These fears about technology are nothing new, as every major technological advance has given rise to similar concerns. But this latest one comes against a tense global socio-economic backdrop that is conducive to the publication of alarmist studies. The International Labour Organisation (ILO) has published a detailed report analysing the impact of generative AI on "the quantity and quality of employment". Its findings are instructive, as they call for nuance and aim to inform informed policy responses.
At the beginning of the 20th century, the introduction of assembly lines in the car industry radically transformed employment. The automation of manufacturing and assembly processes increased production and reduced costs. This model, known as Fordism, spread to other sectors such as the food, textile and pharmaceutical industries, as well as the production of electronic components. Similarly, the first mainframe computers in the 1950s and 1960s revolutionised document and archive management methods, as well as data processing. With a tenfold increase in computing capacity, they are now used everywhere. These technological advances have had a considerable impact on our society, our economy and our lifestyles. They have brought significant benefits, such as the lightening of arduous tasks and increased productivity, but above all the creation of new jobs. However, they have also had negative repercussions in the form of widening socio-economic inequalities. What about the generative AI revolution? How is this development different?
In the same way as Fordism and computerisation in the last century, the advent of generative AI offers companies the opportunity to optimise processes by transferring part of the workload from humans to AI, using natural language. These tools, which are capable of handling increasingly complex tasks, offer a host of application opportunities in a variety of fields, such as the creation of computer code, image recognition and even basic medical diagnostics. At the same time, they raise ethical issues relating not only to respect for personal data but also to corporate social responsibility. AI therefore has the potential to radically transform the labour market, and it is legitimate to ask how many jobs will be impacted and in what way. Recent studies have sought to shock public opinion by announcing massive job losses in order to boost the financial value of these technologies: 120 million in the world's 12 largest economies according to IBM, 300 million in Europe and the United States according to Goldman Sachs! Who can beat that? The aim is always to achieve a sharp increase in the GDP of countries thanks to these "incredible" generative AIs. The OECD is also concerned, estimating that 27% of jobs are highly exposed to the risk of automation.
Past experience shows that technological progress has always created more jobs than it has destroyed. The quantity of jobs, in this context, is less to be feared than their distribution among workers. The real issue, then, lies in the reorientation and transformation of employment. Because this technology is part of an ultra-globalised and interconnected economic context, conducive to the extremely rapid and widespread dissemination of these technologies.
An enlightening analysis
The ILO's analysis is based on an international classification of occupations (ISCO: International Classification of Types of Occupation) associating tasks with 436 occupations. An exposure score measures the risk of a task being automated by generative AIs, and makes it possible to assess their effect on the occupation. Among the main results, the ILO estimates that 24% of office tasks are highly exposed to GPT4 and 58% are moderately exposed. Cashiers, data entry operators and administrative positions would be particularly vulnerable, but not for all their tasks. It should be noted that these jobs have already begun to be impacted well before the arrival of generative AI. Non-office jobs are in fact less at risk, with 1 to 4% of tasks highly exposed compared with no more than 25% of tasks moderately exposed. While these figures relate to tasks, they do not necessarily call into question the viability of the jobs that depend on them. On the contrary, this technology could free up employees' time to devote to new tasks. However, company management must be alert to the risk of increasing work intensity once technology has been integrated into a job. Social dialogue to facilitate this transition is essential.
The ILO's assessment also reveals disparities according to countries' income levels. The structure of the labour market differs greatly from one geographical area to another. High-income countries are more exposed than low-income countries (5.5% of total employment compared with 0.4% respectively). Although this study predicts a significant increase in employment (13.4% in high-income countries and 10.4% in low-income countries), it does not take into account the differences in infrastructure that could hinder this technological penetration in low-income countries. As a result, there is a significant risk of widening the productivity gap, and hence of widening inequalities between nations.
There is another important point to note. There is a strong gender bias when we put the data into perspective: women would be affected more than twice as much as men by generative AI. This analysis is mainly explained by an over-representation of women in sectors and professions more exposed to automation, and an under-representation in scientific, digital and advanced technology professions. This finding highlights how far we still have to go in terms of equal representation of men and women in employment, and the need to reduce gender inequalities in several sectors of activity and levels of responsibility.
A positive outlook
A more significant effect of generative AI could be on the quality of employment rather than its quantity. This would provide an opportunity for job transformation rather than outright replacement. In this case, companies and governments will have to be prepared to reshape jobs to adapt to this new situation. Governments, social partners and all stakeholders must act to proactively design policies for an orderly transition, solutions to encourage gradual change. As is the case in many projects involving transformation, it would be preferable to avoid over-reactive change management.
The ILO provides us with an objective and detailed analysis of the risks, but above all the opportunities, associated with job-generating AI. The interpretation, broken down by profession, is much closer to the day-to-day challenges faced by companies and employees.
Like any major technological innovation, it would be unrealistic to make generative AI disappear. It should not be seen as a threat, but as an opportunity to transform existing professions and create new skilled jobs. To achieve this, we need to anticipate the ethical and social challenges it presents. It is therefore urgent to act, but to act responsibly, to make this technology a formidable tool at the service of the evolution of our human societies.
Dr. Sophie Fayad, Dr. Laurent Cervoni, Talan Research and Innovation Centre.
Gmyrek, P., Berg, J., Bescond, D. Generative AI and Jobs: A global analysis of potential effects on job quantity and quality. ILO Working Paper 96. Geneva: International Labour Office, 2023
Jill Goldstein, Bill Lobig, Cathy Fillare, Christopher Nowak, Augmented work for an automated, AI-driven world, IBM Institute for Business Value, Août 2023
Joseph Briggs, Devesh Kodnani, The Potentially Large Effects of Artificial Intelligence on Economic Growth, Goldman Sachs, Mars 2023
Marguerita Lane, Morgan Williams, Stijn Broecke, The impact of AI on the workplace: Main findings from the OECD AI surveys of employers and workers, OECD Social, Employment and Migration Working Papers No. 288, Mars 2023