Loading summary
Transcript37 lines
- [00:00]
A
Foreign.
- [00:06]
B
This episode is brought to you by Indeed. If I had to hire someone for this show, I wouldn't want to pick up just anyone off the street. They need to have the right skill set and background. If I wanted to hire an editor, I'd probably want someone who knew how to use editing software. If I needed a writer, it'd be nice to have someone with experience in journalism who closely follows the political and tech world. When you're running a business, you shouldn't settle for anyone, but the best Indeed can help you find the best. With Indeed sponsored Jobs, you can stand out from the crowd, reach qualified candidates faster, and increase the amount of people who see your job listing. By the end of this ad, companies like yours will have made 27 hires, according to Indeed data. And that's just in one minute. Think of how many hires are made per day. Get the results you want with Indeed Sponsored Jobs. Listeners of this show will get a $75 sponsored job credit to help get your job the premium status it deserves@inn Indeed.com plane that's Indeed.com plane right now and support the show by saying you heard about Indeed on this podcast. Indeed.com/terms and conditions apply. Hiring do it the right way with Indeed. This episode is brought to you by Ninja 1. Ninja 1 understands that it teams today are stretched thin, trying to manage too many disconnected tools and rising security demands keep raising the stakes. Ninja 1 unifies everything in a single intelligent platform, from endpoint management and autonomous patching to backup and remote access. Fewer tools, lower costs, higher efficiency. Trusted by more than 35,000 customers in over 140 countries unifi it to simplify work with ninja1. Learn more@ninjaone.com I've covered many subjects on this show. Inflation, GLP1s politics, loneliness, aliens. But the biggest divide in my audience, and the biggest divide among the people I read and listen to and trust and is on the subject of artificial intelligence. One way to characterize this divide, and maybe it's a simplistic way to characterize it, is to say that on one side of this debate you've got slobbering sycophants of the technology who say everything is about to change. The lid is off on history. AI is years or months away from automating every single human task, about to transform every nook and cranny of human experience, and on the other side, you've got hardened skeptics who think this is all complete nonsense because the underlying technology is little more than a parlor trick that Silicon Valley billionaires want to shove down your throats. The news and discourse space as I see it is divided between these outrageous extremes. This technology is a joke versus this technology is 12 months away from changing everything. I think this is a false choice, and today I want to suggest a better, more sophisticated way for this conversation to proceed. First, there are some cultural differences I think there are at play here. When I'm in San Francisco visiting with people who are building this technology, I spend time with folks who've chosen to devote their work life to this invention, even when they themselves are not entirely sure what they're summoning into existence. When I'm in D.C. or New York City visiting with people on the receiving end of the Bay Area's exports, I'm spending time with folks who regard the last 20 years of Silicon Valley products as being somewhat regretful additions to the human experience, such as social media, which has left many of us with anxiety disorders and left Silicon Valley with hundred million dollar mansions. So there is a cultural difference here, where some view Silicon Valley as rich parasites who should be profoundly distrusted, while the Bay Area tends to regard itself as the font of technological progress. But beneath that cultural distinction, there's a deeper substantive divide over AI. And it's not really one disagreement. It's more like four distinct divides. So here are what I'll call the four great divides of artificial intelligence. Number one, is AI useful? I have brilliant friends who use this technology every day of their life and say it's transformed their work, and I think they're right. And I have brilliant friends who have tried repeatedly and insist that it does nothing for them. And I trust that they're right. AI is not like a light bulb which provides the same wattage to all users, depending on your job and the AI model you're using and the quality of your prompts, and a thousand other factors. AI is more like a light bulb that offers some people a million watts when they turn it on and offers others nothing more than total darkness. Number two, can AI think? Are these tools engaging in something like human thought, which combines memory, sense prediction, taste? Or are they blunt instruments for synthesizing average work across several domains? Average data analysis, average student essays, average art I want to pause here to point out something I think is really important. I've seen many people suggest that AI can't think and therefore it isn't useful. But those are separate questions. AI can help a scientist draft a paper or a bibliography, even if it doesn't meet our philosophical or neurological definition of thinking. It can be useful to many people, without being technically thoughtful, the third great divide that I've observed is one that I've most directly participated in. Separate from the question of is AI useful? And can AI think? And is the question is AI a bubble? This is principally a question about the speed and timing of the technology's adoption and its revenue curve. The hyperscalers in Frontier Labs are currently spending hundreds of billions of dollars a year training and running artificial intelligence models. If they don't see ferocious revenue growth from AI in the next few years or the next few quarters, a lot of companies, especially those that take on debt, are going to find their current position untenable. They'll either face a markdown on valuation or a layoff or something even more catastrophic. Once again, you can believe that AI tools like, say, Claude Code can do economically meaningful work and therefore believe that AI isn't a bubble. Or you can believe that AI does significant work and it's still a bubble, because there's no way these companies will make back the money on time. In fact, that was more or less my position for most of last year. The fourth and final great divide here might be the widest and it's hard to capture succinctly. So I hope you'll excuse me, going a bit broad here. Separate from the questions of is AI useful or is it thinking or is it over levered is a question that's something like is AI good or bad? At one end of the spectrum you've got the venture capitalist Marc Andreessen proclaiming that AI will quote save the world. On the other end of the spectrum, you have the rationalist writer Eliezer Yudkowski arguing that if anybody builds superintelligent AI, everyone dies. There's a lot of real estate between those positions. Save the world, everybody dies. Maybe you think AI won't usher in the end of the human species, but you worry that it will make the most beautiful things in life, movies, art, human relationships, more slop filled and shitty. So this is the landscape of the AI debate as I see it. What seems on the surface to be one debate between pro and anti camps is really several different debates that are becoming conflated and mushed together. Is AI useful? Can it think? Is it an economic bubble? Is it good for us or bad for us? Those are separate questions, and our ability to bring wisdom to this topic depends on our ability to see that separateness. In the spirit of trying to be specific about AI, today's guest is here to help us answer a very specific question. What will AI do? To jobs. In his Atlantic cover story this month, author Josh Tieringel points out that the people who have built AI and have spent much of the past few years predicting the most dire possible effects it will have on the economy. In May 2025, Dario Amade, the CEO of the AI company Anthropic, said that AI could drive unemployment up to 10 or 20% in the next five years, wiping out half of all entry level white collar jobs. The CEO of Ford estimated that it would eliminate, quote, literally half of all white collar workers in a decade. Last month, new tools like Claude Code and Codex were unveiled that caused a shiver throughout the tech world as many top coders claimed that they could take over enormous chunks of their own work forever. All of that sounds incredibly spooky, incredibly scary. But when you look up from the scariest and spookiest predictions about the near future to study the labor market of the present, it's actually hard to see any, any effect of AI at all. Unemployment's under 5%, hiring is slow. But it's not particularly obvious that AI is doing anything to the labor market as we speak. So how do we balance these pieces of evidence? The audacious predictions of tech CEOs, the enthusiasm for tools that do seem to automate certain tasks right now, and the current calm of the labor market? What does it all add up to? Whatever it is to quote Josh, America isn't ready. I'm Derek Thompson. This is plain English, Josh Tieringill. Welcome to the show.
- [10:21]
A
Thanks, Derek. It's great to be here.
- [10:23]
B
You have the COVID story this month in the Atlantic entitled America Isn't Ready for what AI Will do to Jobs. And I want to start outside the scope of the article and talk a little bit about the debate about artificial intelligence as I see it. I'm oversimplifying here, but I think there are two groups that are going to come into this episode from very different places. Group one says this technology is going to have a massive impact on the economy, on jobs, on the future of productivity, maybe even on our own sense of who we are and the value that we can provide our families and our companies. Group two, however, I think, insists that AI is basically vaporware. This is a lot of obsession about basically nothing. This is billionaire pumped nonsense that hallucinates and doesn't really help anybody do anything. We're going to spend a lot of time talking to Group one, the folks who believe that this technology might be revolutionary. But before we do that, I would actually love you to address Group 2 directly. How sure are we that they're wrong. Why are you convinced that this really is a revolutionary, transformative technology that deserves our very close attention?
- [11:43]
A
Yeah, I think you put your finger on it, is that there is this divide, right? And there's some people who think AI is going to be a deity, and there are other people who think it's a parlor trick. And I will tell you, I'm not in between. I actually think that its power is incredible. But I also understand why group two is greeting this skeptically. And I think a lot of that has to do with the context into which AI has arrived. Right? So number one, let's take the biggest possible picture, which is we are not short on existential risk in our lives right now. Political risk, climate risk, just nationhood risk, right? Whether security risk. Everybody is feeling risk from something. We came out of a pandemic in which we were all just super hunkered down and that was the logical way to greet the world. It just was. And so right at the end of that, this thing shows up. And I think a lot of people looked at it, they looked at the hairball of motives behind the people who were creating it and trying to disperse it, which include massive investment, personal wealth that may come from it. We're all sort of now victimized by 15 years of sort of like social media bullshit, and our defenses are up. And so they looked at this, they came out of the pandemic, they looked at the Trump administration creating new risk every day and basically said, not for me, I'm out. I don't want anything to do with this. I don't like the people making it. I don't believe in it. I've heard these promises before. In the end, it will just sell me ads. I don't want to do it. I actually more than sympathetic to that response. I think it's actually kind of a logical response. Here's the thing. The tech is amazing. And I say this as a person who is skeptical for a living. It's dazzling. The things that it can do are remarkable. They are compromised in people's perceptions by the way that this has been marketed. And largely that's about the way it's been financialized. Everything. It's so expensive to make that the labs. And when I say the labs, that means places like OpenAI Anthropic, the people who are making cutting edge models, have had to pivot very quickly into the language of money in order to keep chasing new breakthroughs. And so that has colored a lot of. I think in my neighborhood, I live in the East Village, there's a lot of people who are dismissive of it because it is this sort of collusion of technology and money. And. And if you can imagine the tech without the tech companies, you will have a different response. And I wrote a book that's coming out in May that's basically that, what can we do with this? If we can just get away from the bullshit marketing. So I'm a believer. I'm a believer that the tech is amazing and that it is essentially boundless eventually, and even in the short term, the things that it can do are remarkable. I'm also a believer that it's entering a fractured system that makes the likelihood of its misuse pretty enorm.
- [14:51]
B
I agree with a lot of that. I also, and I want to be sympathetic to this second group of skeptics. I think one thing that makes AI interesting and maybe unique in technological history is that unlike other general purpose technologies, the capabilities of AI are exquisitely personal and local. So, for example, you think about something like the train or electricity. Well, if a train takes a bushel of wheat in 1870 from Chicago to New York, everybody can agree on what that train has done. It has transported a commodity from one major city to another. There's an objective understanding of what that technology just accomplished for an electricity, for electricity in the 1880s, 1890s, early 1900s, as it was catching on. And I think we're gonna talk about electricity in a bit. If a light bulb turns on, if a Broadway marquee turns on, everybody can see exactly what electricity has done. It has created light. But you compare that to generative artificial intelligence, which is an interaction between a unique user, a person, and their computer or their phone. Every interaction is unique, every prompt is unique. And so some people are using technology, are using AI, and coming up with answers that are, I think, sincerely not useful to them, or especially if they're using older models, hallucinatory. But a lot of people, especially in the last few weeks, software programmers who are using the most recent editions of Claude Code or Codex and OpenAI, are using these tools and saying, holy effing shit, this could change the way I work forever. And so I guess I wonder, even if you zoom into journalism, for example, I was having a conversation with some folks that do sports commentary the other day, and I was trying to help them understand how they might be able to use this tool. And it was interesting because honestly, they pointed out to me that it's actually not that useful sometimes. If you're an NBA commentator on espn and you want to figure out who are the best point guards when facing a zone defense. They have a very easy way of looking that up with tools that exist. But if I, an economic analyst, want to understand the share of spending on groceries in major cities over the last 100 years, there's not one portal that's really easy for me to use. And so using AI actually gives me access to technology or, excuse me, to information that used to take a long time to access. So I'd love you to maybe just dilate a bit on, on this point too, that the utility of AI is so fractured such that some people, even well meaning, good faith people, can have a really negative experience with this technology, while other people are also saying, in good faith, it could change my work forever.
- [17:37]
A
Yeah, 100%. Right. And we'll start with journalists because we are them and we know them. And a lot of their experience was, oh, I can't trust this. And their initial hallucination that they experience from the AI would scare them off. And I get it. Our job, first and foremost is we have to be right. And so if you can't trust the information, you're going to discount that. Also the writing, your mileage is going to vary. The writing from some of these tools is still a little corny. It's still very cliche. If you write for a place in the Atlantic and you submit an AI written article, I mean, it wouldn't get past the first minute of an editor's read. So there's reasons to distrust it. Then I talked to my friend friends who work in financial services where immediately they're like, yeah, we do sometimes have to check the math, but holy shit. Like, this is the brain that I've always wanted to have at my side, right? In terms of the capability to do arbitrage to create decks that used to take me six hours, it's taking six minutes, and now I can actually do the thinking that I want to do and oh my God, yes, I'm concerned that my boss is going to eliminate my job, but in the meantime, holy shit, this is great. In medicine, it's really exposed this divide between younger doctors who are much more willing to embrace life hacks. The medical profession is ruled by electronic health records. For better or worse, their orientation is around having to fill out forms. Younger doctors have figured out AI scribe tools can help me fill out those forms automatically. They're saving me an hour or two a day. Older doctors don't want to change their workflow. And so what you're going to See, is that based on your profession, based on the way you work, your mileage of success will vary, for sure. And I also think there is something to be said for the phenotype of the person, right? So forget the profession. Break it down to just, are you a person who loves routine, who has specialized, Your technique is a little bit resistant to change because you like the doing things the way you've been doing them. This is going to be very hard for you, because in order to do things differently, I'll just give you a simple example from doctors, right? Scribe Technology is stuff you just turn on and it listens to the patient, doctor interaction. It notes everything. It can fill out the ehr, it can fill out prescriptions. You have to make a subtle change, which is your doctor will be vocalizing their entire exam. And so when I went to somebody who was doing a test case, he basically said, all right, Josh, here's what I'm going to do. I've got my stethoscope. I'm listening now to the left lung, and I'm moving down to the bottom of the lung, and there's some congestion there. I don't like the sound of that. Now, that's different, right? When I go to my actual general practitioner, who I'm pretty sure doesn't know my name, I don't think that there's more than like, 10 words exchanged. He would never be able to use Scribe software. And my guess is that he also probably doesn't fill out the EHR very well. But that's just like one example of how each profession is going to have a different time in its relationship to the tech.
- [20:47]
B
I'm really glad we spent some time on that because I don't think this is just commentary on the debate, like why some people disagree on the utility of the technology. I think this is a point that has incredible resonance for the question, how will AI affect the economy? The fact that you can take two people in the same industry, journalism, medicine, and they can have completely different approaches to that technology. They get different things from the technology. That's going to help, I think, us understand how fast this technology moves, how fast it's adopted, and therefore how fast it changes the economy writ large. So this is how I want to structure the rest of this conversation. I want us to consider essentially three different scenarios for AI and the economy. The first scenario that I want us to talk about is that AI won't cause much job displacement or automation at all. It's basically a normal technology. The second scenario that I want to consider is that the changes will be significant, but they'll be slow. And the third scenario that I want to consider is what some people think of as the more apocalyptic scenario, which is that this time is different, things are gonna move very, very fast, very, very soon. And therefore, we need, I think, a big. Not just sort of cultural conversation about this, but political conversation about this. So that's how we're gonna go. Scenario 1, 2, 3. Let's start with scenario 1, why job loss won't happen. If you look at technology, and I believe this is in your essay, like the ATM or the Excel chart, someone might have been in our position as journalists in the 1950s, 1970s, and this technology was coming online and saying, oh, there's not gonna be any more bank tellers. There's not anybody working with spreadsheets in the future. Absolutely wrong. There are still bank tellers. Everybody works with Excel. And so these are technologies that didn't destroy jobs at all. In some cases, they just amplified jobs. So walk me through this possibility that artificial intelligence is in some ways a normal technology that will essentially sit with knowledge workers in the future without actually replacing them from the labor force.
- [22:51]
A
Yeah, and listen, there's a lot of people, a lot of economists in particular will say, yeah, that's exactly what's going to happen. There'll be a little period of adjustment, and then we're going to create even more jobs, even better jobs with even higher wages. The case for that is that the tech, first of all is expensive, and it's all really localized around speed. It's expensive. It takes a while to move into our various industries and systems. And in that time and space, we have the time to imagine ways in which the tech amplifies employment. Right. And listen, I'll be honest, like, I think this is the least likely scenario based on what I know about the technology. But in this world where things take a little longer to get into the enterprise, where your CTO is like, look, it's great, but it doesn't work with all these other systems. It's going to take several years. We can imagine a slow pivot where we begin to say, okay, so look at the doctor example, right? That just saved an hour or two. An hour or two over the course of a year is an immense amount of time. It reduces attrition, which is a huge factor in employment in the medical field. It actually gives us this massive corpus of data which we can now use to create even more jobs. One of the fastest growing jobs in the 21st century is data analysis and data visualization. Now we have this massive corpus of public health data. Think of all the new jobs that can be created, analyzing that. Those are the kind of rosy scenarios. And I'm not going to say that's impossible. I think it's very possible. And I think some fields that are forward thinking will absolutely do that, will absolutely take the money and invest in new jobs. I think when you look at financial services, there's a real split, right? There are some people who think this is technology that will just duplicate jobs we're currently paying a couple hundred thousand dollars for from kids right out of school. It'll just duplicate that. It'll take care of the report making and the analysis. And there are other financial services organizations that are like, no, great, now we're going to redeploy those brains in all sorts of new ways to find ways of investing and making money that we just have never had the time to do. So that's like the case, right? So again, I don't think it's impossible. I think it's dependent on the technology moving at a very particular pace now. It's also dependent in some ways on the discipline of the people who have employees. Because not only do they frequently have employees, they also have investors. And so when we talk about this, none of this happens in a vacuum. All of it is influenced by the human behavior of investors who want to See results. Different CEOs have different messages for Wall Street. There are some companies, and I think, most notably we would say Amazon and Apple, that have historically said, oh, are you investing in us because you want to see a nice quarterly return? We don't want your money. We play in decades. We're basically the Chinese government. Our CEOs will be here for 25 years. We are always looking two turns ahead. If you think this is your quarterly, no, we're not for you. There are other companies where the CEOs simply don't have that will and investors will attack them and, and they are presented with seeing gains more immediately. And if you are managing month to month, quarter to quarter, the sort of gradual landing that you're talking about becomes much harder. And that has nothing to do with the technology. That's just human behavior.
- [26:23]
B
I do think that one of the strongest reasons why we should think that this technology will move a little bit slower than some of the most audacious predictions from Silicon Valley is that practically every general purpose technology has been adopted quite slowly. I mean, you think about, you know, the canonical example, I think, is the telephone. The Telephone is patented in 1876. In the first few years, there's basically no telephones that are manufactured because like, talking across long distances is not a part of anybody's life. And so initially people are like, I don't even know how I would use this thing. I believe the initial applications were mostly commercial. The adoption of the telephone did not pass 50% of American households until the 1940s, according to our best records. Right. So it took 70 years for half the Americans to essentially pick up the phone. That is the typical story of general purpose technology adoption. I mean, you tell the story of electricity in the piece. And before we contrast these stories with AI, I think it's important to tell these stories because this is how technology typically works almost all the time. You have the invention, but implementation happens at an enormous lag. So tell us a little bit about electricity and how power stations really lagged the ability to actually generate AC DC technology.
- [27:43]
A
Yeah, I mean, look, electricity is the transformative technology of the last two centuries. It basically took four or five decades to be reasonably dispersed into America, which is a very powerful and very capital friendly country. And part of the reason is that it's just when you create something that new, there are a lot of vested interests. So there are lots of factories that literally were built on steam. Right. And when I say built on steam, I mean they constructed a steam engine in the basement and then they built the factory on top of the steam engine. So when you hear as a factory owner, hey, we've got this brand new tech, it's so much easier, it's going to be cheaper. You're like, uh huh. Well anyway, I've got this building and I spent all of my money to create the building. Call me when I can plug in. Right. And so a lot of those factory owners waited and they waited for their materials to become obsolete. They waited for price pressure to make it impossible for them to change. They also waited for something really important, which is the government to pay for the rollout of electricity. Right. Electricity largely was subsidized, particularly in rural areas, by the government saying, we're going to get everybody connected to electricity. So that's just a very simple example of friction. And some of those factories I referenced, they did take 50 years. There were steam driven factories 50 years after the discovery of electricity, because their materials still worked, they were still making money. The expenditure was enormous. Now I'll give you the counter. The counter is that AI is already probably the fastest growing consumer technology in the history of technology. OpenAI has 100 million plus users. Anthropic is not far behind. The other point that people like Anton Kornick and others will get to is that electricity didn't roll itself out. It was a material. And it required all of these human hands to build the infrastructure, to create electrification in factories and in homes and create a grid. We didn't have a grid. A lot of the ads that people see on sporting events that they do not understand are for AI tools that help companies roll out other AI tools. And so we're dealing with smart machines. And that, more than anything, is the hardest thing to wrap our head around. These are smart machines that, given the proper instruction, can create a pathway for other smart AI technology to infiltrate its way into your home, into your factory, into your enterprise. That's going to take us some time to get our heads around.
- [30:24]
B
Yeah. I think the strongest reason why artificial intelligence is either not going to displace jobs at all or going to have a slow impact in the labor force is essentially one word. It's history. Right. Historically, changes, even changes that come from general purpose technologies like electricity happen very slowly. They don't appear in the macroeconomic data for years. I think it was what a Paul Krugman joke that computers exist everywhere but in the government statistics. Right. They you could see computers when you walked into an office, but whenever you looked at the government statistics, it didn't seem like there was a productivity boom until, I guess, maybe in the second half of the 1990s. So the strongest reason, I think, to believe that history will repeat itself is essentially that history has kept repeating itself. It has been very. Unemployment right now is under 5%. So it's very difficult to say that this technology has already had some kind of significant displacing effect. But this gets us, I think, to our final scenario and the scenario that you focused on the most in your piece, which is the case that this time is different and that AI could in fact move very, very quickly. Is, in fact moving very, very quickly. You already alluded to the idea that I believe, According to Pew, 50 to 70% of Americans say they use this technology every week. This is not electricity. Right. People were not using light bulbs three years after Edison patented them in the 1870s. It took 50 years. So what is the strongest argument that you've heard? That this time will be different? That we could really see a significant displacement or rejiggering of the labor force within the next few years?
- [32:04]
A
I think there's two, and I want to separate them because it's really important to understand the difference, one is technology based and one is Wall street based. And that's literally the fire in the cave and the shadow that the fire casts. So let's talk about the tech. As I said, a lot of AI tech, if not quite, there's a term called recursive AI, which means AI that can teach itself. Now, there's a lot of debate, for all sorts of reasons, about whether we have achieved recursive AI, but it's sort of moot. AI is capable of rolling out AI. And what we're seeing with Claude code, which is another sort of really important development that sort of came to market last fall, is that you can actually tell an AI I need to install AI in this function. I want to connect it to this data set inside your company, inside your life, make little life hacks. It's not perfect, it's not as good as, you know, this sort of pristine human code. But software is all about. Everybody will take bad software that works over perfect software that takes forever to ship. That's just been the rule of software. And so what we're seeing is people are able to create AI driven systems inside their companies really, really rapidly. And that's not just gen AI, right? So generative AI is. You can easily understand how generative AI can replace a sort of average customer service function. Right? You train a generative AI on lots and lots of your manuals and tools and protocols. In some cases, it's actually way better than the human friction of customer service. You can see how those jobs might go away this year, but you can also begin to see how data driven companies can use AI to essentially run all sorts of different functions really, really quickly, because the AI is actually helping you roll it out, troubleshoot it, and it's adaptive very quickly with limited human interaction. So I do see that as a very real thing that's happening this year, certainly in software also in lots of other places. You've seen the launch of OpenAI with its health tool, and we can talk about the difference of the business models between OpenAI and Anthropic and others, and how that will also impact jobs. But these are very real, very competitive things happening right now. And so I do think that's one argument. Now, the other argument is that, as I said, a lot of companies, a lot of traditional Fortune 100 companies that have invested in AI have spent billions of dollars to catch up to implement systems. Even if it takes another year or two for those systems to be perfect, the pressure on those CEOs to show results when I Spoke to a bunch of CEOs, many of whom went to ground. In the time I was reporting this, they would not speak on the record. They were petrified of being the face of job loss. What they basically said is, look, I actually like my workforce. I actually think this could take time and we could perfect it. Wall street has no patience for that. They're expecting me to show financial results now. And the way they show financial results fastest is by cutting jobs and replacing those jobs with automation, even if the automation isn't perfect. And so that is the shadow on the cave wall that concerns me even more than the speed of disbursement of AI.
- [35:32]
B
You have some stories in the piece about how exactly this could play out, how we could get companies, and we want to be specific here, I think for the next few questions and answers. How specific companies, say, in consulting, could use this technology in a way that significantly displaces thousands of jobs. So why don't you just take the baton there? Let's hold on one industry, say consulting firms. Can you walk me through a plausible story of how the application of this technology could reduce employment in consulting in the next few years?
- [36:06]
A
Yeah, and let's not be shy about it. Let's go straight to the M word, right? Let's talk about McKinsey, which is the most powerful global consulting firm in the world. They charge an absolute boatload of money for a 26 year old who's never run a company before to be an analyst to help you with your strategy. I mean, that person's time is billed probably at close to a million dollars over the course of an annualized budget. McKinsey has loads of records of its own case studies. McKinsey has access to the very best cutting edge enterprise technology. If they said, what do we need this for? We have a corpus of our own expertise. Let's train a bespoke model on all of that so that when a customer comes to us, let's say for instance, I don't know, you pick the company, but let's use a real name.
- [36:55]
B
A company that comes. I mean, I remember very famously when Condon asked, went to McKinsey and everyone thought they were just going to McKinsey to get permission to fire 10% of their workforce. And then McKinsey was like, hey, surprise, you need to fire 10% of your workforce. So yeah, let's talk about a big media company. We can talk about HBO Condonest.
- [37:11]
A
Yeah, let's say Conde, Conde comes to them and says, we need to be positioned for the future with AI. Okay, reasonable ask. McKinsey will say, traditionally, well, that'll be $5 million, please. Conde says, great, you're covering our ass for all sorts of things. Absolutely. And traditionally, again, an army of 26 year olds with MBAs goes to work. They do all the interviews, they do the analysis, they submit a report, sure enough, it says, let's cut a bunch of jobs. And that, by the way, takes three months. With AI looking back at all of the reports McKinsey has done, inputting all of Conde Nast's current trajectory and future and the competitive set, you could probably make that report and deck in a day in the first draft. You could probably have a team of more expert humans, take a couple of days to go over it and say, oh, well, let's go back in and try this and try this and try this. And that report's done in a week. And I'm not exaggerating. That report could be done in a week, could be done probably better. Now you can say, but they would never do that. Well, McKinsey is not the only consulting firm. There are other consulting firms that would not charge out their people at the same exorbitant rate and might produce a better, faster result. And all of a sudden, McKinsey has massive competitive pressure and they will fall in line because they're not going to lose business over this. So that seems like a pretty reasonable thing that will happen in 2026.
- [38:46]
B
There's something about the game theory here of who moves first to adopt AI in order to reduce the price we charge our counterparties that is really, really interesting to me right now. I mean, I was just reading Bloomberg, Matt Levine's newsletter, and he was recounting some reporting on kpmg, the accounting auditing firm. And KPMG reportedly asked one of its counterparties, its own auditor, to allow KPMG to pay them less money for auditing work. And Matt points out, like, it's not unusual for a company to ask its auditor to, you know, reduce the size of the contract and, you know, pay less for accounting. But it is quite unusual for an accounting firm to announce to another accounting firm that accounting is now less valuable than it used to be. So please accept less money. Because doesn't that encourage everyone who works with KPMG to turn around and say, why should you be charging us the same amount that you've been charging us? And so this idea that AI makes accounting more efficient and therefore reduces the price that you can charge for six months of accounting work can ripple through that field and maybe reduce revenue and maybe reduce employment. Right. This is. I'm speculating here. There's not currently some bloodbath of employment in accounting, but this is the way I think one could imagine a kind of freaky scenario playing out. Whether it's consulting. You told the McKinsey story, accounting, auditing, the KPMG story, or maybe in the legal industry, right? A law firm telling another law firm that it's working for, hey, can you bump down your hourly fee a couple hundred doll? Because we know that you can do it for cheaper. And then that realization rippling through the legal industry. This is the sort of thing that I don't know the degree to which it's happening right now, but it strikes me as exactly the sort of thing that the fast timeline people should be looking at very seriously.
- [40:48]
A
Yeah, Derek, it is happening right now, specifically in software. And you're seeing that because some of this is intuition. But Wall street has hammered software firms over the last week or two. And I've seen this up close with friends who are software developers. They can do the work that used to cost hundreds of thousands of dollars. They would bill out their individual's time, they would bill out the testing. All of the things that it takes to make bespoke software for somebody's company you really can do in a day with Claude code. Again, it's not perfect, it's not pristine, but my God, it's good enough. And the software actually ships. And you can do it essentially with Vive coding. And so I want to explain what Vive coding is, but vibe coding is basically, basically using English to cajole a model to make an app or software that really, really works for you. Okay, There's a lot of bad vibe coding. There's a lot of sloppy vibe coding, but in the hands of a really good practitioner, you can actually make something that's ready to go. And I see it with my friends in the industry. They're like, this is amazing. It also is going to drive my costs down to nothing because essentially, if I'm using cloud code in a $200 subscription per month, and I need two or three people to operate that model to create software, I don't need anybody beneath it. And I know that I'm in a competitive business, and when somebody comes to me and says, in the future, I just need you to build something fast, they'll be able to go to a dozen places that will build them something fast and cheap, and eventually they may even just take that in house. So you are talking about these incredible efficiencies. On the one hand, and for those of us who use software, it's amazing, right? And on the other, what happens to all the people who used to get paid doing that? Because this is not happening over the course of decades. This is happening in 2026. By the end of 2026, the software field is going to look very different. Very different. And software is a huge business in the United States. What does it mean to live a rich life? It means brave first leaps, tearful goodbyes, and everything in between. With over 100 years experience navigating the ups and downs of the market and.
- [43:05]
B
Of life, your Edward Jones Financial advisor will be there to help you move ahead with confidence.
- [43:11]
A
Because with all you've done to find.
- [43:13]
B
Your rich, we'll do all we can to help you keep enjoying it. Edward Jones Member, SIPC this episode is brought to you by Lincoln. Whether it's bonding digitally or exploring the world Together, the 2026 Lincoln Nautilus Hybrid is built for connection with lots of smart tech that helps bring worlds together both on and off the screen. So help turn everydrive into an opportunity for discovery with the 2026 Lincoln Nautilus Hybrid. Learn more at lincoln.com available connectivity, features and functionality vary by model package pricing, trials and term lengths vary by model. Video streaming and games are only available while parked.
- [43:55]
A
This episode is brought to you by Firehouse Subs. Who just dropped a game changing sandwich. The French dip. Literally one of my favorite sandwiches slash subscribers. Roast beef, caramelized onions, melty cheese, little freshly toasted garlic butter roll and the warm savory au jus. I've been eating these forever since I lived on the east coast in la. I think to me this versus the cheesesteak, the French dip, no contest, way better. And I think it's really because of the au jus. I don't know anybody who doesn't like au jus. An elite game day sub. Fun to order by the way, if you want it delivered. Because they usually put the au jus in the special little container you can pour it on. Knock yourself out. The French dip here for a limited time. I wish it was longer. Only at Firehouse Subs. Limited time at participating Firehouse Subs restaurants while supplies last.
- [44:54]
B
Can we think about economists here a little bit? I mean you spoke to a lot of economists for this piece and I wonder whether you got a satisfying answer to a question like this. If a unit of accounting work suddenly becomes 30% cheaper and a unit of legal work somehow becomes 30% cheaper and a unit of consulting work suddenly becomes 30% cheaper, what happens next? I mean, one sort of side question I have here is what happens to those firms? But also, like, where does the extra money go? Like, what becomes more valuable in the economy as some industries become less valuable in the economy? Where does value flow to? Did you get good answers to those questions of like, what happens Next? If the KPMG McKinsey scenarios that we're talking about right now start to billow out over the whole macroeconomy?
- [45:53]
A
I got a single answer from all sides, which was, you just have to tell me how fast this is happening. Over the course of a couple of decades, the economy changes dramatically. The example I used in the piece is toll takers, elevator operators. The slow sunset of those jobs took decades. And so you would lose 3%. 3% of those occupations would go away every year. You didn't notice. But over the course of a decade, it's like, oh, wow, we have fewer of them until eventually they don't exist. But because it took so much time, there was attrition, nobody went into those fields. They automatically started to fill out other jobs. And so you don't notice the economy adjusts. If all of a sudden that happened in a year, if we lost all the toll takers in one year, believe me, you would notice. You would hear about it. And so the way the economists have sort of lined up is I'm going to, I'm going to be really just blunt about it. It's about their belief in the tech. There are older economists who are much less likely to have spent time onboarding their own work through the technology, who believe this is yet another thing, like Excel or social media, that, yeah, yeah, yeah, everybody's talking about it. They're all talking their own book. They're all trying to gin up investment. I'll believe it when it's implemented everywhere. There's a cohort of largely younger economists who use AI every day in their work, who are aware of its power and who say, you guys cannot compare this to electrification or the engine or any previous general use technology. It's going to happen now. And if it's going to happen now, this is where to get to the answer to your question. They don't know what happens because largely all of the wealth that is created by these industries, all of the pricing power will accrue to the makers of the AI. And because AI is made by just a handful of companies, many of whom entered the space already very powerful, they're incredibly concerned that you get this massive concentration very rapidly in the hands of very few people who control the AI.
- [48:09]
B
You'd think that if any story, any tech story, was bound to attract the attention of Washington politicians, it would be a tech story whose bottom line was, and this thing might displace millions of jobs. So you tell me, do you think Washington is paying attention to this story?
- [48:30]
A
No. I mean, no. So I spoke to a bunch of people in D.C. the one person in the middle who actually is pretty well read about it and pretty upset about it is Gary Peters, who's the retiring senior senator from Michigan. He's from Michigan. So obviously issues of industrial obsolescence are very important to him. His number one priority is reauthorizing the largest job training act in America, which has retrained about 5 million industrial workers. Larger, largely. That bill was passed because of global trade. That bill expired several years ago. Washington has not even bothered to reauthorize it, which gives you some sense of the absolute ignorance of the Senate. This is an easy win. And Gary Peters has just been banging his head on the desk saying, can we at least do this? And the answer is no. Now, he's not the only one. I think some of your listeners may be surprised, but Marjorie Taylor Greene was an enormous advocate of taking action about AI fueled impact on the labor market. Josh Hawley's done a little bit of it, but the center, such as it is, has completely ignored the issue. Where you're seeing interest and action, and I believe it's very sincere, is on the far left and on the far right.
- [49:50]
B
So let's talk about that a little bit, because you, in this article talked to both Bernie Sanders and Steve Bannon, making this the rare article that quotes at length boast both Bernie Sanders and Steve Bannon on the same issue, and especially in that rare case, quotes them in ways that they seem to agree that AI presents an enormous threat to the future of America. Tell me a little bit about these conversations with Bernie and Bannon, because I was really surprised by the way they engage with this issue.
- [50:17]
A
Yeah. So listen, Sanders wrote what I think is a really smart first step at what government should do in the face of AI coming for jobs. It's basically. I compared it to Martin Luther's 95 Theses. It is an angry document. It collects all of the predictions from Fortune 100 CEOs. It's very knowledgeable. And what it proposes is basically immediate action to protect jobs, to really do a massive effort at retraining, to shorten the work week, to create job shares in The Face of what's Coming. And I would say that it was greeted with absolute silence and probably wasn't read by the rest of the body. Steve Bannon read it. And so obviously Bernie Sanders has a long history of worker protection. This is an age old argument for him. It's just a new instance. Bannon. Bannon used to be a documentary filmmaker. He bought the rights to the Alan Kurzweil book and wanted to make it into a movie, the book about the Singularity, because he thought it was a horror movie. And so he's been on this issue for about 20 years for his War Room podcast. He hired an AI correspondent three years ago. So he's not new to this. And also the politics of it are very poor for him inside the Republican Party. And so when I went to his house to talk to him about it, it was like he'd been waiting for me for years. He was so excited to talk about the fact that DC is just wrong on this, his proposal. He thinks Bernie doesn't go far enough. He suggests that the government ought to have a 50% stake in the labs because we have provided as a country, we've provided the infrastructure, the energy, the technology, the protection for these labs to thrive. This technology has the potential to put tons and tons of people out of work. So we ought to tax them in a way that is redistributed to workers. He very explicitly told me, the right is going to hate this. They're just going to hate it. But I don't care. I believe this. We went a little deeper, obviously. And I said, well, look, one of the people who has the most power to do something about this is a guy who, not at the time, a couple weeks before Bannon said, well, he should run for a third term. He should have a third term. And Bannon's explanation is basically, well, he's being hoodwinked. He's getting bad information from the likes of David Sachs, who is the federal government's AI czar, also the crypto czar, so busy guy. And he's also invested in more than 300 AI companies, and he's blaming them for the President not having his eye on this. But you can absolutely see a coalition of the far left and the far right building around fighting off AI and protecting workers.
- [53:05]
B
This is one of the dynamics of this conversation that I think is really most interesting and is going to be most important in the next six months, two years, for the 2026 election, the 2028 election. It's how AI is going to split both the Democratic Party and the Republican Party. I see AI as being a complexifier in the Democratic Party because on the one hand, you have moderate Democrats who are sometimes considered technocratic, who are interested in this technology, curious about what it will do, and therefore might have a little bit more of a hands off. Let's see, see exactly how it's used the next few years before acting too fast to shut it down. And then you have the, you know, we can call it more populist left, which is incredibly skeptical of Silicon Valley, incredibly skeptical of the billionaires that are being minted by this technology, and rightly or wrongly, see it as an enormous threat to the American workforce that they see as the future of their electorate. So there you have the Democratic Party. But it's the fissures on the Republican Party that I think are probably subtler to people who are outside of, like the nitty gritty here. You just mentioned that there's Steve Bannon and the populist right wing that sees AI as being this kind of demonic force that's gonna unleash an apocalypse in the world when we have the singularity. And so they're anti AI, but then you look at what's happening in the White House and you look at who they've put in charge of AI policy. David Sachs. This is a guy who I feel very comfortable saying is basically a neoliberal. Like one of the points that I've been shouting about for the last few months is that the IR irony of Donald Trump's trade policy is that on the one hand, we have this high tariff protectionist trade policy toward just about everything that we can touch. And for AI, we have tariff exemptions in the tens of billions of dollars. We have a free trade policy, essentially with AI. We're willing to sell some of our best chips to even our adversary China. And so AI policy being run out of the White House is essentially neoliberal and free trade and welcoming of sort of, you know, globalization in a way that runs counter to both Trump's trade message and Steve Bannon's more anti AI populist message. I mean, what makes this idea or makes this technology so innocent to me, just for the next few months in politics, is, again, it's so unclear to me how the chips fall and where the parties fall on exactly how they treat this technology. I mean, that was a bit of a rant, and I'm sorry for just getting it off my back, but I wonder how you just go a little bit deeper, like how you feel this is going to play out because there's no way, given the level of capital expenditure and given the kind of predictions that are being thrown around here, that AI isn't going to be a major issue in 2026 and maybe the major issue in 2028, someone's gonna figure out how to talk about this. I just don't know yet exactly what they're going to say.
- [55:53]
A
Yeah, look, I think you're right. And at the time that I started the reporting, a bunch of people said to me, this is going to be the major political issue of 2028. You watch. And then by the end of the reporting they were like, no, it'll be the major issue of 2026. So it's coming quick. A lot of what they believe. And some of the people that I'm going to reference did not want to speak to me on the record, but were happy to speak to me off the record. These are cabinet secretaries and ex cabinet secretaries. I think it's safe. I could protect them with that. They basically said, look, I think that the White House is making a bet that Donald Trump scares enough Fortune 100 CEOs that they're not going to do the massive job cuts before the midterms. So that's interesting. And then again, as the reporting went on, what I heard from Fortune 100 CEOs is, oh no, we'll be out of jobs if we don't show some sort of results well before then. And so I think that the question again is like, when will we start to see the impact on the labor market? Because if we do, even the most out of touch politician understands that when your constituents don't have money, they are coming for you. They will blame you, they will demand action from you. And so a lot of this really is going to come down to when will we see it in the reports? So I referenced the Bureau of Labor Statistics in the piece. Right now, there's nothing to see. They don't see major impact from AI. If they do, it's very marginal, it's very much quiet. Now, some of that is we have not funded the BLS properly for two decades and so there's not enough survey data to really go at the question of employers and employees. How are you using AI and is it supplementing or augmenting or displacing you? So we do need better data. But by and large, the very well informed people I spoke to said you will see those impacts in 2026 and you will definitely feel. And what I mean by that is, even if you don't see it in the economic data, we're now getting so much cognitive pressure around AI that people are going to attribute job loss, no matter what, to AI. And so I think what you're going to see is some breakage in the parties. I think just because politicians don't act don't mean they're not smart. I think you will see the far left and the far right begin to coalesce around worker protection fury. Frankly, what concerns me. Let's talk about the benefits of populism, just for a minute. We have the 40 hour work week because of populism. We have weekends because of populism. And these intense doses of populism when it comes to stretching our system too far, are really important. That's how an organism like the United States survives. At the same time, when you talk to Steve Bannon, he is gifted at that shaggy zone between productive populism and revolution. And so when people reach for pitchforks, you don't know what they're going to do with them. We have built this incredible civilization that is oftentimes dynamic and flexible and adjusts to change. Every time you get that, you are entering an unknown zone. And so that gives me great pause and concern because I do think people are already furious about what they're seeing in their country. And I referenced this in the story. But one way that the world is different now is that we are kind of a society defined by two objects, phones and guns. With your phone, you can see exactly how well people are living who are not you. You are surrounded all day by people who have it better than you in imagery and in videos and with guns. You could do something about it. And I don't mean to sound dark, but at the same time, like, that is the reality of the country we live in. And so, as I stare down the path to 26, I'm very concerned about both of those things.
- [59:54]
B
The last point that I want to make that I'd love you to respond to, and this pivots off of something that you just said. If you look at the jobs data right now, the effect of AI is not obvious. Unemployment is low. The economy is adding jobs. Now, it's true that young college graduates definitely have an elevated unemployment rate. Maybe that's AI, maybe it's something entirely different. But this is a theme that I've repeated on a few shows. Narratives can take off in the absence of underlying fact. You could, in a weird way, have a reaction to the fear of artificial intelligence before we have definitive proof that Artificial intelligence is having the kind of effect on the economy that would justify that fear. And that's one thing that I think makes the near future so uncertain, is that you could have. There's a world in which the doomers, who are incredibly optimistic about the future of this technology could sort of find common ground with some of the skeptics of this technology to essentially form a kind of anti AI, AI pessimistic sort of cadre. And they could do things, whether in politics or in something else that could have a significant effect in the next few months.
- [61:22]
A
I think it's likely. I think that your point about narratives is really well taken. The AI labs and the larger software companies have spent hundreds of billions of dollars on advertising and marketing to just saturate the market with talk about AI that's not properly guided. And so they are freaking people the fuck out. They're delivering these ads that people often don't understand for products that they're reluctant to use or if they're using, are a little bit freaked out by. And so, yeah, I think the real risk here is that the narrative will drive fear, paranoia and responses. I think that there's genuine risk to the employment market as well. And that may or may not show up over the next couple months. But I think we're already living in that narrative and it's just a matter of time before people start to exploit it nakedly for their gain. Whether that's political gain, commercial gain, I think that's already happened.
- [62:22]
B
I think I have one last question for you, Josh, and it's a question about control. I feel like this theme that we've been circling is this idea that workers, politicians have this fear that AI is something that they can't control. It's going to displace jobs without workers having a say in their jobs being displaced. But if I hold on that theme of control, one of the really spooky and strange aspects of artificial intelligence is that even the people who are building it will tell you on the record or off the record that they're not sure that they can control it. And that is a truly unique feature of this technological moment. I mean, to go back to the old examples, I keep harkening you back on who was talking about trains as being something they couldn't control. Trains were on tracks, they were on roads. Of course you could control them. They had brakes. Electricity is something you can't control. Of course you can control it. It has an on off switch, it has filaments. This, in a way, seems to me to be a unique technology in that the people who you would think have the most control over its evolution are telling themselves and telling us that they're not sure that they can control it. And I wonder how that features into our ability to predict or forecast its effect on the economy and the labor force.
- [63:56]
A
It's a, it's a really important point. The makers at the labs will tell you we are only just beginning to understand what this technology can do. We in the labs and so regulating the output is going to be really hard because we don't know what we're making day to day. We're beginning to know more, but we don't know what we're making. Anthropic put out a report yesterday about Claude, which is widely viewed as the most humanistic, most value based AI. That said, CLAUDE is capable of contributing to the creation of heinous crimes quotes heinous crimes. So that gives you some pause. Yeah, I'd say so, yeah. I mean the issue here is that someone actually has to take the lead. And I think what I've learned during this story is that you have these institutional entities, be it corporations, economists, actors in government and it's basically the pot fly nobody thinks they need to call for. And so it's just going to fall on the ground ultimately. One of the things that I think you're referencing is like a lot of the social media companies and a lot of the tech companies over the last 15 years have given off nation state vibes, right? They've looked at government and been like, oh, you're feckless, you don't know how to do what we do. We're transnational. We're basically the East India company of the 21st century. We don't need you. They definitely need government. I think when you scratch them they will tell you. It sure be nice to have some leadership on what they expect us to do because this is a problem of the common good. We're not going to stop if the competitor is not going to stop. We're not going to go through checkpoints as far as safety if our competitors are not going to do that. And this is regulation. And you've had these very companies bashing the need for regulation now for a decade or two at minimum. You've had half of the United States believing that regulation is terrible and stopping the government and stopping the economy from growing at the speed it needs to grow at. No matter that we're growing pretty well. But this is a regulatory moment and it's going to be real hard because nobody actually knows the answers. But I think that's what you're getting at is that when you don't actually know what your output is, you need someone to step in and say, well, we're going to try something and you need to do it quickly. I think that if there's one thing I can reinforce, it's that speed is going to tell the tale here. If you think this is going to take 10, 15 years, you don't need to act that urgently. The people I speak with, who I really believe say it's not going to take 10 or 15 years. Go lower, go five at most. And so you do need action.
- [66:40]
B
Yeah, it's. I'm sometimes a little bit just almost exasperated in terms of how I think I should respond to it. Because fundamentally what you have here is a trillion dollar cluster that doesn't know exactly what it's building, sending out this technology to individual users who don't know exactly how to use it. And here we are trying to forecast its effect on the economy and employment in the next five years. Who knows? This is just such an uncertain place to place our bets and yet so unbelievably important. Where I absolutely agree with you is that you said earlier, earlier we thought this was going to be the issue of 2028. I find it very hard in the first few weeks of 2026 to think that by the time we reach November, AI is not one of the hottest issues in national politics as well, even if a lot of people in Washington aren't paying close attention to it yet. I just think this is just one of the most massive stories of our time. And what makes it spooky is that no one, not the people using the tech, not the people criticizing the tech, not the people building the tech, know exactly how to control it. It's unique indeed. Josh Tierengel, thank you very much, Derek.
- [67:56]
A
Thanks so much for having me, Sam.