Findings from the National Bureau of Economic Research indicate that artificial intelligence (AI) might not greatly boost productivity and could potentially exacerbate income inequality. Daron Acemoglu, a professor of economics at MIT, suggests that predictions regarding AI's effects on productivity and wages are excessively optimistic.
Skepticism Over AI's Economic Benefits
In his paper “The Simple Macroeconomics of AI,” Acemoglu scrutinizes assertions that AI will produce substantial productivity gains. While some estimates, like a 100 percent increase in GDP in the next decade, lean towards high optimism, others, such as Goldman Sachs' projection of a seven percent rise in global GDP and a 1.5 percentage point boost in productivity over ten years, are more conservative.
A McKinsey Global Institute report aligns with these lower estimates, suggesting an annual GDP growth boost of 0.5 to 3.4 percentage points in advanced economies over the next ten years. Acemoglu remains skeptical, drawing on historical trends in automation where benefits favored business owners and executives, leaving the workforce with negative repercussions.
Modest Productivity Gains Expected
Based on studies indicating that about 20 percent of American workers could have half of their tasks automated by large language models (LLMs), Acemoglu forecasts potential labor cost savings of 27 percent, equating to an overall cost reduction of 14.4 percent. Despite these savings, he argues that the effects on total factor productivity (TFP) will be limited.
Acemoglu's calculations suggest a TFP growth of just 0.064 percent annually, predicting a GDP growth of between 0.93 percent and 1.16 percent over the next decade. He warns that these numbers might still be optimistic as future tasks could be more intricate and less amenable to automation. He therefore anticipates a modest increase in TFP and GDP, around 0.53 percent and 0.90 percent, respectively.
Inequality Concerns
Acemoglu also addresses how AI might affect wages and economic disparity. He argues that AI is unlikely to significantly raise wages or mitigate income inequality. Even if AI enhances the productivity of lower and middle-tier workers, it may still negatively impact the real earnings of lower-educated women, particularly those who are white and native-born. His findings suggest AI could widen the income gap between capital owners and labor, worsening existing disparities.
In October last year, leading economists expressed concerns over the divide AI could cause in the workforce. The study, based on government data from a 2018 survey of 474,000 companies (representative of around four million businesses in the US), found that less than 6% of the businesses were employing AI technologies at that time. However, the adoption rate among larger businesses, with more than 5,000 employees, was above 18%.
