It took just a few weeks for freelance writers, especially the most qualified, to lose contracts and income after the launch of ChatGPT, US university researchers have found. Are they the canaries in the mine, the first of the “300 million jobs” set to disappear? For an expert, it is not so much artificial intelligence as “business decisions” that are in question.
92,547
Number of freelancers whose public data was used between April 2022 and March 2023 on Upwork. This vast platform, available in 180 countries, connects some 5 million customers and 18 million professionals. “Due to the flexibility of this niche market compared to traditional jobs, it is a perfect tool to explore the short-term effects of ChatGPT,” explain three American researchers, Xiang Hui and Oren Reshef, from the University of Washington in Saint Louis, and Lupofeng Zhou, of New York University, in their study The Short-Term Effects of Generative Artificial Intelligence on Employment, published on July 31, but not peer-reviewed.
Serial declines
Since ChatGPT is a generative artificial intelligence (AI) that primarily uses written language, the researchers focused on the freelancers most likely to be affected by its arrival, those offering “writing-related services” like creating content, synthesis, editing, writing. They noted a marked drop in the number of contracts obtained, of around 2.6%, five months after the launch of ChatGPT.
In terms of income, the drop is even more spectacular: around 10% during this period.
The risk of quality
Secondly, the researchers attempted a bold approach: establishing whether the “quality” of the freelancer lessened the fall. “We used several measures to determine this quality, including their past jobs and income, the skill level required for those past jobs, their compiled performance, and their hourly rate,” the authors explain.
Surprise: “high-level performance and high-quality service do not help to lessen the effects of the introduction of generative AI”.
Through a complex combination of data, we arrive at an even more worrying conclusion: “There is evidence to suggest that top employees are disproportionately hit by AI. »
The researchers wanted to conclude their study on a note of hope. “Our report focuses only on the effects [négatifs] on workers. Assessing the full implications of the arrival of generative AI on stakeholder well-being is a task beyond the remit of this study, but remains a promising direction for future work. »
Many warnings
The three American researchers are not the first to sound the alarm about the risks of artificial intelligence in terms of employment. Their study, however, has the merit of demonstrating empirically, and in the short term, the effects of a generative AI like ChatGPT.
In February 2023, based on a survey of 500 European companies using ChatGPT, the firm Sortlist Data Hub concluded that 26% of software and IT companies planned to cut jobs as a direct result of the arrival of the AI.
Last March, a report by Goldman Sachs economists estimated that more than 300 million full-time jobs worldwide, mostly white-collar, could be automated with the use of AI like ChatGPT. The effects would be more marked in developed countries.
Since November 2022, a number of media outlets around the world, including the Washington Post And India Todaypublished testimonials from creative professionals who were surprised to be replaced by generative AI.
The wrong target
Frédérick Plamondon knows the theme of threatening AI well. This doctoral student in industrial relations, former lecturer at Laval University, notably addressed hot topics such as the environmental cost of AI, privacy and the construction of myths in the 73 episodes of the podcast show IA Café, of which he is one of the co-authors.
He notes from the outset that the study focuses on freelancers, and not more generally on “white collar workers” as some articles present.
This is a big difference: white-collar workers are professionals who are generally salaried, office workers, middle managers, supervisors, department heads […] Freelancers are another context: it’s not an employee relationship, they don’t have social protections, they are people in a precarious situation.
Frédérick Plamondon, doctoral student in industrial relations, former lecturer at Laval University
We can indeed expect profound transformations in the job market, but they are not inevitable, believes the expert. “Is it the machine that causes this, or are these not more management decisions? It is not artificial intelligence that is causing job losses, it is managers who choose to move towards simulated synthetic expertise because they think it will suit their clientele. This is above all the issue. »
Mr. Plamondon pleads, as many speakers have done since last March, for better regulation of AI. Not necessarily to slow down its development, but to control its impact on our lives. “It’s not that artificial intelligence is becoming dangerous, but these new technologies put a lot of power in the hands of companies that will use it to generate more profits […] And employers choose this solution rather than paying human beings to work and develop their skills. »