The rise of artificial intelligence (AI) has long led to concerns about its impact on the labour market. Historically, a good portion of these concerns have been exaggerated, says the European Central Bank (ECB) based on research.
AI is a technology whose development has accelerated in recent years. Partly due to the rise of generative AI, for example in the form of OpenAI's popular ChatGPT and Google's Bard. However, AI is certainly not new; the term was defined by English mathematician Alan Turing back in 1950.
Concerns about impact on labour market
Nor is it new that the development is leading to concerns about the impact on the labour market. For instance, all kinds of AI technologies have been emerging for some time. These include deep learning, machine learning, image recognition, automated translation tools and algorithm-based recommendations.
ECB researchers point out that the rise of these earlier AI technologies has also raised concerns about the impact on the labour market. "Waves of innovation are generally accompanied by concerns about the future of jobs. These fears persist despite history suggesting that previous concerns about labour becoming redundant were exaggerated," researchers Stefania Albanesi, António Dias da Silva, Juan Francisco Jimeno, Ana Lamo and Alena Wabitsch write on the ECB website.
Surveys give varied picture
In an article, the researchers cite several previously conducted studies that give a varied picture of AI's impact on the labour market and jobs. For instance, they point to Felten et al. (2019) and Acemoglu et al. (2022), both of which show that occupational fields that are more exposed to AI do not experience a significant impact on employment.
At the same time, Acemoglu et al (2022) do show that the deployment of AI does lead to the fact that less related to AI can be affected. Based on this, the researchers conclude that although new tasks are created, AI also replaces human labour in a subset of tasks.
Focus on EU
Researchers also point out that many studies conducted so far have focused mainly on the United States (US). In a new research paper (Albanesi et al. 2023), they therefore focus on the link between AI-enabled technologies and employment in 16 European countries in the period 2011-2019. These years included the rapid rise of deep learning applications such as natural language processing, image recognition, algorithm-based recommendations and fraud detection.
This shows that a quarter of all jobs in European countries are significantly exposed to AI-assisted automation. This can be both an opportunity and a risk, according to the researchers, depending on whether AI supports or replaces jobs. Relative to jobs exposed to the impact of software developments, jobs touched by AI employ more highly skilled workers. This, according to the researchers, supports the claim that AI-based technologies can compete with jobs for highly skilled people.
At the same time, researchers point out that while the impact of AI varies by level of education, fewer differences are apparent when looking at age groups.
Positive impact
The various studies cited by the researchers used different methods to calculate the impact of AI on employment. The researchers report that regardless of the method they use, they see a positive relationship between AI-assisted automation and changes in employment. This impact ranges from 2.6% to 4.3%.
The researchers also point out that the impact of automation can vary by level of education, and therefore also affect incomes in uneven ways. For instance, middle-skilled workers in routine jobs are more susceptible to the impact of computerisation, while high-skilled workers are more susceptible to AI-assisted automation.
Increase in employment
"AI-based automation is thus associated with an increase in employment in Europe - especially for high-skilled occupations and younger workers. This is at odds with data from previous waves of technology when automation reduced the relative share of employment of middle-skilled workers, leading to polarisation," the researchers write.
"However, we find no evidence of this pattern of polarisation for our sample, even when we examine the impact of software-based automation, as shown in the software exposure by Webb (2020)." The researchers see no relationship between software exposure and changes in employment for the pooled sample. There is also no evidence that software replaces routine jobs with mediocre skills.
Further information is available here.
Author: Wouter Hoeffnagel
Image: Gordon Johnson via Pixabay
