A top economics researcher is making the case that generative AI could be good for workers, as long as there’s a course correction in how businesses plan to use the technology.
Why it matters: The economic consequences of AI are a big unknown. But if this outlook is correct, the economy could see the upsides of the rapid economic changes AI might bring while avoiding the pitfalls of uneven labor market outcomes, like those seen during the automation boom of the 1970s.
What they’re saying: “The right way to think about generative AI is to view it as a flexible tool that’s usable by human workers,” MIT professor Daron Acemoglu said at an event hosted by the Group of 30 at the International Monetary Fund on Friday.
- “If we can do that — not just for managers and the top-level workers, but for electricians, plumbers, nurses, educators — I think there is a chance of turning this into a ‘pro-worker’ phenomenon,” Acemoglu told the group of elite current and former global officials.
- “But, unfortunately, that’s not where we’re heading.”
The big picture: The U.S. economy is in the midst of a productivity boom, driven largely by a recovery in the supply side of the economy that was disrupted by the pandemic.
- Economic policymakers say it’s unclear how long that lasts and what role AI is currently playing in this boom — and whether it could cause productivity gains to pick up further.
- AI optimists say the technology could improve productivity, in part, by handing off more menial tasks to AI.
Yes, but: Acemoglu pointed to the late 1970s, when automation hit America’s factories in a way that reshaped the industrial landscape.
- “People from Silicon Valley, journalists and some economists might be excused for saying, ‘Ultimately, things have worked out,'”Acemoglu said. “But ‘ultimately’ may be a very long time,” Acemoglu added.
Flashback: Acemoglu cited the widening earnings gaps between workers that happened in the years after the 1970s technology wave.
- “Low education groups experienced real wage declines, while productivity is increasing, while capital owners are doing well, and well-educated workers are experiencing rapid wage growth,” Acemoglu said.
- He added that some demographic groups saw their income shrink — in real terms — over the next 40 years.
- That was a result of other factors, including the rise in globalization, but “automation, in particular how we have used digital tools,” was an important part of it, Acemoglu said.
What to watch: What widespread adoption of AI means for worker outcomes ultimately comes down to the priorities of their employers — and whether the bias is toward automating as many tasks as possible.
- The alternative, though, is to use generative AI to complement (not replace) workers’ tasks, much like customer service representatives use technology to, say, more efficiently help callers.
- “We need to think of digital tools as useful things for humans — not our overlords, but our helpers,” Acemoglu said.
One roadblock to that reality is allowing workers to understand why AI tools give certain recommendations over others and then allowing them to combine that with their existing knowledge and expertise, Acemoglu said.
- “That’s currently impossible,” Acemoglu said, adding that large language models are akin to a black box.