Is AI overrated or underrated?

0:00This message comes from WaterAid. Imagine having no clean water to drink, cook or clean. That's the reality for one in four people worldwide. You can change that. Donate now and your gift will be matched at WaterAid. dot org slash podcast. This is Planet Money from NPR. Greg Rozelski. Darian Woods. You write the Planet Money newsletter. You've been writing a lot recently about artificial intelligence and whether or not it's too overhyped. Yes, I have. But, you know, publishing this kind of stuff can be a little bit risky. Like, you know, if AI does end up transforming the economy, like I don't want to be the guy who, like once said, like predictions like that were overblown.

0:45Yes, you do not. So today, Greg and I are going to face off. We make the case for why AI will have a massive effect on the economy and for why it won't. And just so nobody confuses each segment with our own, you know, personal prognostications, we are going to decide who reports on which side with a coin flip. A digital coin flip. Let's get ChatGVT to decide. What? I didn't even know we were going to do this. This is freaky. So, Greg, heads, you take AI is overrated. Tails, AI is underrated. ChatGPT, can you please flip a heads or a tails randomly? Sure. It's heads. It's heads. Greg, you're going to look at whether AI is overhyped.

1:34Hello and welcome to Planet Money. I'm Darian Woods, co-host of The Indicator. And I'm Greg Rosowski. Today on the show, we bring you two episodes from Planet Money's daily show, The Indicator. First, I'm going to make the case for why AI, like chatbots and image generators, will have a huge effect on the wider economy. And then I'm going to argue the opposite. May the best human win.

2:01This message comes from Vanguard. Capturing value in the bond market is not easy. That's why Vanguard offers a suite of over 80 institutional quality bond funds, actively managed by a 200-person global team of sector specialists, analysts, and traders. They're designed for financial advisors looking to give their clients consistent results year in and year out. See the record at Vanguard.com slash audio. That's Vanguard.com slash audio. All investing is subject to risk. Vanguard Marketing Corporation Distributor. This message comes from NPR's sponsor, Odoo Business Management Software. Some say Odoo is like fertilizer because it promotes growth.

2:44Others say it's a magic beanstalk, scaling with efficiency. Odoo, exactly what a business needs. Sign up today at odoo.com. All right, Greg, I want to convince you that generative AI will have massive effects on the wider economy. Huh, okay, I guess I'm all yours. I'll start with some examples of how existing products and services are being made more productively right now with AI. Paul Daugherty is the Chief Technology and Information Officer for Accenture, which is an IT consulting company. We've had$2 billion of generative AI sales just partway through our fiscal year. So in a gold rush, you are teaching the gold miners where to look for gold.

3:27We are, yeah, we're maybe providing some of the maps as to where to go. Where do you find the value? And how do you get there exactly? Okay, so AI is great for consultants, huh? Yeah, that is true. I understand your cynicism, Greg. But Paul gives a bucket load of real examples. Take this government agency in Europe for pensions and welfare. They received 20 ,000 plus emails and letters every day. They didn't have enough people to even read them all. So generative AI could analyze the letters. It summarized what they were about, and it flagged those that required more attention. And it eliminated a backlog of 8 million lesses.

4:06It reduced how long people were waiting to hear back from that government agency. So six to eight weeks faster response time. Big increase. This would have taken a 700 % increase in their human staffing to do it if they had to do it with people. Also, there's the example of call centers. Classic AI use case here. Yes. So one company used it to understand what the customer was inquiring about and find the right solution. The result in this case was over a 30 % improvement in productivity, and it was over a 60 % increase in the measure of customer satisfaction. Paul says generative AI is also helping businesses expand.

4:46So take the insurance industry. Insurance might not think of it as the most exciting new industry to adapt technology, but in this case, they are at the forefront of applying generative AI. So when you apply for insurance, the company has to assess your risk. And for certain types of insurance, that's a lot of work. So let's take business interruption insurance and other commercial insurance. Paul says many insurance companies only take in about 20 % of new requests. And that's because they simply don't have enough qualified underwriters to assess all of the thousands of pages of documents. With generative AI, you can do that all a lot more effectively.

5:24And in fact, we're seeing them add new underwriters because they can now onboard people with this technology to help them become proficient much more quickly. And look, Greg, I admit these examples are not the sexiest. Yeah, it's no AI-generated video of like an alligator in space. Yeah, maybe that would convince you. But, you know, I would argue that this is the point. A lot of the boring behind-the-scenes processes that make our economy tick are being quietly transformed in ways that'll make all of our lives better and easier. Maybe, but we're not talking plumbing and electricity here. Bear with me, Greg.

5:58I've got more examples. So those were examples of how generative AI is helping existing products. There is a second category of how AI is reshaping the economy, which is completely new products. There's this company called Game Changer, and it has this product where you add in the scores for your kid's little league team, and then using a type of AI, it spits out this new super style article afterwards with a recap. Play-by-play coverage of your son or your daughter or the kids on the team. And yeah, you could envision all sorts of creativity and fun with this. I think it shows the promise and potential that the technology will have.

6:36You know, the wider implication of thousands and thousands of companies doing little things like this all across the world could be pretty vast. Tyler Cowen is an economist and a friend of the show. We're already using them to find new materials in the laboratory, to find new antibiotics, to crack problems in biology. Tyler sees those examples of what we're already doing now, and he thinks there is real promise ahead. It's going to increase the pace of scientific advance. So 20, 30 years from now, we're going to have much higher living standards because of generative AI. Tyler says generative AI has gifted us a tool that we've never had before.

7:15What is really scarce in our world is intelligence. Material resources matter, but the way you get more of them or get them more cheaply is to somehow apply intelligence. Now, for really the very first time in human history, we've created a fairly general kind of intelligence that for many tasks is already smarter than we are. Think of it this way. Every person has at his or her disposal now a free research assistant, a free colleague, and a free architect. That's not everything. It doesn't mean you have a free carpenter or free energy, but that's an awful lot. And just the number of projects that will stem from human ingenuity, I think it's going to astonish us.

7:57Tyler puts generative AI on par with one invention in Europe in the 1400s. I'm inclined to cite the printing press, which brought great benefits and a lot of knowledge and intelligence. It did make the world more chaotic in some negative ways. I think we need to work very hard to make sure we do better the next time around. But this seems to be a truly transformational breakthrough. You know, even if we're not seeing it exactly now, the way Tyler sees technology is that it takes time for it to diffuse through the economy. What we do now is we have processes already in place, and we tack the AI on top of those processes.

8:34And that can be a gain. But the real gain is much longer run when you reorganize how the whole thing is done. So to truly benefit the wider economy, all our processes, our norms, and our businesses need to adjust. Just because we had the technology of video conferencing software didn't mean that many people were working remotely pre-2020. The norms needed to change. There's human bottlenecks, institutional bottlenecks, inertia, fear. So again, that's a 10 to 20 year project, even in cases where the quality of the AI is good enough. And Greg, I'm not sure if you're aware, but Tyler has been using generative AI in his own work.

9:14He wrote this book himself. It was called Goat, Who is the Greatest Economist of All Time and Why Does It Matter? And it has this accompanying chatbot based on the book. Also on his podcast, Conversations with Tyler, he has this interview with the long dead Irish writer, Jonathan Swift. The answers were from ChatGPT and a human voice to the answers. I did enjoy the Jonathan Swift interview. Oh, yeah, yeah. He was very good. Yeah, better than many human guests. Not as good as you, though, Greg. Wow. Was there some sarcasm in your voice there? I'm saying there is the promise of AI, but it will have things it can't do, like the warm human touch of Greg Rosowski.

9:55Okay. Well, for a nice and warm human touch and a counterpoint to Darian's argument, stay tuned for the case for why AI is overrated. That's coming up after the break.

10:23This message comes from Babson College. Ever notice how one idea can change an entire market? At Babson College, graduate students learn to turn bold ideas into real-world impact, whether launching a startup or rethinking business as usual. In Babson's MBA and specialized master's programs, you'll gain the skills and experience to lead change. Ranked number one in entrepreneurship, Babson is shaping leaders who don't just study the economy, they build it. More at babson.edu slash grad school. This message comes from LPL Financial. What if you could have more control over your future? LPL Financial removes the things holding you back and provides the services to push you forward.

11:08Because when it comes to your finances, your business, your future, LPL Financial believes the only question should be, what if you could? LPL Financial, member FINRA SIPC, no strategy assures success or protects against loss, Investing involves risk, including possible loss of principle. All right, so I made the case for why AI is underrated. Now, Greg, it is your turn. Tell us, why is AI overrated? Darian, there are just so many reasons to believe AI is overrated. I could talk about the fact that productivity growth remains super disappointing. If AI were revolutionizing the economy, we would see it in the data.

11:46We're not seeing it. I could talk about the fact that AI companies have yet to find a killer app and that perhaps the biggest application of AI could be like scams, misinformation, and threatening democracy. I could talk about the ungodly amount of electricity it takes to power AI and how it's raising serious concerns about its contribution to climate change. But alas, I will hold off on those, Darian. Very disciplined. Instead, I will focus on just three reasons why AI is overrated. And I will be enlisting the help of an economist extraordinaire, the one, the only, MIT economist, Daron Asimoglu.

12:23Okay, so Daron Asimoglu, if you haven't heard of him, maybe you should Google him. Big deal. He's like one of the leading economic scholars of technology. So a lot of hype around AI. Oh, boy. Yes, you can say that again. Given that hype, should we expect AI to usher in revolutionary changes for the economy in the next decade? No. No. No, definitely not. I mean, unless you count a lot of companies overinvesting in generative AI and then regretting it, a revolutionary change. Ouch, those are some barbed words. That's right. I mean, I guess you could call that a change or like, you know, a huge bubble popping.

13:03Anyways, Daron, not bullish on AI's impact on the economy in the near future. And let's talk about why. Reason number one, this so-called artificial intelligence we have isn't actually that intelligent. I think most importantly, it's overrated because we are overrating its current capabilities. When you first use something like ChatGPT, it might seem like magic. Wow, a thinking machine able to answer questions or write or generate anything in an instant. But when you look under the hood, it's more like a magic trick. These chatbots are a fancy way of aggregating the internet and then spitting out a mishmash of what it finds.

13:42Okay, I have a slightly more optimistic view about the sophistication of the underlying technology, but I guess the metaphor you're using is a glorified photocopier passing human work as its own. I mean, basically, yeah, it's a copycat. And perhaps the worst part of it is a good chunk of the stuff AI is copying is copyrighted, which is why there are at least 15 high-profile lawsuits against AI companies asserting copyright infringement. Yeah, we actually have an episode about this. There's some artists that are suing AI companies. I've also read that AI companies have begun paying media companies for their content.

14:18Do you know, have AI companies used your academic papers and books in their models? And did you give them authorization? No, I didn't. I didn't give them authorization. And yes, they do. I don't even have to be talking to you right now. I could just talk to ChatGPT and have them regurgitate your thoughts in written form. They can do some extent. Although, you know, I did try it just out of curiosity, how it would summarize some my thinking. It's not horrible, but you'll do a better job, Greg. You'll do a better job. Oh, stop it, sir. So you can talk to yourself. You can talk to yourself. I'm just a humble homo sapien.

14:54He's flattering you, Greg. Watch out. Oh, yeah. I was blushing. You can't see that in a podcast. But yeah, sure. Generative AI, it is a breakthrough, but the hype around it has gotten out of control. The reality is the so-called AI we have now is not even close to the holy grail of AI researchers, so-called artificial general intelligence. What we have now, it's a bit lame. The technologist Dirk Hondel has called it autocorrect on steroids. These models learn from patterns in heaps of data to predict the next word in a sentence. So what's your perspective on that? Is AI actually intelligent? Is this all just like a misnomer?

15:33Yeah, I would have been so much happier if the whole word intelligence wasn't part of this discussion. The reality is these systems do not have judgment. They cannot tell right from wrong. They cannot tell true from false. Which brings me to the number two reason why AI is overrated. Reason number two, AI lies. The industry has come to call AI falsehoods and errors hallucinations. But like artificial intelligence, that actually also might be a misnomer because that makes it sound like, you know, it works well like always. And then every once in a while it drinks some ayahuasca or like, you know, eat some magic mushrooms and says some like trippy made up stuff.

16:12But this issue seems to be bigger and more common than that. Do you have any data to spot this? Yeah, well, there's one study that found that chatbots hallucinate somewhere between 3 % and 27 % of the time. Here's a real world example. Google was recently forced to revamp its AI overviews feature because it was giving users some pretty bizarre results. CNBC covered this. Kelly, some of those questionable, bizarre results centered around the AI overview telling users that you can eat a certain amount of rocks a day for your health. It's healthy to do so. Many AI researchers are saying we cannot end the problem of hallucinations anytime soon, if not ever, with these models.

16:53That's because they don't know what's true or false. Which brings me to reason number three why AI is overrated. Because AI is not intelligent and because hallucinations make it unreliable, it's proving incapable of doing most, if not all, human jobs. There are all of these jobs that AI was supposedly supposed to kill right away, like computer coder jobs or translator jobs, but AI is proving to be unreliable and too dumb to replace humans. Here's a recent example, drive-thru attendance. For about three years, McDonald's piloted a program to use AI at some of its drive-thrus. Yeah, I saw this. and it was not the best look for AI.

17:32Yeah, there were all of these viral videos showing AI making ridiculous errors, like trying to add$222 worth of chicken nuggets to someone's order. And then there was this instance where there was a bizarre addition to someone's dessert. You added bacon to my ice cream. I don't want bacon. Don't knock it till you tried it. I mean, it could work. I don't know. Maybe AI has something there. I don't know. But yeah, even the jobs AI was supposed to be like really good at, it's having big problems. At best, AI is proving to just be like a tool for humans to use. Humans still have to be there though.

18:09They have to probe the work of AI. And this has some like serious economic implications. Daron recently wrote an academic paper that analyzed AI's potential effects on the economy over the next decade. First off, he says, there are just humongous chunks of the economy that generative AI will barely touch, like construction or food and accommodations, factories, basically anything outside of an office AI won't even come close to being able to do in the next decade. Then Derone narrows in on office work and finds that there are just a whole bunch of tasks that current AI models are just incapable of doing.

18:43Derone finds that AI will impact less than 5 % of human tasks in the economy, less than 5%. And here, there will only be some mild cost savings. In the end, Durone predicts AI will not boost productivity or economic growth very much in the next decade. And here's the other big thing. People claim that AI is getting exponentially better, but there's actually mounting evidence that improvements to AI, it may be slowing down. AI companies may be running out of data to train their systems. There are all these problems with the microchip supply chain, and it's proving costly to create data centers and pay the power bills to run AI systems.

19:22So yeah, AI may not end up having a big effect on the economy, at least not anytime soon. And a big reason why is because human beings are just way more capable and impressive than these machines. So a lot of people in the industry don't recognize how versatile, talented, multifaceted human skills capabilities are. And once you do that, you tend to overrate machines ahead of humans and underrate the humans. That's right, Darian. It's not AI that's underrated. It's humans that are underrated. Go team human. Well, Greg, you've made your case. I'm inspired, but I don't know. I think we will see AI do more and more that we haven't even dreamed of yet.

20:05Well, Darian, I guess you got to keep dreaming.

20:11Hey, and by the way, me, human being, warm touch, I write the Planet Money newsletter. If you're interested in this topic and want to learn more, I wrote a newsletter that goes into way more detail about the reasons why AI is overrated. Check it out in our show notes. Also, subscribe to the Planet Money newsletter, npr.org slash planetmoneynewsletter. By subscribing, you are keeping human jobs in existence. These episodes of The Indicator were originally produced by Corey Bridges and they were edited by Patty Hirsch. They were engineered by Valentina Rodriguez-Sanchez and Neil Rauch. They were fact-checked by Sierra Juarez.

20:48Cake and Cannon is the Indicators editor. I'm Darian Woods. And I'm Greg Wazowski. This is NPR. Thanks for listening.