(At the outset, economist Daron Acemoglu is keen to say that he has nothing against artificial intelligence (AI). He sees its potential. “I’m not an AI pessimist,” says he has barely started the interview in his office at the Massachusetts Institute of Technology (MIT).) If Mr. Acemoglu comes across as a Cassandra obsessed with the economic and financial perils that AI portends, it is because of the contrast between what he says and the incessant media hype which, according to him, boosts investment and inflames the stock prices of tech companies.
Gain of barely 5%
As promising as it is, AI will not live up to the expectations thus created, believes Mr. Acemoglu. According to his calculations, AI can replace barely 5% of jobs within 10 years. Good news for workers, certainly, but a cold shower for companies which invest billions in AI by banking on strong productivity gains.
“A lot of money will be wasted,” warns Mr. Acemoglu. Five percent doesn’t make for an economic revolution. »
Mr Acemoglu – one of the most credible voices among AI skeptics – warns that the frenzy on Wall Street and at the top of corporations goes too far. A professor emeritus at MIT, Mr. Acemoglu first became known to the general public 10 years ago, when he co-authored Why Nations Fail (Why Nations Fail), a bestseller of the New York Times. The advent of new technologies, including AI, looms large in his economic analyses.
AI enthusiasts promise that it will automate a large part of professional tasks. Thanks to its capacity for self-improvement, they also herald a new era of medical and scientific breakthroughs. According to Jensen Huang, CEO of Nvidia, the company that embodies the AI boom, the growing demand for technology services from the private and public sectors will lead to spending of approximately $ 1 trillion to modernize the equipment of data centers. data.
But these claims are increasingly skeptical: investments in AI have increased costs much faster than revenues at companies like Microsoft and Amazon. But most investors remain willing to pay a lot for shares of companies likely to ride the AI wave.
Mr. Acemoglu envisions three possible futures for AI:
– The media hype is gradually dying down and we are seeing “modest” uses of AI taking hold;
– The stock market bubble swells for another year or two, then its bursting leads to the disillusionment of business leaders, investors and students;
– The bubble doesn’t burst for years, prompting companies to make massive cuts in skilled personnel and invest hundreds of billions in AI “without knowing what they’re going to do with it”, then finding themselves left behind. running in all directions to try to rehire when the results are not there. This is the most serious scenario, because “in this case there is a widespread negative impact throughout the economy,” says Mr. Acemoglu.
Most likely? A mix of the second and third scenarios, he says. Business leaders are so afraid of missing out on the AI boom that it’s inconceivable that we’re taking it easy: “And when this kind of bubble intensifies, generally, the landing is not soft . »
Tens of billions in three months
The second quarter figures illustrate the scale of the spending: Microsoft, Alphabet, Amazon and Meta Platforms alone invested more than $50 billion in capital spending, much of it in AI.
Large language models, like OpenAI’s ChatGPT, are impressive in many ways, Acemoglu admits. Can’t they, then, replace humans, or at least help them a lot in many jobs?
There are reliability problems, he replies. And these robots lack judgment and wisdom on a human scale. So, it’s not anytime soon that many white-collar jobs will be outsourced to AI. The same goes for physical jobs, like construction or housekeeping, he adds.
“It would require extremely reliable information or the ability to faithfully replicate steps that were previously up to workers,” he said. “AI can do this in some areas under human supervision, like coding, but in most cases, not. »
“Sooner or later, reality catches up with us, and that’s where we are right now,” he added.