The AI Revolution Is an Ownership Problem, Not a Tech Problem — Economist Justin Wolfers

My guest today is one of the most widely cited economists in the world and we're here to talk about AI. AI is a cognitive revolution. All of this technology over the past century has made blue collar work less valuable because machines are a substitute for brawn. But the folks who were getting substituted are now the guys with the white colors, not the guys with the blue colors. That's just totally different and worth pausing on. Justin Wolffers is a professor of economics and public policy at the University of Michigan and a visiting professor at the University of New South Wales. You may know him from Prof G markets with Scott Galloway and Ed Elson or as one of the top 25 economists under 45, shaping how we think about the global economy.

In this conversation, Justin and I explore why AI isn't just another technology but an economic revolution on the size and scale of electricity. White ownership and competition, not AI itself, will decide who wins and who loses and how each of us can adapt to avoid being left behind. What we have here is not an AI problem. It's an ownership problem. The technology is extraordinary and it can do work for us. In one case, the technology is owned by the worker and the worker gets freedom. In the other case, it's owned by the boss and the worker gets emissoration. It speaks to a very fundamental issue, which is how the AI is owned shapes who wins and loses.

Competition is actually saving us right now. We don't have to regulate ownership because when you have enough competition, the forces of competition prevent the worst outcomes. And so the personal advice that I give people is not to become an AI expert, but become just feeling. I am doing well. How are you this morning? Good. Where are you coming from? Is it the morning for you? Are you in England? Maybe it's the evening. That's an Australian accent. Don't make that mistake twice. Jesus. We're off to a great start here. No, no, we're not. No, I'm from the Midway. I live in Ann Arbor, Michigan. I teach at the University of Michigan.

OK, and I actually knew that. You know, it's 8 a.m. for me. I am not the... Oh, you're a let's first. Sorry. I am. Sorry to get you up so early then. Oh, it's OK. It's OK. It's good for me. It's not my natural, like, circadian rhythm to get up early. But yeah, yeah. So I watched your cold plunge one. No. And I woke up this morning and had a cold plunge. So that's what... Nice. That fixes my rhythm. First thing in the morning, for the coffee. Absolutely. I do the same thing. I mean, if I have to get up, that's my way I can function. Yeah. That's how you know we're both middle-aged blokes. That's right.

It does seem to come with the territory. These... How can you tell someone cold plunges? They just told you. Yes. Yes. Yeah. Yeah. Yeah. Oh, I see why you're such a hit on the Galloway already. Justin, Professor. Justin Wolfers, welcome to the show. Mate, let's go with Justin. OK. What's that? Let's go with Justin. We can forget the Professor Pat. OK. Fair enough. Fair enough. Thank you, Justin. Justin, I loved your conversation on prof g markets with Scott Galloway and Ed. And if anyone is interested in how tariffs are affecting the economy and the state of the economy right now, I'd highly recommend that to anyone listening.

But one of the things that you guys started to get into and didn't quite have enough time to cover was how AI is affecting everything. And I thought it was interesting that you said, right now, it's so easy to get distracted by that relentless news cycle and tariffs and things going on politically that we often aren't spending enough time thinking about the greatest technological development of our lifetimes. And so I wanted to jump in with you. What should we be focusing on regarding AI? What should we be paying attention to? Right. So I think it's not just the greatest technological development of our lifetimes.

It's that I think it potentially is one of the most important economic developments of our lifetimes. So let me just give you one way of thinking about the last 50 or 100 years of economic history. We figured out mechanization, where we have machines that can pick things up for us. Maybe this is the entire Industrial Revolution. So steam could move stuff that previously took people to. And so now when you visit a forward plant, there are these big things that can carry a car above you and that can then come and solder a car door on and all these things. That when you go to a farm, you don't see a person walking down the road, scattering seed and burying each individual seed.

What you see is these big harvesters that do this. What you're seeing there is technology replacing or doing the work of, sorry, doing the work of Braun, muscle, sin. Yeah. And that has been very, each of us in this beautiful Earth has been born with a set of talents. If you're a big, buffy bloke, muscles, what you got. Some of us are skinny, nerdy academics. And I don't have much Braun. But what I have is up here. So what I do is cognitive work. I write down economic models. I try and understand how the economy works. I teach. I have conversations for a living. Isn't this weird, Nick? You and I, we get to talk for a living so weird.

Well, you're getting paid on that. But many of us do cognitive work and the language, historically, we've used for this is blue collar versus white collar. And so all of this technology over the past century has made blue collar work less valuable because machines are a substitute for Braun. And that's been great for guys like us because white collar guys, I know I'm wearing a blue collar. It's very confusing because stuff that blue collar guys make gets cheaper when machines can make it instead. So it's good for white collar lives. It's bad for blue collar lives because if a machine can do your job, your employer is not willing to pay you as much.

And so that's been the story of the last 50 years, which is working class jobs have suffered wage stagnation and we've seen white collar jobs enjoy enormous gains. And so white collar is also often synonymous with college educated. And the difference between economic fortunes of those with a college degree and those without has risen tremendously through time. So why is AI interesting? AI is a cognitive revolution. It's doing what we do. It's the same thing, which is it helps us make a bigger pie with the same resources. But the folks who were getting substituted are now the guys with the white colors, not the guys with the blue colors.

So that's just totally different and worth losing on. So, you know, how do I feel? What do I, what am I good at? I'm an economist as a my particular specialty in economics is I'm a mile wide and half a mile deep, which I know a lot about a lot of things, but I'm brilliant on nothing. And my talent to the extent that I have one is reading widely and then being able to transform complex economics into single simple sentences. I hope I'm doing that now. If I'm not, that I'm not even good at that. What is chat GPT good at? Reading everything. Everything. It's better read than me. It's two miles wide and two miles deep and transforming it into simple sentences.

There's actually have an AI trained on my textbook. This is my textbook. You can barely see corner of up here. And maybe you should just run the interview with it because it knows a lot of economics and it can do it in the charming Australian accent. And so right now I feel like a Detroit auto worker in the 1970s. An auto worker in the 1970s saw the machines coming, the robots coming, saw their way of life under threat, saw their economic prosperity potentially disappearing, was told by Washington that this is all in the interests of creating a bigger pie and a bigger economy. And so just copper on the chin.

And they saw much of what they value disappear. And maybe there is actually more to value there than that of a white color academic, but it does mean that those of us who've been on one side of this equation for 50 years, we're now on the other side. I'm not saying we're more valuable. I'm just saying, wow, that's interesting. Eric, we should probably talk about it. Yeah. Will that drive wage stagnation and the worth of that type of work down for white color workers just like it did for blue color workers? I mean, it's the same economic forces. Maybe. So, you know, there's a million complications.

So one question is, is it a compliment or a substitute? Um, so let me tell you a completely unrelated story and then try and come back to that. But the question is, so compliment is, does it go with, help me do a better job or does it do my job instead? It substitutes to do my job instead. So you remember when the VCR came out, Nick, you're old enough. You will remember this. I do. I do. My aunt had a VCR. My whole family would get in the car and drive to my aunt's to watch a movie on the VCR. It was like really early. Only Jenny. She was wonderful. Um, and everyone said, well, that's the end of movies.

So they thought video cassettes were a substitute for going to the movies. In fact, it wasn't the case. It led more people to be more interested in the movies. Um, the Grammy Awards became central to our national conversation. And so in fact, people went to the movies even more. This turned out to be a compliment, not a substitute. So you can see how easy it is to think it's going to be one thing and it turns out to be another. That's the simple point I want to make there. So two blokes on a podcast aren't going to be smart enough to figure this out. So let's be humble. Um, but I do. So what do we know from it?

So so if you're thinking about, I want you to read the entire economics literature and summarize the literature on monetary policy. Uh, you could have called me five years ago and I would have done that for a five figure sum of money. Now you can type it into chat GPT and it'll be probably click clearer and cleaner and it'll cost you up to two cents. Um, so that does sound like it's going to drive down white color wages. Now, the thing to remember is the pie is going to get bigger, right? Because if we can do things more efficiently and we have the same inputs, then we must be making more stuff.

So someone's got to be getting it. Um, so the first thing you might think is well, now, Braun is valuable because someone's rich. That rich person really, really wants their garden to look beautiful. Uh, and a robot can't make your garden beautiful. It can mow the lawn, but it can't make it beautiful. So maybe we'll see that cause an increase in demand for blue color worker. And so that will narrow blue color white color differences. Maybe. So that would be a force for greater equality. Um, you might still be saying, well, if we're making more stuff in the white color guys aren't getting it and the blue color guys are only getting it downstream, then who's getting it?

And that's where that's a deep question. So my guess is you want me to tell the story of robots, Nick. Is that right? I do want you to tell the story of robots. Okay. So to the audience right now. Imagine for your next birthday, I give you a present. The present is a robot. Happy birthday to you Nick. Happy birthday. Here's a robot. This robot can do your job for you. You own this robot. Okay. My guess is Nick, you'd give me a really big hug. You'd say, thanks, Justin. Cause what I'm going to do is I'm going to take the Nickbot 2000 and the Nickbot 2000 is going to go to my job every day to all of my work for me.

Because my work's done, my boss will still pay me. So now all effectively what I've given you with the Nickbot 2000 is free time, so much free time. And you will write poetry and walk the beach or invent things or just play with your kids. But what an extraordinary gift if I can free you from 40 hours of drudgery, probably the greatest invention in the history of the world. Right. So let's now tweak the story. I'm going to call you a boss, Nick. Nick, do you actually have a boss? I don't, but for the sake of the story, let's pretend I do. Something that owns this whole podcast and the company I own and everything I do.

Let's, yeah. I was bad to say, I was surprised you didn't say your wife, but that might be asking more than you want to tell me. So what if instead of giving Nick the Nickbot 2000, I call Nick's boss and I say, hey, do you want to Nickbot 2000? And he says, yes. He now owns a robot that can do everything Nick does. Why would he employ Nick? So he's going to come back to Nick and say, I've got a robot that can do your job. I'm willing to pay you up to $1 a day to keep the job. Otherwise, I'll have the Nickbot 2000 do it for me. Mm-hmm. Nick now is poor. He's in the gutter. He's penniless. He's miserable.

And he hates technology because it just stole everything that was of value to him. I want to come back now and compare the two stories. What's the same in the two stories is the technology, the Nickbot 2000. The technology is extraordinary and it can do work for us. In one case, the technology is owned by the worker and the worker gets freedom. In the other case, it's owned by the boss and the worker gets emissoration. So what we have here is now the robot, to be clear now, is AI because Nick does cognitive work. So what we have here is not an AI problem. It's an ownership problem. If you own the AI that could do your job for you, then you'd still get your paycheck and you'd enjoy all this extra freedom.

If your boss does, you're out of work and out of luck. I like that story because it focuses the mind on what the problem is. There's a lot of people who are anti-AI, pro-AI, blah, blah, blah. This isn't really a question of what the controls around AI are or should be and whether you should use it at university or at work or any of the stale debates we have. But it speaks a very fundamental issue, which is how the AI is owned shapes, who wins and loses, to an extraordinary amount of money. That says, in turn, that public policy, our system of laws and the way we regulate AI, can have an enormous effect.

That in turn is why I say this is why this is the most important argument we're not having. And that's not the only thought experiment I could share with you. The point is to show you one thought experiment in which small changes in laws have massive effects on what this technological revolution, who this technological revolution was, and what the only thought experiment was. What is the most important thing that this technological revolution will serve, which in turn is just meant to convince you, let's have this discussion. Yeah, because there is a lot of discussion that this will be a technology on the order of magnitude as when electricity was first captured and utilized.

And we decided that was a utility, a basic utility. It wasn't that Edison, the company founded by Thomas Edison, got to own electricity for the rest of time. And I like the way you frame that as a question of ownership, because something that concerns me is the differences in how ownership might play out, depending on who gets their first with a extraordinarily advanced model of AI. If it is Meta and Mark Zuckerberg, who I don't know him personally, but just looking at his indifference to the societal harms caused by social media. And it's not like they have to wipe out social media. They could make changes to their algorithm to make social media less harmful, especially to young people.

And they don't, because it might knock a few cents off the share price. And that sort of indifference and meta profiting at the expense of everyone else is a frightening scenario of that who ends up with the ownership of this type of technology. And in contrast to that, I would say there's a, I'm sure I'm going to get his name wrong, but Demis Habass, who was the leader of DeepMind that was then, he sold DeepMind to Google. And he said, well, if I had waited five years to sell it to Google, I could have netted a couple billion dollars. But I thought that would give away five years of doing the most important research we can.

And the deal he made with Google is I'm taking a lot less money. But when, if my theories prove correct, I want what we develop from this to be given to the world. And to Google's credit, they went for it. And then DeepMind ends up inventing AlphaFold, which was the first AI, well, really no one had figured out how to, how proteins fold, and AlphaFold figured it out. And once they did that, they could have charged for access to that data in perpetuity and made a ton of money from it. And he said, no, we're going to give away the mapping of some 215 million proteins to the world. And it's already creating great science and advancing our ability to cure diseases.

And so I feel like those are the two examples of ways this might go. And let me be terrifying and one's exciting. Let me bum you out more and then come back and cheer you up more. Okay. So so far, when I told the story, either I give you the Nickbot 2000 or I give it to your boss. But of course, the Nickbot 2000 is itself produced. It's an AI product. Maybe it was produced by OpenAI, maybe by Mater, whatever. They're all engaged in a space race. If one of them wins, one of them produces the brilliant AI and patents it. Then we now have a monopoly provider of Nickbot 2000s. Let's say Google wins.

In fact, let's just say, let's say XAI Elon Musk wins. And he's now the monopoly provider of the Nickbot 2000. That means that your boss can either call you and pay you $500 a week or call Elon Musk and ask for a Nickbot 2000. Elon Musk is a monopolist. He realizes your boss's next best choice is to pay you $500. So he'll sell the Nickbot 2000 for $499. So now your boss buys it. You're out of a job again. Your boss is $1 better off. So both workers and capital is conventionally defined, gets screwed. And Elon Musk is rich as. And so we see a lot of technologies that end up monopoly-ish and can search.

That a lead in technology can be self-perpetuating. There's learning by doing sometimes there are network effects. And so the tech sector is full of these giants that are often ruled by themselves in a sector. And maybe AI has those characteristics. And if that's the case, then the monopoly provider essentially will gobble up all of global GDP because they'll sell the robot at the value it creates less a penny. The employers will buy that. The employers make a penny. The workers make nothing. Right? So it's actually worse than you think. And even worse than that actually, the monopolies, there's a standard economics result which is monopolists have a tendency to raise the price and sell less than is socially optimal.

So we'll actually get less AI use than we should. They charge high prices monopolies. And so therefore fewer bosses use them at all. And so GDP might even be smaller. Okay? There's a weird and I think frankly miraculous thing happening right now. There is genuine and profound competition in AI. There's open AI, there's enthropic, there's matter, there's deep-seek, there's a long-list. And it's genuinely competitive. So what is the price of AI right now? Well, in a competitive market, prices fall, it's always worth servicing another customer as long as you can get a price at least as high as your marginal cost.

The marginal cost of a thousand queries might be at one cent. And so therefore the price of a query is a fraction of a penny. And that's actually where we are right now. Like it's kind of extraordinary how cheap open AI's products are and enthropics are. And in fact, those prices are falling, they're halving every few months. This is a dramatic deflation in the, well, it's not deflation because it's not general. It's a dramatic price reduction. It's sort of a Moore's law type thing going on, but even more rapid and even more dramatic. So because of the miracle of competition, Elon must can't charge your boss $499 for the NICBOT 2000.

In fact, he and his rivals will keep underbidding each other until they get down near their marginal cost, which is awfully close to a penny. And so now they sell the NICBOT 2000 to your boss for a penny. And now at least your boss makes a profit. So competition is actually saving us right now. We don't have to regulate ownership because when you have enough competition, the forces of competition prevent the worst outcomes. How confident are you that competition will continue? I don't know. And so another way of saying this, remember a moment ago I said it's not that we have an AI problem, we have an ownership problem.

Now I think some of the most important consequences would suggest it's not that we have just an ownership problem, we have a competition problem. If we don't maintain competition in this space, someone will monopolize it, gobble up all of GDP and maybe even slow innovation. So now regulation of competition in this space seems really, really, really, really important. So that's the next debate we should be having. And by the way, competition doesn't have to be just at the level of many AI providers. So we're in this funny world right now where there's enormous competition between open AI and anthropic and all these others, but they all use Nvidia chips.

So all of a sudden you've got competition at this level and down in the lab market you've got competition, but you've put a monopolist in between them. Now Nvidia can say I will only sell you the chip that will create the Nickbot 2000 at number of the believably high price. So instead of Elon Musk getting all the world's riches, Nvidia does. And just in case this sounds like a weird economic theory, it's basically what's happening. I mean, I'm overstating the case that Nvidia is the most valuable company in the world right now. I believe so. And they don't have, to your point, a direct competitor at this moment.

Right. And so it's not just enough to have competition between competing AI firms. You need competition in all the essential inputs, which at the moment is graphics cards, but it could be other things. You know, later on it could be electricity or other things. But again, I just want to make the point that small changes in market structure, whether there's an Nvidia competitor or not, fundamentally transform the role that this technology will play in our lives and who wins and who loses as a result. This episode of the Nick Stanley Show is brought to you by Zapier. If you've ever felt buried in repetitive work, copying data, moving files, sending follow-ups, you know it's like death by a thousand mouse clicks.

Zapier has always been the tool that fixes that. It connects over 8,000 apps, Google Drive, Slack, Notion, Gmail, MySpace, you name it, so your tools can finally play nice together. But here's the big shift. Zapier now lets you create AI agents with their chance EPT integration. Think of them as tireless teammates who never complain, never take lunch, never get bored of doing the boring stuff. I've actually made several of these little agents myself. No Python, no JavaScript, just pure vibe coding, which is my favorite kind of coding. Because even though I have no idea how to code, I just talk to the AI, tell it what I want, almost magically, works.

For example, when a podcast guest books a time, Zapier can send them a personalized email from me preparing them for the show, create a draft of show notes, update my calendar, and even prep a social media post automatically, all of which frees me up to focus on the important stuff, like taking credit for the amazing work my agents did. And now, Zapier just took it up another level for developers. Zapier MCP is available right inside Chat GPT, which means you can connect those 8000 plus apps and trigger workflows just by writing what you want to happen. Literally, you tell Chat GPT, send this file to my team in Slack, update the spreadsheet, and draft an email, and Zapier plus Chat GPT, figure out the right tools, do it for you.

Getting started is simple. Head to the Zapier Chat GPT MCP server and add the tools you want Chat GPT to access. Follow the steps in the Connect tab. And if you're an admin on a Chat GPT enterprise account, you can enable MCP across your entire workspace. So if you're ready to stop wasting time on busy work, join the AI revolution and make a little automation magic of your own. Why the Zapier Chat GPT integration using the link below. What's your sense on the markets right now? Not that we're giving investment advice. We do not give investment advice on the show. Have to put in that disclaimer.

But you have these incredible valuations of these companies involved in AI. They must see that they're going to be, well, not see, but suspect, anticipate that they will be hyper profitable down the line to justify these incredible investments. And what are the odds that they reach that profitability horizon in time to justify the valuations? Because in looking back through a little bit of history, I was surprised to find that with, say, the railroads, which were a transformative technology, there were a lot of people who lost their shirt, seeing that railroads would change transportation economies.

And they did. But the initial investments didn't all pay off. A few people got really rich. A lot of people lost a lot of money with investments. And we had the same thing with the dot-com bust right around 2000. You had people that could see the future. They did see the internet would change the way we live and work. And yet there was this huge bust, a bubble and then a bust in those markets. I mean, I was shocked to find Amazon dropped 92% of its market capitalization. Almost went bust just before they essentially inherited the world and took over retail as we know it and became one of the most profitable companies in the existence of capitalism.

Is something similar happening in these markets right now where there might be a bubble and a bust? Any thoughts on that? I'm going to just start by canceling humility. If you and I knew it was a bubble, you and I would not be talking about it on a podcast. And if you and I had a long track record of spotting bubbles, we might be having this conversation, but it would be on a private island. And anyone who thinks they know what's going on, who is not on a private island, is overconfident. So the other thing to realize is it's not that the bubble has to, it's not that the AI boom has to exactly follow a particular script.

It might just be enough. There's a possibility. And is there a possibility that it works out that way? So let's go back to the dot com bubble because it's kind of interesting. If you interpreted that as a bet that the internet will be transformative economically, it was. So it was right. If you interpreted it in a little more granular level that founding a company called pets.com that had a sock puppet that would remind you that pets can't drive and that you would make money by shipping 50 pounds bags of cat litter to people's houses and not charging them shipping, that seems obviously absurd on its face.

Yeah. So the worst excess is, I think we're obvious then, and are utterly obvious today. Having said that, Amazon used to be a bookseller. So who would have thought a bookseller would come to dominate global commerce? So was there a possibility back in 1997 that the internet would be incredibly transformative and those valuations made sense? Maybe. Did it look to have achieved that by 1999? No. 25 years later, is it obvious which side was right? I don't know. So let's come back and just ask, you know, it's really trendy. If I wanted to become really probably, I'd say, Nick, it's a bubble. Believe me, I'm very smart and I'd use all my white man confidence.

And there's a lot of it. But I want to counsel humility here. So is it possible these valuations make sense? Here's a very simple adding up economist way of doing it. Work down and take every job in the American economy. And then for each of those jobs, say, potentially AI could do X% of this job. So for you and I right now, we're doing a podcast talking about economics. Google already has a product that can do this. What's Google? Google? No, Google? No, Google. Right. And so you could go through when you could do this, right? So you would say mowing the lawn, robot mowers can mow the lawn, flower arranging, no, they can't do that.

Short order cook, that doesn't seem like an AI thing. Translator, it does. And so there are a bunch of education. My sense of many LLMs is they're clearer teachers than many humans. There's a different question. People learn from bits of silicon as opposed to flesh and blood. So we don't really know whether all teachers are replaceable, but maybe they are and maybe they're not. And you could do this through every single job of the American economy. And you're going to come up with some number, right? Let's do you want to make up a number? I'm happy to. But if you want to make one up instead, we can use yours.

I feel like your number would probably be better than mine. Okay. Let's say a quarter, a quarter of all work, like all the financial analysts are gone, all the translators are gone, half the teachers are gone, all the podcasters are gone, the people who break up your clips, Nick and make tiktoks out of it, they're gone. Blah, blah, blah. Right. The US economy produces $30 million a year, $30 trillion a year. So that's $7.5 trillion worth of output now being done by machines. Maybe it costs $500 billion to produce that. So that $7.5 trillion becomes $7.00. But that's just in one year. Right. And this is forever.

So let's discount at a 5% discount rate, which is way too high, but I just want to keep the math simple. So now I'm going to take my $7 trillion, multiply by 20 and I'm at $140 trillion. What is the valuation of all these companies right now? It's not that right. Right. So that says it's not hard to think of a world in which that much value is being created by AI. So it's possible. It's all I'm saying. Maybe you could have one of your WizKids look up the combined value of the Magnificent Seven right now. I don't have it in the top of my head, but I know it's going to be less than $140 trillion.

Right. Without a doubt. I mean, there's a lot of AI. You could even ask an AI if you want. You could even ask an AI if you want. Right. Right. I actually could real quick. Yeah. I'm just going to make for absolutely compelling viewing as we both type into our eyes. Oh, it's OK. We'll cut a little bit of this out just the dead time. But yeah, AI already cuts up the shorts for this show using clips. Yeah. Yeah. Yeah. Although it does require some human oversight, but it's getting better all the time. OK. I've got a number of 20 trillion, 20.3. Yeah. OK. OK. So what we may just not have told the viewers is we just dropped out and asked AI what is the total market cap of the Magnificent Seven?

What did you get, Nick? I got 20.3 trillion. And how much did we say it would make in the US? It could produce $140 trillion worth of value. Oh, I just made a mistake, Nick. I'm only talking about the US. Right. The US, I'm going to make up a number. I'm going to say is roughly one-fifth of world output. So let's take my $140 trillion multiplied by five and we're at $700 trillion. Can that $700 trillion in value creation potentially support a total market cap of $20 trillion? Yeah, you better can. Right. So look, the point is not to give investment advice. You already told me not to. The point is just to say it's plausible this all makes sense.

Yeah. And every number I gave you in our calculation, 25% of tasks, $30 trillion, one-fifth of global output, you can quibble with any of them, but I did it out loud so that you can put your own number in. Mm-hmm. I guarantee whatever your number is, it's going to make it plausible that these companies are worth $20 trillion. Now, another possibility, by the way, is that all those companies compete like crazy in a perfectly competitive market. You can produce a lot of value, but make no profit. And so it could also be true, we have an AI revolution and people get rich. That's when we give you the NICBOT 2000.

Right. Right. And the companies don't get rich. In which case that market cap makes no sense. Let me reverse course. A different thing to say is, markets are often pretty smart. Markets reflect the aggregated wisdom of lots of different people, all these WizKids on Wall Street with their fancy suits and their computer models and so on. They're making big bets as if they believe there's a substantial chance that AI could be transformative. That's the whole sentence. That's all I want you to say. So next time someone struts around the schoolyard and says, oh, this, it still makes mistakes and it's not very good and it's not going to change my life.

I'm just going to say a bunch of sophisticated people believe otherwise. And in a minimum, I think we should take them seriously. So let's come back and talk about AI. Yeah. You mentioned there the mistake of thinking about only the US economy there. Now, I mean, China and the US are leading the race by leaps and bounds worldwide with the development of this technology. Where does that leave the rest of the world? So one answer is, okay, here I'm going to shock you. It depends. Yeah. But so one answer is it doesn't matter at all. So I'm in Australia. We didn't invent the Nickbot 2000. But if I can still buy the Nickbot 2000 and it can do my job for me and if it's being sold at a reasonably good price, it transforms the Australian economy.

Yeah. If I can't buy at a reasonable price because there's no competition, then I might buy at a terrible price and Australia gets poor and America gets rich. So again, it depends on the market structure. Depends on how much competition there is. You might think, so I am less worried about where it's produced. Now I'm not a national security guy or anything like that, right? As the number of human beings working in AI is tiny. This idea, as I understand it, came out of five guys at Google sitting around thinking about math. Australia's got five guys too. Maybe they'll come up with something altogether different.

And by the way, it's also possible that China creates the best AI but they need Nvidia, in which case it's America, the profits. So again, I just want to come back to it. It depends on the structure of markets, not just the market for AI, not just the labor market, but also the intermediate markets in between. I will say when I look at the current state of economic policy in the US, where there is an obsession about bringing back manufacturing. And one of the claims is that this will help our national security. I can't imagine a more important national security tool than something that can process vast amounts of natural language.

Think about that as a spying tool, for instance. So where I had to be making, I'm not a big fan of making national security informed bets and don't have much about national security. But it seems likely, if I've given you economic reasons to care about AI, I'm willing to also say, I think defense says you should as well. So again, this is the, it might be the most important defense discussion we're not having. Right. Because it does seem like the future of warfare is one country's drones attacking another country's drones. There could be one country's bots typing on another country's Twitter. There are so many ways this could go, but it's hard to believe that what we want to do is subsidize American saddle manufacturers rather than technology firms.

So speaking of that, what mistakes are we making right now in government? One obviously, there doesn't seem to be a lot of open discussion about all of this, the ownership issue, the competition issue, but just in terms of actual policy, what are you seeing and taking note of right now? So I think the United States in lieu of a policy that's actually a policy, the absence of a policy is a policy, it's laissez-faire. Laissez-faire makes a lot of sense in a lot of contexts. And if we were talking about the market for t-shirts, I'd be 100% in favor of it. When you're talking about the possibility of monopolization, when you're talking about enormous shifts in distribution of income, the distribution of work, future meaning, the role of people in society, I suspect the answer may not be laissez-faire.

So that just says, I understand I'm saying the same thing, do something. I understand that's dull because you're asking what, and that's what I'm going to remind you, I'm not very clever. I think that's a great place to start this. And if you would, just real quick, laissez-faire, just define it real quick. Laissez-faire. Oh, laissez-faire. Oh, it's how you sound French and you impress a date. Laissez-faire just free the market. Yeah. Let it rip. Let's just get out there and beat each other up. Now, of course, by the way, we're not actually pure laissez-faire, which is we have a patent system that protects the development of intellectual property.

And you could say either that's not just free markets because everyone should be able to use ideas or you could say at the other end, actually ideas are owned and it's unfair that they go into the public domain after 20 years. I am struck by how the same fundamental technology, the transformer algorithm, is somehow fueling many different companies. That does sound to me like that underlying idea is not protected even if expressions of it. Okay, so what else do we need? We probably need international cooperation. You know, this is the, you can't have a public swimming pool where you can only pee in one end.

Whatever happens in America shapes the world, whatever happens in the world shapes America. I think I haven't talked about yet is, you know, if you look what happens with social media and the deleterious effects, people believe it too much. They lean into it that's used as a tool of influence. Right now, I believe what chat GPT tells me, but should I? Could that be manipulated? Is it being manipulated? So there's a whole range of questions there. You know, you don't only have been manipulated in little ways on a few topics to keep it almost undetectable and have a huge influence. Absolutely. And not only is the question, I mean, you raised the question is it detectable and with statistics, we can find tendencies and we're all very good at that, but we don't even have detectors set up though.

If I typed into chat GPT, who's the greatest president of all time and, you know, 51% of the time it said President Trump, I might sort of wonder whether it makes sense. And I'd have to actually be checking if it's 51% or 14% or 62%. We don't have any institutions checking. And actually, it's a really nice question because in training data, it probably is the case. More people have said President Trump is the greatest president of all time than said that of President Biden. I think that is more a statement about governing style than substance. One of them demands subservience. The other does not.

Right. And so, you know, there needs to be, okay, so what else do we need to do? Let me come back because the answer is a trillion things. So I'm going to start with the things that I teach. So the first, I teach in university. The first thing we need to do is understand students have access to chat GPT. I have been absolutely depressed by the unwillingness of academics, not uniformly, but too often, to acknowledge the simple reality students have tools they didn't once have and will treat. And my view, the advice I give my fellow colleagues is at home, high stakes assessment should be over. Never have them again.

Yet, if you want to get into Harvard, your kid will write four admissions essays. Mm-hmm. How many of those are getting written by chat GPT? Does that mean Harvard's now going to lead in the cheats rather than the imperfect, honest people? And so that it is just absolutely trivial way we're failing to rise to the moment. And any of your listeners who have kids at college will recognize this. There's a somewhat deeper question. OK, so this is like phase one, recognize the threat part, right? And so we should have understood this three years ago and I've been giving online webinars to fellow economists, trying to make this point.

And I've had some success, but it's been wildly imperfect. Phase two, can this help me do my job better? Right? So we know lots of things that students struggle with. One, they struggle to talk to their professors. They're intimidated by them. We also know that students often have problems that need to be solved at 3am, but their professors are asleep. So as part of my economics textbook, we have designed, as you're working through your problem set, we've got a little AI that's basically me on steroids. Well, you can say, hey, I'm not really understanding question three. And it'll repose the question.

It'll give you an example. It'll never give you the answer, right? And so this is a way in which we can make an AI a compliment to what we do. We can turbo charge. So I am in my students room at 3am. I mean, I'm not the Justin Bot 3000 is. Yeah. And that's kind of cool. And once you start thinking that way, there's a million trillion things you can do. And I'm developing a bunch of tools in my narrow slice, which is teaching economics. And there's probably a thousand flowers blooming, but there's also sort of many that he has yet undiscovered. And then there's a much deeper question. I teach introductory economics.

And if you took the results, if you took the questions on my exam last year and put them into chat GPT, chat GPT would have got 100% literally 100. Does it make sense then for me to test people on stuff that they could ask a computer about? What is human knowledge in an AI world? Right? So we have analogies. The value of knowing your times tables is much lower in a world with calculators. So we spend less time on times tables. But equally, a lot of people are suspicious of spending no time on times tables. And I think actually for very, very good reasons, right? Sometimes you just need to do the mental math or you need to be able to tell when did I frame the question the wrong way because the calculator is giving me an answer and it doesn't make sense.

So then what does it mean to learn economics or literature or political science or biology or physics or medicine in this world? That, you know, our universities create enormous value. But if we keep creating education of the form that was necessary in the year 2000, rather than skating where the puck is and thinking about what skills are necessary in the year 2040, our value goes from trillions of dollars to much, much, much less. So why aren't universities in the midst of a revolution? And I don't want to say that university is a special. I'm talking about my industry because I understand it.

So then the question becomes to each member of your audience right now, whatever industry you're in, why isn't your industry in the middle of a revolution? And you might say it feels unstable and it feels like everything's changing. But really at a deep level, we've inherited how we do business from history. Maybe little of what we inherited is relevant for this new world. Have we kept too much of it? And my guess based on the case study of higher education is yes. Now you see here some tech firms now that are like, we will never hire another person unless you tell me why I can't get an AI to do it.

And that's the most extreme view we have at the moment. Probably still not extreme enough in my view. So you know, there's a lot of work to do everywhere. We've got to understand the tools. Let me go back to AI literacy 101. You've probably read a million think pieces by some mid-tier journalists who sits down with an AI and it doesn't do anything amazing. And then they declare it a pointless technology. And the problem isn't the technology, it's the user. If you type into AI, write me an essay about blah, it'll write you a blah essay. But if you ask it to write in the voice of Ernest Hemingway, and to be as persuasive as possible and to draw on the following resources and aggregate this and do that and so on, it'll be incredible.

So we have an industry of think pieces by people who don't know how to use AI. And they're getting through because they're from editors. Have a vested interest in a world in which the talents of editors are highly valued. Yeah. But actually editors are the next ones to go. Sure. They're after the translators, but they're not far off. So we've all got work to do, brother. In education, I think that's a really interesting space to look at some of this stuff because it strikes me as wrong to think about that students are just cheating if they use these large language models because I use them every day for work.

And when you know how to prompt them well, just as you illustrated, it makes all the difference in the outputs. And I feel like that would be an extremely valuable skill to teach within primary education and higher education. And yet at the same time, I'm struck by this paradox where learning how to write well helps you to think more critically, to understand your own thought process just by getting it down. And so that is still an essential skill that needs to be taught. I mean, an example that comes to mind, you mentioned the 5% discount rate when you were calculating the size of the US economy.

And I had a finance class in business school where the professor wouldn't allow us to use spreadsheets to solve problems like that. Figure out the discount rate and blah, blah, blah. And everyone hated the class. I mean, it was students were complaining, making them, they thought it was ridiculous that we couldn't use these tools that we had available to us. And yet I will say coming out of that class, I developed an understanding of what these actual financial mechanisms were that I would not have if I just punched it into a spreadsheet, got the answer, got an A on the test. And so at the end of the process, I really understood why the professor wanted us to learn it the hard way, then go use the tools.

And by the end of the class, that was the big reveal, the big surprise is like, okay, now you have an understanding, now start using these tools. And at the time, we were just talking Excel spreadsheets to solve more complicated problems. And some sort of marriage between those two things seems to be the revolution that needs to take place within education. I think that's right. And I just want to add, and it's hard to be confident. Yeah. Yeah, it is. So if what we said was these are tools students going to be learning in the workplace, therefore what university should do is teach the use of AI.

You actually said high schools or elementary schools. But we could create a class on prompting. It would be about three hours long. And at the end of three hours, you'd be not terrible. Right. And if someone wants a good reference, by the way, I don't get paid for this. If you want to move from zero to being part of this conversation, I recommend Ethan Mullick's book, what's Ethan Mullick's book called Co-intelligence? It's just a wonderful primer and introduction. So but three hours later, I've sort of taught you how to prompt. So either I declare higher education now is three hours instead of four years, or I've got 3.99 years to fill in.

Yeah. And that's where I think you're right. And this sounds like the sort of the old liberal arts education defense. I'm not sure I mean it that way, but sometimes it's knowing what question to ask. Yeah. And I've got to know a lot of economics or you've got to know a lot of finance to know what is sort of the heart of the matter. And so that's not a question of prompt engineering, how to write the prompt. It's sort of what am I trying to get out, right? And you sort of got to know a lot of economics to get something great. Now I want to come back, I want to just knock over a few things that I think are clearly wrong, not that you said but other people do.

So there's some really light and easy answers that I think are done. So one is university should just teach prompt engineering and then move on. Okay. You can do that. Once you've taught prompt engineering for three hours in the first day, then the student can get through the rest of their college degree, hitting control C and control V. That would mean an entire degree. Right. I'm not sure we should give a degree for three hours of work. Yeah. Some people are like, I know what we should do. We should have students practice what they're going to have to do in the workplace and what they're going to have to do in the workplace and what you do every day, Nick, is you put stuff into the LLM and it gives you a bunch of output and then you evaluate, improve it and correct it.

Guess what's really good at doing that? Another LLM. Yeah. So the idea that you can do anything that way I think doesn't make sense. And then you can say, well, let's make it an optional tool. Now, here's the problem with making it an optional tool. An LLM never produces work below an A minus level for an American undergraduate. That may be a statement about American undergraduates. I don't really care. What that then means is no student can afford not to use the LLM. Right. Because everyone else is at least an A minus level. It also creates a tremendous problem for the faculty member because people say, well, if people have new tools and can be more productive, all we need to do is change our grading standards.

But that means that if a student submits something that's an A minus, I know that took one line of prompting. Therefore, I should fail that student. They have to be producing A double plus work. Right. If I fail a student for submitting A minus work and I say, no, that earns an F and their parents come and see my Dean and the Dean reads a beautifully written essay, a guarantee to you the Dean's not on my side. The world just isn't made for everyone, as it currently is constituted, for everyone to have the capacity to produce pretty good, at least pretty good work. So I'm still at something of a loss.

I don't have the answer. I can tell you the tool that I'm developing. It's super cool. I love this. So one of the answers people give is, well, the problem is that we use standardized tests in America. If you go to Oxford, the way an Oxford's education works, is you and two other Oxford students sit in your lecturer's office for what's called a tutorial and you will have written a 1000 or 2000 word essay. And then they'll say, Nick, what did you mean in the third paragraph? And I'll drill you on it like that. Right? It's very intense. People say, well, that's what we should do now in America. So first of all, it was an economic problem.

I teach 500 students at a time. So that means the Dean would have to pay another 199 times my salary to do that. So the economics of that don't work. But actually, if you think about the whole, that the lecture is playing there, actually an LLM could do it. And so what we're developing for my classes is something that looks like an Oxford tutorial where I sit down. Wow. The computer says, hey, Nick, do you do the readings? You're like, yeah. And then rather than asking you to set a standardized question to control C, control V, hey, can you think of an example that it has something to do with why demand could have slipped down?

Now you could still cheat. But one of the things is if we make it a conversation, then the returns, the cost of cheating is still to paint in the ass. And the benefit is your grades going to be like 0.001% better. So we're just trying to change the economic incentives for whether you want to cheat. I'm not sure that this is right. As I said, we're experimenting, 1,000 flowers, bloom and so on. But that's what students who use my textbook hopefully going to be able to do within a year or two. So I do think you're right to focus on education. This is a sector right for disruption. It's one that traditionally moves slowly where folks like me have tenure and therefore very little incentive to wake up to the world that we see.

And it's one that I think is remarkably well suited to the strengths of LLMs. It's cognitive work, right? So we're in the right workplace. So will it change? Higher education has been a sage on a stage, a bloke in front of a blackboard for about 800 years in a row now. And about every 30 years, someone says it's over and they've been doing that for 800 years. Could be over this time. I could be wrong. Who knows? Yeah, it strikes me as more important than ever before. If you're going to be able to have unique thoughts and ask unique questions. So I guess, as you said, it really is a defense of the old school liberal arts education.

I mean, it could be a swing back to that being the most important skills just from a individual selfish incentive point of view. If I want to be able to earn a living and find a place where I can be augmented with my talents, can be augmented with these tools, then a liberal arts education that teaches me to ask the right questions and think critically about them. Maybe the biggest advantage I can find in higher education. It might. I'm feeling you're saying it might not. Yeah. So first of all, studying philosophy is not going to help you figure out the economy any better. But it's still the same idea, right?

You've got to learn some economics to know what to ask about the economy. I think I mostly agree with you. Now let me put an asterisk next to that for one generation and it's the generation currently going to school. With any technological revolution, they take a huge amount of time to actually do their work. The Industrial Revolution took decades. We had to reorganize factories, electrification likewise. The PC revolution, we didn't see it in the economic data for decades, which is we had to fundamentally change how we do and conceive of and think of work. And so if you used history as a guide, you'd say, oh yeah, AI could be big, but that'll take two to three decades, maybe even longer.

In which case, the real economic gain would be hurrying that up. In which case, you'd say the real economic challenge for the kids who are just entering the labor market right now is how can they enter the workforce not to use AI but to adapt work in new ways. And if that's the case, let me try this out on you. Maybe something we should be teaching kids is how to code a simple API. How to work with the tools that they're surrounded by. I taught myself at age 40 something and it's fun. I've never felt more powerful. But a million people are going to have to learn how to plug into these APIs in order to figure out all of the different ways in which we can adapt our work.

I've been struck by in conversations with fellow economists, I'm regarded as moderately techy even though I'm the worst computer scientist on Earth. And it's because I've worked, played with, worked with, been in the sandbox, tried to develop, tried and failed lots of things. And so maybe we need a little bit more of that broader literacy of how do we plug into a system, how do we change it, how could we code around it, stuff like that. Again, I'm not sure I'm right. That seems like a short term thing is it's going to have to be a lot of adaption and change within workplaces. And that seems like something that would be important at least for the next decade.

Absolutely. Can you code Nick? You know, like you, I am spending a lot of time with the vibe coding, which is incredible. I mean, basically you're talking to a large language model and then it is doing, it's writing what you want to have happen in Python or whatever the programming language is behind the scenes for you. Never felt more powerful. That's a great way to describe it because it's amazing. It's always been a brick wall for me throughout my life where I just go, I really, I don't quite see the matrix in the numbers there with coding. And all of a sudden, I can see it and make things happen and create an agent that goes out there and does repetitive work day after day.

And I'm only clearly scratching the surface. I mean, so the first time I drove a tractor, I thought it was incredible. We could do huge amounts of physical work or lift, you know, with a bulldozer or something, a lift, huge amounts of dirt. Like the small amount of muscle here was somehow lifting tons that felt strong. What coding felt like to me, and I'm just sharing an emotional experience. I'm not being an economist here. Yeah. Is it felt like that for cognitive work? I can code a loop where the computer will do things a million times. Holy cow. Yeah. Like I've never felt more powerful. And then that computer program only does the loop a million times because it's a program I might not have to press the button.

I can do a million times whenever I want. It's, oh, it's lovely. So if I do nothing else, I just want to share the emotional feeling that this can be really exciting. And if your audience finds that of any value, there's a million ways to dip your toe and you might discover that there are ways you can adapt this technology so that you can live your life in better, richer and more powerful ways. Yeah. And anyone who's curious about it. I find the Zapier agents to be incredible and they are sponsoring some of our episodes. So I do need to acknowledge that. But I got involved with the Zapier agents prior to that sponsorship.

That's actually what led to the conversation. And it is, I wouldn't say incredibly easy to do, but it's by an order of magnitude easier than it ever has been before. And if you're willing to put in some time and learn, you can have that same amazing emotional experience that Justin here just described. Justin, if as we wind this down, if you are listening to this and you want to think more deeply about this stuff, you share some of these same concerns, especially regarding those questions you brought up about competition and ownership. Any thoughts on what an individual can do in order to feel like they have a voice in this conversation?

I admit that I've not thought hard about that question, but I've thought hard about a cousin. So I'm going to just ignore your question and answer the one I haven't answered to as that perfect. I think many of us feel powerless as we're looking ahead to a future where robots might be doing our work. And I think that's a tremendous source of anxiety. And so the personal advice that I give people is not to become an AI expert, but become the best in your office. Because if you're the best in your office, you're the one who's interviewing the vendors. You're the one who's thinking about new work processes.

You're the one who's leading change and that makes you the most valuable guy there. To be the smartest guy in your office on AI is depending on your office and I don't know your office. But in my office, turned out not to be that hard. And actually it was kind of fun. And I think that's the single best advice I can give you for future proofing your career. Because that's the last guy fired. The guy who's still putting the boxes together or licking the envelopes or putting their spread sheets together by hand, they're the first guy fired. So I would just say try to be the innovator in the office.

This might sound intimidating because if I said go out and write code, you'd be like, well, code's hard. What's different about this is this is a technology that uses English. This is a technology that's explicitly modular. That's what an API or application programming interface is. So the on ramp to using this technology is probably the easiest on ramp of any technology that's changed your workplace. And so the big secret is Nick and I can sound moderately coherent while actually not being particularly expert. And it's because the competition is so terrible. And so because your peers are scared, because they're not investing, if you invest a little, be ahead of them, and then you'll play the leadership role.

I don't see how we can end on a better note than that. Alright, man. Actually I'm going to land on one better. I want to challenge your audience. What a friend of mine told me to do and I think it was brilliant advice. Every single day, open up a new tab in your browser. And your job is to think of a new goofy question that you can pose to an LLM and see if it can solve it for you. And what you'll discover is sometimes it'll be terrible. But what you'll discover is things you've never thought of. So of course, if I want to know what the capital of Poland is, I know that I could go and ask an LLM.

But one of my buddies said, I went on a vacation to Tasmania. One of my buddies said, did you have CHACCHE-PT, organized your itinerary? I was like, no, why? CHACCHE-PT? And he's like, think about the problem, right? You're trying to, it's a basic computational problem. What parts of the state should I visit? In what order? What are the really fun places to go? It turns out it's unbelievably good at that. Maybe that was obvious to you already. But his point is, have it right, good nights to already your kids tonight. Just ask it something different every day. And after 365 days, my guess is you'll have 300 successes, which means you will discover 300 ways in which you can use it to do your work more effectively and more efficiently than the guy next to you.

And so that's the challenge. Every day, open the browser and don't close it until you've tried it on one new type of task. I mean, all these stuff. That's a great exercise. And just to add onto that, something that you can do with it. So my father, who is 75 years old, does something very similar? And then you can share the entire chat with somebody else. And so he'll ask a weird question and will text me and say, look at this wild response I got on this topic when I asked it X, Y, and Z. And I find just his weird questions very insightful. And this is a gentleman who's on the back nine of life has no tech experience whatsoever.

Just getting in there and experimenting and trying things is the first step. Yeah. Yeah. Thank you so much, Justin. This was an absolute pleasure. And you've given me a lot to think about and all of our listeners as well. And thank you so much for your time and insights today. Oh, and if anyone wants to find you online, where's the best place to find you? OK, so what is your only fans account? Yeah, that's right. I'm at Justin Wolf is on Twitter and threads and blue sky. And here's the thing. I'm trying to get up the courage to go sub stacking or to start a YouTube channel. So if you want to follow me at one of those places, if I find enough followers, I may well get the courage.

So that's where I'm hoping to meet people, but I'm not there yet. I really hope you do it because you do have a real talent for simplifying incredibly complex things. And I would say in a way that at least to this point, the LOMs can't do in the same way you do in the same human way. So I would I would love to see that happen. Thanks, Nick. Appreciate it. All right. Okay, everybody. Until next time. Questions don't accept the status quo and be curious.