Demand for professional skills has changed significantly since the launch of ChatGPT in late 2022, according to a new study. But the picture is complex. While there were fewer jobs in areas such as copywriting and translation, the demand for other skills increased.
Will artificial intelligence (AI) revolutionise the labour market and make millions of jobs redundant? Or is there a tendency to exaggerate the impact of ChatGPT and other forms of generative AI?
A new international study published in the Journal for Economic Behaviour and Organization (see box) concludes that the truth is probably somewhere in between. Artificial intelligence is rapidly replacing jobs in some areas, but it also creating new jobs in others.
The main takeaway is that we are seeing significant changes in the composition of the labour market.
The study is based on statistical analyses of more than three million freelance jobs with different skill profiles posted on one of the leading international freelance platforms before and after the launch of ChatGPT in November 2022.
"We can see that the demand for skills such as translation or copywriting and simple programming decreased significantly after the introduction of ChatGPT. But the study also shows that the demand for other skills has increased," says Ole Teutloff, who is affiliated with the Copenhagen Centre for Social Data Science (SODAS) at UCPH.
"The main takeaway is therefore that we are seeing significant changes in the composition of the labour market. But we are so far not talking about new AI technologies wiping out jobs across the board, making human labour redundant."
Job posts down by almost a quarter
Overall, the analysis shows that the number of freelance jobs offered in areas where ChatGPT can largely take over has decreased by around 24% compared to jobs where AI is used as a complementary tool or has no impact at all.
The jobs most negatively affected were writing about real estate and 'about us' pages on websites. Here the number of posted jobs fell by more than 50%. Conversely, the number of freelance jobs increased in areas such as machine learning, chatbot development and more creative forms of text and video production.
Negatively affected (less jobs):
- Writing 'About us' pages
- Real estate content writing
- Web design with Java script
- Proofreading
- Translation (W. European languages)
- Business analysis
- Data science and statistics
Positively affected (more jobs):
- Chatbots
- Creative writing, explainer videos
- Machine learning
- Travel planning
The decline was particularly strong for short-term jobs of 1-3 weeks and for jobs aimed at freelancers with limited experience. According to the researchers behind the study, this makes sense, as more complex tasks continue to require human involvement.
Johanna Einsiedler, co-author of the paper and PhD student at SODAS, cautions against placing too much emphasis on specific numbers in the study, as the field of AI is developing rapidly. However, the overall trend is clear:
"We are potentially facing big changes. When we see the number of jobs available in the most AI-sensitive sectors falling by a quarter in just two years, it's obviously significant. And we probably haven't seen the full impact yet, with the freelance sector being affected faster than the rest of the labour market," she says.
New demands for flexibility
According to the researchers behind the study, this poses new challenges for both employees and employers. Integrating AI into work processes will create winners and losers, depending on how AI technologies are implemented and what skills individuals have. Some jobs will disappear, and others will emerge.
Some of the most important skills will be the curiosity to learn and the ability to adapt.
This has happened before when new technologies have taken off. But in this case, policymakers should be paying special attention to how fast things are moving, Teutloff stresses:
"It is a particular challenge when we talk about education and training, which takes time to change. Fortunately, Denmark has an education system that focuses on general skills. In my opinion, this will become even more important in a rapidly changing world. Some of the most important skills will be the curiosity to learn and the ability to adapt," he says.
Figure: Estimated labour demand shock by AI exposure category (relative to unaffected job categories)
About the study
The study 'Winners and losers of generative AI: Early Evidence of Shifts in Freelancer Demand' has been published in Journal for Economic Behavior and Organization.
Behind the study is an international group of researchers. In addition to first author Ole Teutloff and Johanna Einsiedler, both from SODAS at the University of Copenhagen, contributors include Otto Kässi of ETLA Economic Research in Finland, Fabian Braesemann of the Oxford Internet Institute, Pamela Mishkin, USA, and R. Maria del Rio-Chanona of University College London.
The study analyses freelance jobs posted on an international platform (anonymised) focusing on how the introduction of ChatGPT changed the demand for skills partitioned into 116 clusters. These were grouped into three aggregated clusters:
- Skills that can be substituted by AI tools
- Skills where AI tools are complementary
- Skills that are not affected by AI
The latest version of ChatGPT was used when defining the individual and aggregated skill clusters.
