A significant portion of the vast investments being funneled into artificial intelligence (AI) infrastructure may fail to deliver on their promises, warns MIT Economist Daron Acemoglu. In a recent interview with Bloomberg, Acemoglu expressed skepticism about the high expectations surrounding AI, cautioning that much of the capital could be wasted.
“A lot of money is going to get wasted,” Acemoglu stated.
According to Acemoglu, only around 5% of jobs are truly poised to be taken over or meaningfully enhanced by AI technologies in the next decade. This estimate suggests that the anticipated economic benefits from AI, such as substantial productivity gains and efficiencies, might not materialize, at least not anytime soon.
“You’re not going to get an economic revolution out of that 5%,” Acemoglu added.
One of the key concerns is that major investments in AI by cloud hyperscalers like Microsoft, Amazon, and Meta Platforms—particularly in Nvidia’s AI-enabled GPUs—may not result in the expected revenue increases. Should investors start scrutinizing profit margins and the timeline for returns on these investments, there could be a sudden cooling of the AI narrative.
Acemoglu foresees three possible scenarios for AI’s future, none of which appear particularly promising. In the most optimistic outlook, the hype surrounding AI subsides, allowing some applications of the technology to take hold. However, a more bearish scenario involves the AI frenzy continuing until 2025, only to result in a tech stock crash reminiscent of the dot-com bubble. In this scenario, disillusionment with AI could lead to a period of “AI spring followed by AI winter.”
The third scenario suggests the hype could persist for many years, with companies replacing human jobs with AI technologies without fully understanding how to utilize them. Eventually, these companies might scramble to rehire workers when they realize that AI technology does not perform as expected. Acemoglu believes the most likely outcome is a combination of the second and third scenarios. “When the hype gets intensified, the fall is unlikely to be soft,” he commented.
While Acemoglu is impressed with the capabilities of large language models like ChatGPT, he emphasizes that reliability issues will prevent these systems from replacing humans in the workplace for quite some time.
“You need highly reliable information or the ability of these models to faithfully implement certain steps that previously workers were doing,” he noted.
He further added, “They can do that in a few places with some human supervisory oversight, like coding, but in most places they cannot. That’s a reality check for where we are right now.”
 infrastructure may fail to deliver on their promises...){kind=link}