We are back and we are joined now by Brian Merchant, technology reporter in residence at the AI Now Institute, publisher of the newsletter Blood in the Machine. It's also the title of his 2023 book. Previously the technology columnist for the LA Times and senior editor at uh Motherboard back then.
Uh Brian, thanks so much for coming on the show today. Yeah, thanks so much for having me. Of course.
So you wrote a piece for your Substack, as I mentioned. It's called Blood in the Machine. It's called the AI jobs crisis is here now.
So why were you compelled to write and indicate that the AI jobs uh crisis is already here that people have been warning about for a while and what is its impact looking like in the present? Yeah. Well, I I I kind of wanted to dive into that question, right?
like I mean it's it's always put in the future tense like what will AI mean for jobs? What is going to happen? We how it's it's sort of positioned as this unknowable thing.
Um but over the last 2 years uh I have been sort of reporting on and talking to people who are dealing with sort of this frontline impact. And even though it's not necessarily sort of being tracked on like a macroeconomic level, we need more good data on this. I just I've had enough sort of anecdata.
I've done enough investigative work. I've talked to so many workers who are impacted by this stuff that I decided I just needed to start putting this all together and to formulate a theory of of what is happening now. Um and sort of the most recent catalyst was the fact that um I'm sure uh folks maybe saw in in the news that Dolingo had announced that it was going to be again like in the future it was going to be an AI first uh you know company and that they're going to start phasing out contractors and and writers and translators.
Um and I soon heard from actual writers and say there's nothing future tense about this. They've already let us all go. all of the writers and translators, people who sort of have made Dolingo in this quirky an language learning app into a success and that's what people love about it, we're gone.
Like they fired us uh 6 months ago or a year ago and now and and now they're replacing what we used to do with AI. And I've heard stories like this um in the video game industry, in illustration, in the animation industry. Um, and so it's it's it's enough to say, look, the AI companies are providing other corporations with this logic, this idea that they can automate or replace their workers.
Whether or not the technology is ready or it actually can replace the workers, sometimes it doesn't matter. They can just take that logic and then say, well, this is what we wanted to do anyways. We wanted to get rid of you.
Or maybe this is a useful way for us to say all you pesky workers who may be asking for raises or better conditions. Well, we can replace you if we want to yet. So, don't get too uh, you know, aggressive or or or or you know, or or or demanding for for for better conditions.
And I think the best example of the way that this is manifesting is just looking at what's happened with Doge, right? How are they justifying, you know, a lot of this work when they go into um a new agency uh like the General Services Administration um or now even, you know, just about everywhere, but that started with the GSA and they said, "We're going to do an AI first strategy right now. " And that is not necessarily true.
So for this piece that I also spoke to an 18F worker who was one of the the technological services department workers um who uh was let go when the Elon Musk and Doge completely wiped out that entire agency and said you know this is part of our uh AI first strategy and he saw the AI tools that they're actually saying planning on on using in the agency and he's like and it was laughable. He said there's no way that this thing is ready for the but the technology is not there. The techn but but you it like so the rep Republicans right now are debating uh their budget bill and currently they're proposing this part of the budget bill which is just thrown in there clearly by these special interests um that says so much for states rights.
The Republicans want to use the federal government to block states in the United States, even if they have state law, from regulating AI for 10 years. And what's so crazy about what you're describing is do Dolingo is firing these people with this like AI that's not the technology isn't there but what is there is built on from culling from the work of human beings prior to this 100%. I mean, and that's what and we're so the, you know, people often say, well, which is it?
Is, you know, AI, uh, you know, so scary that it's that it's going to replace us all and we need to prepare for that or does it suck and is something that we shouldn't worry about? But the answer is it can be both and it is both uh because it is being used as the tool to do these things, as an excuse to do these things and it can do immense damage in the short term. You you look at a company like CLA, which is this fintech company, this Swedish company that was like one of the first major uh companies to like go all in on AI and like really be a cheerleader.
So much so that OpenAI is using like quotes and testimonials on its own web page. And I think CLA just did a deal with Door Dash so that you can do installment payments for your lunch. uh which is going to just completely we just hit a record credit card debt a few months ago.
Just to give people a sense of how people are getting by just putting it on their cards right now. Putting it on their cards uh you know breaking down sort of you know all of these all of these structures and these and these jobs right like it Door Dash is um the model that kind of predates AI. So, it's useful to bring that into the conversation, I think, because if you want to look at what OpenAI and the AI companies are doing, um, it's like what Uber did on steroids, right?
Arguing that, you know, that it needs special treatment. Uh, that you should it should get rid of regulations that govern, say, taxi work in cities. Uh, but now these AI companies are going right to the source.
They're going to they're saying, you know what, we've mustered enough power. we've got enough buddies in Congress. And I you know, the thing about this moment is it's not just special interest.
It's actually the oligarchic capture of of Congress and and and the and and the federal sort of level. The the Trump administration itself is now run by people who have a vested interest in seeing AI companies take root. So we have this concentration of power and they're saying, "Why don't we just go to Congress and see if we can ban all law making at the state level?
" This is the right the party that just last summer was like championing states rights and now they're saying we want we're going to ban your ability to legislate on this hugely impactful technology at all. And they're sliding it into a budget reconciliation bill. And I I've never seen anything like this.
Right. Again, it's like Uber on steroids. Like this is this technology is too important.
It's so gamechanging. We can't have anything getting in the way. And just the fact that this might even pass is I think is just yet another one of these reflections of this moment and how much uh how much power sort of, you know, the tech industry wields in this administration and this Republican Congress.
And I wonder if that crypto money, the record amount of crypto money that came into the 2024 election is also just AI money. I mean, these are the same players. These are the same people.
They they're all they're all getting rich off of AI and crypto. Yeah. Mark Andre who is one of the biggest sort of uh you know crypto investors through Andre Horowitz the venture capital firm um is also huge into AI.
Every single VC is big into AI now and his fingerprints are all over this uh this this this new law. I there's a strong argument to be made that Andrea now is just as influential as Musk is in in the Trump administration and and in the GOP. And he has this little tech manifesto that came out last year that sort of is the framework for arguing that you know regulations should be done away with except for some very basic things.
And you can see that in when like JD Vance goes to Paris and he makes this blustery speech about how AI is American AI is going to dominate. Um, and then now we see it in this bill that's used that's that's justifying uh, you know, trying to wipe away even the capacity for states to write laws that pertain to AI. Well, uh, and Mark Andre, it's funny you mentioned that.
I guess it was last week. He was saying how there's one job that's too important for AI to take. Brian, you must have seen this quote.
Why don't you tell the audience what that was? Oh, you three guesses. Yeah.
Yeah. And the first student you know I mean what what is the one job? It's the venture capitalist.
Venture capitalists. I mean look you know we have nuclear scientists. We have people trying to cure disease and they can all be replaced by artificial intelligence.
Can I just say if venture capitalists can't be replaced by it then neither can roulette player. Yeah. Yeah.
Well, I think that they've actually done studies on this that like that just sort of did actually pit sort of I don't know if it was a VC but an investment banker um you know against sort of just like an automated system and like the automated systems like do they they win they they they make more money but of course that's immaterial right who gets automated and who doesn't is all a question of power and very little else right let's pull up this graph you cite uh in your piece and It's pulling from uh Derek Thompson's writing in the Atlantic. Um here is a here's this graph. I guess it's from this one must be from the Atlantic here.
Um where you can see uh the unemployment for recent college graduates. Just I mean take a look at that. Obviously, the COVID pandemic is a huge driver of that going below zero at this point, but we're in a worse situation in 2025.
Can you describe for the audio folks what we're seeing here, too? Yeah, actually, Brian, maybe you can as somebody who's maybe more number literate than uh than me, math literate, graph literate. Yeah.
So this graph points out a really disconcerting trend for for for one major reason and that's historically recent college grads like people who just graduate from college with a new degree eager to enter the workforce are typically among the most employable uh demographic period. Right? They're they're young, they're ambitious, they're educated, they're willing to work for relatively cheap.
So they are sort of an an ideal worker in a lot of in a lot of cases. Firms want to invest in them. But this number shows that that number is declining and it's declining relative to the employment of in the economy at large which is really interesting and disconcerting.
Right? you'd expect, okay, maybe the job market itself is tanking and then and and then you would find uh fewer uh college grads able to get work. But while the economy has been relatively healthy by these metrics anyways, you know, by by these metrics, there's a strong case to be made that it's not as healthy as but anyways, it's not in in other words, this number is going down not in tandem with the regular numbers.
And so this could mean a number of things, but one plausible theory is that firms aren't just hiring as much anymore. They're giving those jobs to AI, to automated systems. Because a lot of times the work that a college student or a college grad would do is sort of the entry- level work.
They may be ambitious, they may be smart, but when they come into a firm, they, you know, they do sort of, you know, the the the the easier, less complicated tasks. And that's the kind of stuff that you would expect to be done by AI or that management thinks can be done by AI. And this is just worrying on so many levels if if if new grads aren't entering the workforce and learning these skills and sort of growing uh with with you know in their skill sets and in their knowledge bases uh and AI and instead it's just being handed over to AI or the firms think they can hand it over to AI then we risk sort of a pandemic like condition where you have a group of people who just you know uh can't find employment and uh and and all of the attend dependent social issues that go along with that.
So yeah, this is and and I'll also note that some people have pointed out that this extends before uh the sort of that the chat GBT AI boom, but firms have been talking about AI and automation for at least the last 10 years. And the other trend that's been really big um in the same time period has been uberization and gig work. And a lot of college students have had to get contract work and stuff that's also less likely to show up um in in economic data.
I got an internship out of college and you know a lot of just regular jobs didn't bite if I couldn't live at home uh and not pay rent while I was working. I don't know what I would have done. And that was back when I graduated in 2016.
Um, but that is like I I I I you hear about this all the time and you wonder why there's like this honestly the the they that demographic went more for Trump than typ I mean Kla Harris won young people really young people by I think four points or something like that but it wasn't six points I don't remember the exact polling but it was not the typical shift you would see I mean with young people voting Democrat or or numbers margins whatever. Yeah. Yeah.
No, it it's it's it is it's a worrying moment. Um and I think if this trend continues for another couple years, we'll be able to see again like there's a lot of questions with AI in one of the biggest questions is like long-term like can the firms keep this up? like can they keep shoveling sort of the kind of work that a college grad would do into an AI system and or are they going to regret doing that in two years?
Um so I think we're at this incredibly precarious moment and like I said earlier it can do a ton of damage if companies just decide to sort of downsize, decide to lay people off. Um, and it's going to do damage in particular to those areas and those parts of the economy that are sort of, you know, more vulnerable to AI that have jobs that even if it, you know, uh, if their work isn't as well replicated or isn't as good when it's done by the AI, that maybe companies can just justify it in the long term. And these are some of the most important parts of our culture, right?
artists, illustrators, writers, uh designers, journalists, uh there, you know, AI can in a lot of cases produce a faximile of the work that's done by these people enough to maybe, you know, uh just get a manager to sign off on it or to get used to some lower standards or maybe they decide that consumers, you know, don't need the humanmade stuff as much. And and that's what and and that's what I think is really dangerous about this moment. It's a great sort of dulling down and sort of, you know, just sloppification of not just like the internet but also industry after industry.
Well, you mentioned Doge earlier and I think it's not irrelevant that the that effort those efforts by Elon Musk and the Groers that he hired um and Trump like they were firing people. Yeah. And they were using AI to do it.
Yeah. And uh this is between the tariffs and the recession it may cause and Doge trying to explicitly funnel people back into the private sector or at least unemploy people and then using AI to do it as companies lay off people with AI. I mean, it feels like this is some labor discipline that is connecting all of these interests from the Republican administration to the tech overlords that have backed it fullthroatedly.
Yeah, 100%. That is the through line. Um, and it extends uh directly to what we're seeing happen this week because take a closer look and that's that's what I what I've done this week.
I've been talking to people who who've been involved in writing um some of these AI bills in California that are now threatened uh by this ban. Um I talked to uh Assemblyman Isaac Bryan who has a couple of the bills. Um and and look at what these what these bills are doing.
They are they're not even crazy bills, right? They prevent abuses of AI systems in the workplace primarily, right? They say if you're going to use an AI system to like do hiring and firing, then you have to let people know, you have to put certain measures in place so it doesn't discriminate.
Uh you another bill focuses on workplace surveillance and says you have to notify workers if you're going to use an AI system that ingests their data. So these aren't even really like crazy like no AI bills. These are bills that just seek like I would call them common sense sort of interventions in how an AI can be used.
But the AI companies are looking at this and saying, "Oh, this is going to, you know, this is a going to like cost us a little bit of money if we have to make tools that can help people comply with this. " Um and and B it cuts into the you know primary value proposition of these tools which as you just said is uh is is if it if not automating labor or automating jobs outright then providing new tools for management for labor discipline and and control and anything that cuts into that is not tolerable to Open AI and to Google and to the the companies that are selling what essentially amounts to automation software. And so they're looking at these bills going, "Okay, these these costs could could add up fast.
Maybe we can just go right to Congress. We've got our buddies in there and maybe we can just make the case that we can just block them wholesale, which is again crazy. It's crazy.
" Yeah. Yeah. Yeah.
I mean, and and it's just it's it's as you're when you talk about Uber, it's the same thing. It's just on com on on steroids here because Yeah. the Obama Well, look, the Obama administration was pretty pretty pro gigs.
I mean, that is that's the bipartisan thing here, which is like that we that the Obama administration empowered these people with a lot of subsidies, Tesla, but also like, you know, Mark Zuckerberg and all of it, and now they're here, and this is the monster they've created. And this happens, it feels like, with every industry. Then they just shift to the party that's explicitly pro- capital and they get what they want it looks like.
Yeah. Yeah. Exactly.
And you mean even beyond that, Obama sort of empowered them ideologically as well, right? He helped them create this aura that they're building the future that Airbnb and you know Facebook at the time are somehow inherently progressive because they're tech companies because Google had a tagline that said uh don't be evil. Um and you know they did help and by cozying up you know that was a symbiotic relationship you know Obama and they did it with Starbucks as well who just uh tried to fire a bunch of people with AI but anyway this it is it is a throughine right like the Obama White House had a compelling antitrust case against Google 10 years ago and it declined to take it up.
Uh the same thing it sort of let Uber and and Amazon and the other tech companies um grow in power and kind of turned a a blind eye because it was again yeah it was good for both parties at the time. And now we have a new heavily monopolized or duopolized industry that is dominated by some extremely powerful players. And it's extremely difficult now to unwind this.
And now they quite literally, you know, want to have this power to to to again, as you said, one up Uber. They want to take it to the next level. Open AAI's charter.
And this is really important to understand that what they describe AGI as the way that they're chasing. And there's all this, oh, is this AGI? Is this not is AGI?
What even is AGI? Well, OpenAI has a very specific way of defining that. And that is basically as providing a system that can do the work that humans do for most economically valuable work.
So that's their pitch to investors to backers to partners that we are going building a tool that can replace most humans and do most economically valuable work. When we get there that's you that's how you'll know that we we we've hit it big and this is our pitch to you. you know, hitch your uh wagon to this ride and we will give you a tool that can replace human.
Like this is the end goal. Again, open question if they'll ever get there, but that is explicitly what they are trying to do. Uh Brian Merchant, technology reporter at the AI Now Institute, publisher of the newsletter Blood in the Machine.
It's also the name of his book, but you can check out his Substack um at Blood in the Machine. I'm not sure about the link, but we'll put that below uh wherever people are listening to or watching this. Brian, thanks so much uh for your time today.
I really appreciate it. Yeah, thanks so much for having me. And if you have a story about how AI has hit your job, I'm currently collecting these.
I've already heard from over 50 people with great testimonials. Um please, yeah, just send them through my newsletter. It's an ongoing project.
And thanks so much, Emma. Of course. I mean, before we let you go, here we have some people writing in.
Um, soCal Ryan, "My friend is a social worker and therapist, and her employer wants to record sessions so they can use them to train AI to replace her as a counselor. " As if there's a simple formula to the human soul and our needs. Pardon my French, but f that.
Well, well, save it for the fun half with the cursing, but here it is. Mr Melancholy, data entry employee here. My leadership openly brags to us about how AI will make us obsolete.
Always expecting to be fired. We're gonna get many more stories. So, send it to Brian.
Send it to me. Just two weeks ago, I went to a a a strike where therapists were on a hunger strike in LA in 2025 because uh their work is being automated like an assembly line, they said. So, this is happening and we need to understand it and we need to fight back against it.
Where they're starting the therapy is supposed to be about human connection and this is one of the frontiers. I mean that is like Yeah, we're talking dystopian novels here, of course, but um not not in the take over the world way, the like you have no human connection way. No.
And I mean some of the groups that are fighting hardest against this is where you can see the the real imple I mean a lot of teachers are standing up against the use of AI in classrooms. um nurses who are being forced by hospitals to use AI to diagnose their patients and it becomes a real life or death struggle sometimes to have to override this system in a heated moment to get patients the care and medicine they need and that hospitals are saying you have to use AI first. So you can see sort of some of the battle lines being drawn and and and people really pushing back and it's only going to get uh more heated uh before anything gets better.
Brian Merchant, thanks so much much for your time today. really appreciate it. Yeah, thank you.
Hey folks, don't forget to hit the subscribe button and check out our daily show. We do it every day at 12:00 p. m.
Eastern for about 2 and 1/2 hours. We even take phone calls. You should check that out.