly well-paid jobs and those that require long training, as Eloundou and her colleagues found. At the top of this list are scientists and researchers as well as “technologists” such as software developers, but lawyers and pharmacists can also be found there. Jobs that produce text or code or routinely evaluate information are also noticeably often affected.
For the vast majority of things in working life, people are still necessary, at least to control and monitor them. However, many tasks could be accelerated with technical aids and only a few professions do not come into contact with GPTs at all.
Future uncertain
The study thus describes the status quo of the labor market. How this will develop in the future, however, is difficult to predict, says the team. Whether the use of LLM will change salaries, job satisfaction or the need for workers, for example, has not yet been researched. This also depends on whether further areas of application for artificial intelligence will emerge or whether current areas of application will be withdrawn. It is quite possible that jobs that currently seem to be beyond the reach of AI influence could also be done by LLM in the future if they continue to develop. Conversely, unforeseen reasons could arise why a task would be better carried out by a human.
In any case, Eloundou and her colleagues advocate closely monitoring the development of technologies and their possible consequences for the labor market, the economy and society in order to be able to counteract possible disadvantages in a timely manner.
Source: Tyna Eloundou (OpenAI) et al.; Science, doi: 10.1126/science.adj0998