Cutting through the noise about the impact of AI on employment

Abstract

The paper presents a critical analysis of the literature and anecdotal examples to illustrate the ‘noise’ that makes it difficult to assess the impact of artificial intelligence (AI) on the labour market. It analyses conflicting conclusions and predictions regarding the timeline of AI development, AI’s ability to replace human labour, and its impact on productivity and skill differentials. It shows that although the unprecedented rapid development of AI means that more and more tasks can be automated, the relationship between the theoretical capabilities of foundational AI models and the actual automation of automatable tasks is not linear. As an outcome of the interaction of numerous factors, the impact of technology on employment will vary from occupation to occupation. The study applies the concept of ‘skill-biased technological change’ to the AI era. It concludes that as the automation of knowledge-intensive tasks performed by graduate workers becomes easier and the number of tasks that can be automated expands rapidly, shifting labour market demand in favour of the skilled will become manifest within the category of graduate employees. Demand is expected to decline not only for inexperienced recent graduates, but also for low- to medium-skilled graduates. Competition for graduate jobs will intensify at an unprecedented rate. Only the best graduate employees will be able to retain their jobs. This means that the qualifications and skills of graduate employees whose performance is merely adequate will be devalued by AI, irrespective of the fact that these qualifications and skills had been both socially and individually quite expense to acquire. The devaluation of qualifications foreshadows a crisis in the current structure of the education system, while the devaluation of skills will lead to identity and existential crises for the individual degree holders as well as to a social crisis.