The simple macroeconomics of AI: Daron Acemoglu (2024) (www.nber.org)

In a newly published working paper, economist Daron Acemoglu critically assesses the macroeconomic implications of recent advancements in artificial intelligence (AI). Using a task-based model, the paper suggests that while AI can drive productivity through cost savings at the task level, its overall impact on GDP and aggregate productivity remains modest—estimated at a mere 0.53% to 0.66% increase in total factor productivity (TFP) over the next decade. These modest projections mainly arise because early AI gains stem from easy-to-learn tasks, whereas future productivity increases are expected to be constrained by the complexity of newer, harder-to-learn tasks. Additionally, Acemoglu examines the effects of AI on wage disparities and income inequality. While AI has the potential to enhance productivity for low-skill workers in specific roles, it could paradoxically exacerbate inequality if no new tasks are created for these workers. Notably, the findings indicate that AI's impact on income distribution may be more equitable compared to past automation technologies, yet it is anticipated to widen the gap between capital and labor income. The paper further raises concerns about the potential negative social value of some new tasks brought about by AI, advocating for consideration of these factors when evaluating its macroeconomic effects.