🤖📈📉 AI's economic potential: Goldman Sachs responds to Daron Acemoglu

Alsol: A Quick Q&A with … economists Paolo Abarcar and Caroline Theoharides on immgration and 'brain drain' 🌐🧠

Super-Short Summary: In response to economist Daron Acemoglu's recent paper, which estimates a modest boost to US productivity and GDP growth from generative AI, Goldman Sachs offers a more optimistic outlook. The bank attributes the gap in growth estimates to differing assumptions about the share of automatable tasks, as well the impact of labor reallocation and new task creation. GS sees Acemoglu's assumptions as too conservative, as he only considers tasks that can be profitably automated in the near term. Despite acknowledging the validity of Acemoglu's concerns, Goldman Sachs maintains its expectation of significant long-term impacts on productivity and GDP.

Are we overestimating the potential economic impact of new advances in artificial intelligence? That’s the discouraging conclusion of a new paper from noted MIT economist Daron Acemoglu, “The Simple Macroeconomics of AI.”

As I wrote the other day in an essay about the study, Acemoglu uses existing (and, admittedly, early days) estimates of job exposure to AI to provide “back of the envelope” estimates on two things: first, the share of economy-wide tasks that could be affected by AI; second, estimates of cost savings and productivity improvements due to AI. 

He then estimates the potential increase in total factor productivity (TFP measures the efficiency of capital and labor use in production and reflects innovation's impact on the economy) by multiplying the proportion of tasks that generative AI can automate — adjusted for their contribution to GDP — by the average cost savings from this automation.

As the economist calculates, generative AI will generate a “nontrivial but modest” boost to US productivity and GDP growth over the next decades, by 0.7 percent and 1.1 percent, respectively. As I previously noted, the Acemoglu estimate of 0.07 percent faster TFP growth per year is hardly nothing when you consider TFP has averaged 0.5 percent annually since 2007.

That said, the Acemoglu forecast is way more conservative than some others. For example: Goldman Sachs predicts US productivity growth could rise by 1.5 percent annually over the next decade. McKinsey forecasts the overall impact of AI and other automation technologies could produce a 1.5 to 3.4 percentage point rise in average annual GDP growth in advanced economies over the coming decade. Acemoglu calls those forecasts “hyperbolic” in his paper.

Goldman Sachs responds

But Goldman doesn’t think so. In a response to Acemoglu, the bank attempts to answer why its estimates of the AI boost to productivity growth are “so much larger than Acemoglu’s?” Two explanations are offered that explain the “vast majority” of the gap, according to GS: