- The productivity boost that AI technology will generate is being overestimated, according to a new MIT study.
- AI-led US GDP growth in the next 10 years will rise just 0.93% to 1.16%, economist Daron Acemoglu wrote.
Talk to nearly any artificial intelligence bull, and they'll likely mention the technology's huge expected economic impact.
It's expected to turbocharge productivity, enhancing the efficiency of today's human-led jobs. That, in turn, is seen triggering a new era of growth.
Goldman Sachs is among those bullish on AI's economic prospects, previously forecasting that annual global GDP will eventually see a 7% boost from AI's labor impact. Vanguard also recently estimated that average real GDP could exceed 2.3% between 2028 and 2040.
But one MIT economist isn't so sure.
In a new National Bureau of Economic Research study, MIT economist Daron Acemoglu projects just mild economic upside in the US, stemming from AI advancement.
"My calculations suggest that the GDP boost within the next 10 years should also be modest, in the range of 0.93%-1.16% over 10 years in total," Acemoglu wrote. "Presuming that the technology prompts an investment boom, this forecast could rise to a range of 1.4%-1.56% in total."
The low estimate stems from subdued outlooks on how much AI can really advance total factor productivity; while nontrivial, this figure will also be modest, Acemoglu wrote.
That's especially made true in the face of "hard-to-learn tasks," such as decision-making situations that are heavily context-dependent. According to the study, many of today's bullish forecasts overestimate productivity gains, as they only take into account easy tasks.
But when recognizing hard tasks, AI's boost to total factor productivity falls to an upper bound of 0.53% within the next decade.
While Acemoglu notes that new tasks and products from AI will boost GDP, not not every contribution will be positive. The technology will likely also increase manipulative tasks, pulling down on welfare.
Bad tasks includes deepfakes, false advertising, social media addiction, and AI-led computer hacks, he listed. While these could add 2% to GDP, the impact on welfare would actually be a contraction of 0.72%, he said.