(Bloomberg) – Daron Acemoglu wants to make it clear from the outset that he has nothing against artificial intelligence. He gets the potential. “I’m not a pessimist when it comes to AI,” he says a few seconds into an interview.
Most read on Bloomberg
What makes Acemoglu, a renowned professor at the Massachusetts Institute of Technology, a prophet of doom for the growing economic and financial perils ahead is the relentless hype around technology and how it is fueling a boom in investments and a furious rally in technology stocks.
As promising as AI may be, it’s unlikely to live up to the hype, says Acemoglu. By his calculations, only a small percentage of all jobs – just 5% – are poised to be taken over, or at least greatly aided, by AI over the next decade. That’s good news for workers, yes, but very bad news for companies that are investing billions in technology hoping it will lead to increased productivity.
“A lot of money is going to be wasted,” says Acemoglu. “You will not get an economic revolution with this 5%. »
Acemoglu has become one of the loudest and most publicized voices warning that the AI frenzy on Wall Street and in senior management across America has gone too far. Institute Professor, the highest title for a faculty at MIT, Acemoglu first made a name for himself beyond academic circles a decade ago when he co-authored Why Nations Fail , a New York Times bestselling book. AI, and more broadly the advent of new technologies, have occupied an important place in its economic work for years.
Bulls argue that AI will allow companies to automate much of work tasks and spark a new era of medical and scientific breakthroughs as technology continues to improve. Jensen Huang, CEO of Nvidia, a company whose very name has become synonymous with the AI boom, predicted that growing demand for technology services from a wider range of businesses and governments would require up to to $1 trillion in spending to upgrade data center equipment. in the years to come.
Skepticism about these types of claims has begun to mount – in part because investments in AI have driven up costs much faster than revenues at companies like Microsoft and Amazon – but most Investors remain willing to pay high premiums for stocks poised to ride the AI wave.
Acemoglu envisions three ways the AI story could unfold in the years to come.
-
The first scenario – and by far the most benign – sees the hype slowly dying down and investments in “modest” uses of the technology taking hold.
-
In the second scenario, the frenzy escalates for about a year, leading to a tech stock market crash that leaves investors, executives and students disillusioned with technology. “AI spring followed by AI winter,” he calls it.
-
The third – and scariest – scenario is that this mania goes unchecked for years, leading companies to cut dozens of jobs and pour hundreds of billions of dollars into AI “without understanding what they are going to do.” make some,” only to find themselves scrambling to try to rehire workers when the technology doesn’t work. “There are now widespread negative consequences for the entire economy. »
Most likely? He thinks it’s a combination of the second and third scenarios. C-suites are simply too afraid of missing out on the AI boom to expect the hype machine to slow down any time soon, he says, and “when the hype intensifies, it’s unlikely that the fall will be gentle.”
The second quarter figures illustrate the scale of the spending spree. Four companies alone – Microsoft, Alphabet, Amazon and Meta Platforms – invested more than $50 billion in capital spending during the quarter, much of which was dedicated to AI.
Today’s large language models, like OpenAI’s ChatGPT, are impressive in many ways, Acemoglu says. So why can’t they replace humans, or at least help them a lot, in many jobs? It points to reliability issues and a lack of human-level wisdom or judgment that will make it unlikely that many white-collar jobs will be outsourced to AI in the near future. AI also won’t be able to automate physical jobs like construction or housekeeping, he says.
“You need very reliable information or the ability of these models to faithfully implement certain steps that workers were doing before,” he said. “They can do it in a few places with some human supervision” – like coding – “but in most places they can’t.”
“It’s a reality check of where we are right now,” he said.
Most read from Bloomberg Businessweek
©2024 Bloomberg LP