Is AI really coming for white collar jobs?

0:00Daron Acemoglu:Support for Here and Now Anytime comes from MathWorks, creator of MATLAB and Simulink software for technical computing and model-based design. MathWorks, accelerating the pace of discovery in engineering and science. Learn more at mathworks.com. WBUR Podcasts, Boston.

0:22Jon Finer:The uncertainty is not just about the technological capabilities, but about how we're going to manage this transition.

0:28Daron Acemoglu:Artificial intelligence is getting better, but it's human bosses who will ultimately make the call about whose job it can take over.

0:46Daron Acemoglu:It's Thursday, February 26th, and this is Here and Now Anytime from NPR and WVWAR. I'm Chris Bentley.

0:56Daron Acemoglu:Today on the show, negotiations continue in Geneva to avert war in Iran. There are plenty of hawks, however, like former U.S. National Security Advisor John Bolton, who say, bring it on. If there were ever a moment to try and help exacerbate the instability in the regime, destabilize it, and help bring it down, I do think it would be down. We'll get some perspective on the American pressure campaigns against Iran and Cuba. Also, Flavor Flav was more than a hype man during the Winter Olympics. He actually tried out some of the sports himself.

1:36Jon Finer:I wore my clock on the bobsled, but I couldn't wear it during the skeleton because the clock is on my chest and I'm laying down on the sled.

1:45Daron Acemoglu:That conversation coming up in about 15 minutes. But first, is AI really coming for your job? Or is the hand-wringing about a white-collar job apocalypse just hype? It's probably somewhere in between. But various stabs at answering that question are moving billions of dollars today. Like when a speculative report by the firm Citrini Research from Sunday sent some stocks into a sell-off. That report imagined a world two years from now when AI was replacing white-collar workers in a, quote, feedback loop with no natural break. MIT's Duran Asimoglu says, don't freak out just yet. He won the Nobel Prize in economics in 2024, and he's written a lot on technology and jobs.

2:35Daron Acemoglu:He told Scott Tong, even if the tech improves exponentially, that doesn't mean it has to replace human workers.

2:44Jon Finer:I think it is useful to think about these scenarios because the advances in AI models have been pretty impressive. But right now, much of this is still in the lab, so to speak, meaning it is not spreading to the productive sector of the economy. We are not seeing massive job losses, although some people are concerned about what's going on with entry-level positions. But, of course, we should be planning for how these models can be used and what their impacts will be.

3:17Daron Acemoglu:Well, I guess that's the question. If you or anyone kind of studies AI right now and tries to get a snapshot, does that kind of not take into account what AI could become? Some of the technologists say its abilities are going to be orders of magnitude better than what it is now as far as the possibility of replacing human tasks. Yeah.

3:45Jon Finer:I mean, let me give a two-part answer to that. The first is that what the Citrini Research Report outlines is not completely made up, meaning that there are shreds of evidence that indeed these models are advancing very rapidly. And we know from previous episodes of big shocks, including Chinese imports that led to a lot of job losses in many communities in the United States and automation based on robots that led to large job losses in many parts of the Midwest that used to be the heartland of manufacturing, you can have fairly negative effects because people lose their jobs, they cut their consumption, then that spreads to the rest of the community.

4:30Jon Finer:So that's not completely made up. However, right now, the capabilities of AI models are still not enough for them to take over wholesale large chunks of the economy. Moreover, there's a lot of choice about how these models are going to be developed further, how companies are going to use them. For example, if we were to use the capabilities of something like code to work with software engineers or to work with other white color occupations, the applications would be very different than wholesale automation and displacement of occupations. In fact, in the data, we don't see a decline in software engineering as an occupation because there's a lot of demand coming from partly from the AI sector for people with software engineering skills.

5:20Daron Acemoglu:Dharan Asamoglu, you wrote a paper last year that had a different take from this paper, that if AI indeed makes workers, makes companies more productive, it might actually be good for the economy. Living standards go up. Give me an example of what that looks like.

5:38Jon Finer:The reason why companies are so interested about AI and automation is that it's cost savings. Previously, you were spending$100 million, Now you can reduce that to$80 million. That's a 20 % cost saving. All of that$100 million used to go to workers. Now some of it goes to workers. Some of it goes to AI models. That would have adverse implications for the economy in the aggregate. For example, income inequality may worsen. But from the company's point of view, that's a good thing. If there is a lot of cost savings, that's productivity improvement. and productivity improvements ultimately can lead to better standards of living.

6:19Jon Finer:Now, my assessment was that the effects will not be as large and it won't happen as quickly as many other people are forecasting, but it will happen because these models do really have many capabilities.

6:31Daron Acemoglu:I'm just trying to understand your optimistic scenario. If AI gets good and companies get more efficient, they may not need as many workers, how is that good for the economy? Does that mean it helps me, the consumer? and I don't have to pay as much, and therefore I have more money to spend on other things, and then living standards go up?

6:50Jon Finer:Exactly. So the best-case scenario, you need, as an employer, fewer workers. That's how you save costs. Those cost savings are then reflected to consumers, and the workers who left your company go and find jobs somewhere else. Those two steps are important, and neither of them is automatic. So, for example, you may increase your markups and make more profits rather than passing that to consumers. Or the workers who leave your job may themselves remain unemployed, so therefore the productive resources of the economy, at least during a transition period, are not fully utilized. Both of these scenarios would eat into the benefits of AI, and some of them could become the source of the kind of feedback cycles that Citrine Research talks about.

7:34Jon Finer:But that's why the uncertainty is not just about the technological capabilities, but about how we're going to manage this transition.

7:42Daron Acemoglu:You make the case that workers need to focus on new tasks that can have human skills complement machine skills. We spoke on our show about this to AI researcher Kareem Lakhani of the Harvard Business School. I just want to play a clip from that.

7:57Jon Finer:Today, I have a worry about the AI divide between those that know how to use the AI effectively and those that don't try, don't put the effort in and make that happen. It's now, from my perspective, a matter of investing the time and effort to learn how these tools work and then to apply it to your work.

8:15Daron Acemoglu:So he's saying you've got to try it out.

8:17Jon Finer:Do you agree? 100%, but I think the responsibility starts with businesses, not just workers. Of course, workers have to get better at upgrading their skills for this changing economy. And there are legitimate concerns about our education system, for example, whether it's preparing youngsters for this changing world. But it has to start with firms as well. They have to find ways of using this technology in a way that's both productive and creating jobs. And that's not what we're seeing right now. Everybody is smitten by the idea of automation, automation, automation, and all of this talk of artificial general intelligence is contributing to it.

9:03Jon Finer:That, I would say, is a big concern. And if the Citrini research sheds some more light on that, I would say it's welcome. But I think the Citrini research itself is a little bit too optimistic about the capabilities of these models.

9:17Daron Acemoglu:That's economist Daron Asamoglu of MIT. He's the 2024 Nobel laureate in economics, and he's the author of Power in Progress, Our Thousand-Year Struggle Over Technology and Prosperity, written along with Simon Johnson. Duran, professor, thank you again.

9:33Jon Finer:Thank you, Scott. It was my pleasure.

9:40Daron Acemoglu:Coming up next, we check in on two stories where American foreign policy is dancing on the brink of war. Stick around.

10:00The U.S.

10:02Daron Acemoglu:says it's investigating a shootout that happened off the coast of Cuba. Cuban officials said a speedboat shot at border guard troops as they approached, injuring an officer whose crewmates returned fire and killed four people on board. Cuba's interior ministry called it a, quote, foiled armed infiltration. The boat was registered in Florida and carried Cuban residents of the United States, according to Cuba, along with assault rifles, handguns, Molotov cocktails, and camouflage uniforms. The shootings come as the U.S. said it would allow some shipments of oil to Cuba for commercial and humanitarian use, somewhat easing an American blockade that has been in place for over a month.

10:47Daron Acemoglu:Meanwhile, there's lots happening with Iran, which is in indirect talks with the U.S. in Geneva amid an American military buildup in the Middle East. For more on both those situations, we called up John Feiner. He was a national security advisor in the Biden administration and helped negotiate the Iran nuclear deal in 2015, which President Trump tore up during his first term. John Feiner spoke to Indira Lakshmanan. President Trump has kept up a steady drumbeat of threats against Iran. Here he was at the State of the Union address the other night.

11:21Jon Finer:We are in negotiations with them. They want to make a deal, but we haven't heard those secret words. We will never have a nuclear weapon.

11:31Daron Acemoglu:John, you are extremely well-versed on nuclear talks with the Iranians. Do you see room here for a deal, or are the two sides too far apart now? Well, you know, based on the line that you just played from President Trump, there should be room for a deal. President Trump believes Iran has not said these secret words that it will not pursue a nuclear weapon. Actually, Iran has said that many times, articulated it publicly, verbally many times. It put it in the preamble very specifically of the nuclear deal that Iran negotiated with President Obama that I was involved with as part of the negotiating team and that President Trump withdrew from.

12:11Daron Acemoglu:And the Iranian foreign minister repeated this just the other day, in fact, the day of President Trump's State of the Union address. So if he is looking for some sort of off-ramp in this drumbeat to war, he could claim, look, I've discovered or Iran has reiterated this commitment, and I take them at their word. It does not seem like, though, that that's where this is headed. It seems much more, and the State of the Union was good evidence of this, that he is trying to lay the predicate for possible military action, bringing to public attention a new justification for the war, which is what you said.

12:44Daron Acemoglu:the fact that Iran has a missile program that it is trying to develop that could reach the United States. So, John, though, I have to ask, Trump has said that the U.S. and Israel obliterated Iran's nuclear program in strikes last June. But White House negotiator Steve Whitcoff said as recently as Saturday that Iran is probably a week away from having industrial-grade bomb-making material. What do you make of that contradiction? Well, the administration is having a hard time getting its story straight. What you cited from Special Envoy Whitcoff was mystifying, I think, to many of us because, at least by all public information, and I don't think the administration has said anything to contradict this, Iran is not currently enriching any uranium.

13:28Daron Acemoglu:So if Iran is not enriching uranium at all, how it could get to weapons-grade uranium, which is what it would need for a nuclear weapon, is not at all clear. And Secretary Rubio came out soon after Whitcoff's statements and confirmed that Iran is not enriching uranium in the current moment. So I don't know how to explain that comment. It's a bit alarming given that this is the person who's actually doing this highly technical negotiation, at least one of the two people leading that highly technical negotiation. And, you know, he would have to explain what he meant by that. John, Secretary of State Marco Rubio last night said Iran poses a threat to America now, even without nuclear weapons.

14:06Daron Acemoglu:Do you agree with that? Well, I think the administration should explain exactly what that threat is, because it was not articulated in President Trump's speech other than what I described to you, which is President Trump's statement that Iran is working on developing this capability of hitting the United States with a missile, an intercontinental ballistic missile. As far as I know, Iran does not yet have that capability. The president did not indicate that Iran has that capability or necessarily that it is on the verge of having that capability imminently. My general view about these things is if you're going to take the country to a war that you are choosing, that war should be necessary and you should lay out the necessity of that conflict to the American people.

14:49Daron Acemoglu:The president had the opportunity to do that during the State of the Union address, did not exactly do that. So it is not clear right now what the rationale for war would be, or frankly, what the objective of that war would be, because they've tried a number of these possible objectives, re-obliterate the obliterated nuclear program, as you said, possibly punish Iran for the vicious crackdown that it conducted. And I wouldn't certainly deny how vicious it was against the demonstrators earlier this year, or potentially something more ambitious and actually trying to change the Iranian government, a sort of war objective that has a pretty infamous history in U.S.

15:26Daron Acemoglu:foreign policy. John, I want to turn to Cuba, where officials said that they shot a group of men who they described as armed terrorists on a speedboat from Florida. You were a top advisor to John Kerry when we went together on a trip to Cuba in July 2015, where John Kerry was reopening the American embassy when the U.S. normalized relations after decades. The country's situation has deteriorated significantly since then. Its economy is in crisis. Marco Rubio said last night that the status quo is unacceptable. What do you think is going on there, John? It's funny that you allude to that trip, Indira, which is sort of like the negotiations on Iran's nuclear program feels almost like ancient history today.

16:10Daron Acemoglu:The U.S. approach has evolved so far from that. What exactly is happening with this incursion into Cuban territorial waters is not at all clear. I think it may turn out, we may learn eventually that this is nothing more than people taking matters into their own hands, freelancing, certainly not anything necessarily tied to U.S. policy. But the United States is putting enormous pressure on Cuba. Obviously, the United States removed and is now bringing to justice the leader of Venezuela, Nicolas Maduro, who was one of Cuba's primary backers in its own region and providing significant amounts of energy into Cuba to allow for the functioning of its economy.

16:52Daron Acemoglu:The United States has stifled those energy imports, which is part of what is putting so much pressure on the Cuban economy in the current moment and leading to the humanitarian crisis you described. Whether that policy is related to this somewhat bizarre incident involving the boat, though, that case has not been made yet. Well, just very briefly, John, in the few seconds we have left, we know that Donald Trump wants regime change in Iran. We know he wants regime change in Cuba. Do you feel like something is up, that we should be watching this space? Well, look, historically, the United States does not deploy massive amounts of force to a particular part of the world.

17:32Daron Acemoglu:purely, simply as a negotiating tactic without at least some significant prospect that that force will be used. The president deployed what he described as an armada to the Caribbean to try to threaten and coerce Venezuela to take steps he was demanding. They didn't do so. He decided to intervene militarily. By all objective measures, the United States is seriously considering a significant war in Iran, probably something much more significant than the conflict that took place last summer, the so-called 12-day war. And again, my view is this sort of conflict should be necessary. That necessity should be explained to the American people.