The effect of AI on labour demand: a critical assessment of ‘Power and Progress’ by Acemoglu and Johnson

• Detailed focus on effect of AI on labour demand in ‘Power and Progress’ • Heterodox perspective on ‘Power and Progress’ • Explains how motives behind anti-labour bias are unclear and worker-friendly technologies are hard to identify • Suggests differing policy implications once orthodox assumptions in ‘Power and Progress’ are dropped Task-based models of production have led to a theoretical reappraisal of the effect of new technology on labour demand. This paper critically assesses the policy implications of this research agenda, with particular reference to Acemoglu and Johnson’s recent book, Power and Progress . While Acemoglu and Johnson take welcome steps away from previous orthodoxy, their analysis has several flaws which affect the policy lessons to be drawn, including: (i) the explanation for anti-labour bias is unclear; (ii) worker-friendly technologies are not clearly characterised in theory, and hard to identify in practice; (iii) macroeconomic policy orthodoxy is largely unquestioned. More generally, much of Power and Progress remains unhelpfully constrained by theoretical commitments to mainstream economics.