Transcript of Tech Critic Ed Zitron Interview: DOAC Podcast

So a data center or indeed the GPUs you put inside an AI data center.

STEVEN BARTLETT: (00:06:43 – 00:06:48) Okay, so you’ve got a data center. And then you have these GPUs, which are like computer chips.

ED ZITRON: (00:06:48 – 00:08:01) So AI GPUs are much bigger, much more power intensive. They take a bunch of high bandwidth memory and, because of how many of them you need — you need thousands of them, tens of thousands, hundreds of thousands in some cases — you need a bunch of power.

So an example, OpenAI and Oracle are building a data center in Texas, in Abilene, Texas, 1.2 gigawatts called Stargate Abilene. Within that, with each one of the eight buildings, there’ll be 50,000 NVIDIA GB200 GPUs. So the City of Bristol takes about 780 to 800 megawatts of power a year, right? Well, Stargate Abilene is condensing more power than that, 1.2 gigawatts, into a space around 1,172 times smaller. City of Bristol’s about 1.2 billion square feet. Stargate Abilene’s about 998,000.

So you’re condensing all of this power, all of this money, all of this labor into this one spot. And all of these data centers cost billions of dollars. All of these companies other than Microsoft have now had to take out debt. And the thing is, they’ve spent over 1 trillion dollars so far and they want to spend another trillion dollars next year. And for what? To make tens of billions of dollars, most of which comes from two unprofitable companies, Anthropic and OpenAI.

Is the Adoption Real?

STEVEN BARTLETT: (00:08:02 – 00:08:36) One of the rebuttals to that would be that the customer adoption of people using OpenAI and Anthropic has been absolutely insane. These are the fastest growing products in all of history, especially as it relates to technology. If we just focus in on technology, hundreds and hundreds of millions of people, billions of people are using these tools every single day for things that they have subjectively decided are problems they need solving. So money is a lagging indicator of value. So one would argue that they’re just investing ahead of the monetization options.

ED ZITRON: (00:08:36 – 00:10:09) Let’s start with this adoption. Is it honest adoption when you are forced to use generative AI? When you load Google, when you load Google Docs, Gemini screams in your ear. When you load Word, Copilot’s bugging you. When you use Amazon, whatever ruthless AI is, it wants to, has opinions on what socks you’re buying. This is the largest non-consensual push of technology in history.

ChatGPT, for example — every single media outlet has been screaming about this for three years. They’ve been saying, “This will take your job. You must use this. If you don’t use this, you’re going to be falling behind.” So people are using it because they’ve been told to use it constantly, and they’re using it like search predominantly. And that’s partly because Google fell behind search, and also because it’s better at ingesting queries sometimes.

Sometimes if you use a generative search, it’s like a trolling vessel. It’s not very good at specifics. But if you’re like, “Does this thing exist? Has this person ever said anything like this?” It’ll still probably get it wrong, but it’ll scour the ocean for you. Nevertheless, that’s not worth a trillion dollars. None of it is. The amount of money being sunk into this is just incomparable to anything. Railways, it blows everything out of the water because there is no post-bubble story even for this. AI GPUs are not useful for other things either. It’s a directionless egregore of capitalism, this headless beast that lumbers around desperate to seek out growth everywhere in the hopes that if it harasses people and scares people and demonizes labor enough, people will be forced to use it.

STEVEN BARTLETT: (00:10:10 – 00:11:12) The reason I pause is because I just — I think about my own company. Obviously everybody thinks about their own personal situation. So you have people listening now that don’t use any AI tools, then you have people that are using it for everything from coding new software tools to everything they write to images, whatever. And when you look at the stats around enterprise adoption, it says 88% of organizations regularly use AI at least once for one particular business function. And I’d say in our company, 95% of people use one of these AI tools like Anthropic or ChatGPT or Gemini every day. And that exists on some kind of spectrum, from the super users that are using it probably every hour of every day for almost everything, to someone maybe higher in the executive team that’s using it less, because their job doesn’t require it as much.

And when you look out into the world at how the world is changing from a content perspective, if we’re looking at generative AI, it is obvious that these tools are being widely adopted. Part of the symptom is the AI slop you see all over the internet.

ED ZITRON: (00:11:12 – 00:11:12) Right.

STEVEN BARTLETT: (00:11:13 – 00:11:19) So I don’t know, this idea that it’s not being used, I struggle with.

ED ZITRON: (00:11:19 – 00:12:07) It’s being used. Here’s the thing with the slop. Before we had AI slop, we had SEO slop because Google incentivized doing the lowest common denominator that would rank well in search. It’s a whole story about how they pulled back spam guards thanks to Prabhakar Raghavan, which we can get into, where they made the internet worse by allowing worse content to rank higher.

It’s why, when you used to Google, “Oh, best washing machine,” there’s 11 different horrible blogs that read like somebody got a concussion. They are built to rank rather than be read by humans, or built to be made good. So AI helps weaponize that at scale. Yeah, you can make a bunch of generic slop. We’ve had slop for years. We’ve just found a slop machine.

But then also there’s the problem of cost. So when you use AI services, you burn tokens, and it’s per million tokens.

STEVEN BARTLETT: (00:12:07 – 00:12:08) So that’s a token.

ED ZITRON: (00:12:08 – 00:12:10) So it’s around three quarters of a word.

STEVEN BARTLETT: (00:12:10 – 00:12:11) Okay.

ED ZITRON: (00:12:11 – 00:12:12) So it’s characters.

STEVEN BARTLETT: (00:12:12 – 00:12:19) So the AI companies have a currency in which they charge you, like a taxi in New York has a meter. And they call it tokens.

ED ZITRON: (00:12:19 – 00:12:20) Yeah.

STEVEN BARTLETT: (00:12:20 – 00:12:23) And every word — let’s just say for ease it’s a word — you’re paying per word.

ED ZITRON: (00:12:23 – 00:14:25) About a word, yeah. And it’s per million tokens. So you’ll be charged per million input tokens, the stuff you feed into it, like a document or a code base. And the output tokens are both the stuff it spits out at the end, but also when it thinks. So, okay, you’ve asked me to give you the best restaurants in this area of New York. I should find the best restaurants in New York. All of that’s output tokens as well.

However, when you are paying for a monthly service, you don’t see any of that. All that — put all that crap to the side. They just have rate limits, so you can use them a certain amount, and then when you run out, but they kind of obfuscate what that was. Now, someone recently found — SemiAnalysis actually found this, a big analyst group — they found that on a 200-dollar-a-month ChatGPT subscription, you can burn 14,000 dollars’ worth of tokens. And on Anthropic, you can burn 8,000 dollars, for 200 dollars. That is how most — and even on the 20-dollar-a-month service, you can burn 400 dollars.

Now, most people don’t realize that. Most people have no idea what AI costs. Most people just think, “Oh, it’s 20 dollars a month.” No, all of these companies run at a horrifying loss. OpenAI lost 20.9 billion dollars last year because people can burn as many tokens as they want. And when they tried to move everybody on the enterprise side — so companies bigger than 150 — onto actually paying the cost of AI, in around March of 2026, to quote Sam Altman, they said people have a big problem with it. “I think it’s a huge issue,” which is not really what the heir apparent of tech history is meant to be saying.

But the point is, enterprises immediately started freaking out. Uber burned through their entire annual token budget in three months. So suddenly, after everyone’s saying AI is the most productive thing ever, it’s amazing, it’s changing everything, the moment people actually had to pay for it, they go, “Oh, I don’t know actually. Obviously we all love it, it’s all great, right? But it’s costing too much.” So we now need to reduce the cost because people are just dumping stuff into it, being like, “What do I do here?” And getting whatever the median is out, because that’s what these things do. They provide the median answer.

Subsidized Demand and the Innovator’s Dilemma

STEVEN BARTLETT: (00:14:26 – 00:14:47) So essentially someone like me who’s a power user of these tools, I could be costing Anthropic or OpenAI 1,000 dollars, but they’re only charging me 100 dollars, let’s say. So they are having to subsidize 900 dollars of my usage because of the electricity costs and the costs at their data centers. And so your assertion here is that that is unsustainable.

ED ZITRON: (00:14:47 – 00:16:08) Yes. And just to be clear, they’re probably not one for one dollar. It might be 30 for — we don’t know. I think it’s unprofitable. These companies don’t disclose them even in their audited financials. They play funny games with how they categorize things. But nevertheless, yes.

And on top of that, the way that you stand up inference, which is the thing that creates the output, within these data centers — you’re not just saying, “Okay, turn the inference machine on, let’s go.” You are standing up the GPUs necessary to take in the demand. And if you buy too much, you’ve wasted the money. You have to pay for the hourly GPU use regardless. If you buy too few, your customers can’t use it. They get pissed off at you, they cancel, they go with someone else.

But nevertheless, yeah, they would get demand selling 20 dollars or 40 dollars for a dollar. And that’s what these services do. And really the simplest way to explain it is, if they were actually profitable — if they believed that these services were worthwhile and that they were worthy of the cost, they’d charge it. Regular people wouldn’t be able to get a monthly subscription. They’d just be paying what it’s worth. Unless, of course, there was an economic problem.

And it’s very simple. You pay when you use an LLM regardless of whether you get what you want. When these things hallucinate — say you’re doing something, you’re coding something, and they go through a codebase and they mess up a bunch of stuff and they break a bunch of stuff — you’re paying for that. You’re paying for it whether it works or not. Unless, of course, you’re using one of these subscriptions.

STEVEN BARTLETT: (00:16:09 – 00:16:43) I think the really interesting point is, are they spending ahead of the value showing up, which is I imagine what they would argue, or are they spending all of this money and subsidizing all of their users in a way that’s unsustainable and that will never be justified? Because you think back through the history of technology, you often get people losing money to grab market share, right? And they’re also focusing on bringing the costs down and making it more profitable for them as well. But they can’t afford to underinvest.

ED ZITRON: (00:16:44 – 00:20:07) If they were bringing the cost down, they would have brought the cost down, which they have not. It seems to be getting more expensive. In fact, inference providers don’t seem to be profitable. Even the companies renting out GPUs don’t seem to be profitable.

I imagine it wasn’t like they started out and they were like, “Shit, this is unprofitable at the beginning. We know it, screw it, we’ll keep doing it anyway.” I don’t think it’s some big conspiracy. They probably thought at some point, “Yeah, this will go profitable, the chips will catch up, customers will pay for the overwhelming value,” because you don’t know in 2023 where it’s going to be in 2026. You assume it’s going to go up. That’s the nature of venture capital. They should have stopped in like 2024 when OpenAI lost over 5 billion dollars. They should have been like, “Yep, this is not going to work.” But they kept going because it helped number go up so much. It helped stock values pump. It helped everyone pump. It helped Nvidia pump, Microsoft, everyone.

And not from the revenues — because here’s the funny thing about Google, Microsoft, and Amazon. People for years have been saying their AI bets have paid off. “Wow, their AI bets have paid off,” as these companies refuse to say how much they’re making from AI. But because their existing businesses continue to grow, and did so, by the way, through price increases, changes to how Google and Meta did advertising. Amazon bumped up prices and changed how they did that. Actually, Amazon started a remarkable ad business during this whole time as well, and the selling through the Amazon platform. Anyway, nothing to do with AI, but because number go up, because revenue go up, everyone went, “It’s AI,” because these companies wouldn’t spend 1 trillion dollars for no reason, right?

Except in fiscal year 2026, which just ended for Microsoft — annoying, I know — they made total, according to Bloomberg, about 34.33 billion dollars. 24.1 billion dollars of that was from OpenAI. So that leaves them with about 10 billion dollars in a year when they spent 115 billion dollars on capital expenditures and intend to spend 175 billion dollars next year. The math does not make sense.

I imagine their plan was, “Okay, this is just going to get exponentially more valuable, and at some point the costs will be outpaced by the return.” Problem is that large language models need a bunch of money to train them. They need constant data flows through it. They need customized data. It’s just this big expensive monster.

And when you try and talk to people about it and you try and say, “Hey look, this is really bad. Nvidia has sold, I think it was, 215.9 billion dollars in their last fiscal year worth of GPUs mostly.” And you try and go, “Yeah, that’s to support like 22 billion dollars of revenue total in the entire world outside of these two companies that literally require money being fed into them, sometimes by Nvidia, to keep alive.” When you tell people that, they go, “Well, companies just lose money, right?”

Because, to quote Ed Elson from Prof G Markets, we have this cult-like worship of the wealthy where we think that someone wouldn’t spend all this money for no reason, right? Because reconciling with that — with this idea that the ultra wealthy, the ultra powerful didn’t get there through big brains, they didn’t get there through anything other than luck and opportunism and getting an MBA perhaps with the right people, that they just got there because they’re regular people and they just happen to be in the right place at the right time — reconciling with that and realizing that the world is not controlled by people like a meritocracy is kind of grim. So it’s easy to be like, “No, they’re not making a mistake. I must be missing something.” And that’s what they want.

The Car Versus Horse-and-Carriage Analogy

STEVEN BARTLETT: (00:20:08 – 00:20:17) So I think back through the history of technological breakthroughs, and I think about — I mean, you can look at different industries. And one of my favorite books on this subject is The Innovator’s Dilemma.

ED ZITRON: (00:20:17 – 00:20:18) I’ve read it.

STEVEN BARTLETT: (00:20:18 – 00:21:49) And one of the things it talks about is how the innovation that ends up taking out or transforming an industry often starts worse, doesn’t make economic sense. None of your customers are asking for it. And this is typically why we end up ignoring it. So you’ve got horse and carriages in the 1800s. Amazing form of transport, according to the people of the 1800s. And then you have this thing called cars come along.

Now, the problem with cars is they broke down all the time. It’s kind of like AI hallucinates now. They were more expensive and the economics of it didn’t make sense. You might as well walk than buy a car. There was a law at the time that meant you had to walk in front of it with a red flag and wave, and someone had to employ someone to walk in front of it waving a red flag. Obviously, it’s worse. It’s like a worse solution. However, these things that are disruptive innovations, they have a higher ceiling of growth, and so they eventually overtake the horse.

And when I think about that analogy in the context of all of this, I go, “Okay, it’s imperfect at the moment.” The economic models aren’t perfectly ironed out. They’re still figuring out how to make it cheaper, the infrastructure, et cetera. But if you think about the rate of improvement versus other, let’s say, coding — how much could I train a human coder to improve and increase their output versus an AI agent? One would go, if you just imagine any rate of improvement in these AI tools, at some point, if you just imagine a 5% rate of improvement per month, and then you imagine a 5% reduction in cost, which is what we did with the internet, what we did with cars — Moore’s Law.

ED ZITRON: (00:21:49 – 00:22:03) Moore’s Law is a theory, and Moore’s Law is not with GPUs. So let me actually explain. Nvidia — Nvidia invented, I think it was in the 2000s, they put out something called CUDA, which is the underlying software library and the way to run software on GPUs.

STEVEN BARTLETT: (00:22:03 – 00:22:03) Mm-hmm.

ED ZITRON: (00:22:04 – 00:22:24) Took them a solid decade or more to make it something where they could do data analytics, one of the early things, MapReduce and such. And then when AI came along, they’d had lots of experience with it. But nevertheless, this company has got more money, more attention, more geniuses behind them, more people focused on making their things more efficient than anyone could ever ask for.

STEVEN BARTLETT: (00:22:25 – 00:22:27) And Nvidia, for anyone that doesn’t know, makes the chips.

ED ZITRON: (00:22:27 – 00:22:32) They — so, and that CUDA thing I mentioned, they were the ones with CUDA, and CUDA allowed generative AI to grow.

STEVEN BARTLETT: (00:22:32 – 00:22:37) Okay, so they’re chips. And chips are needed. Those are the things that go into the data centers.

ED ZITRON: (00:22:37 – 00:22:40) And their specific chips are the ones where you can run AI software on it.

STEVEN BARTLETT: (00:22:40 – 00:22:40) Okay.

ED ZITRON: (00:22:40 – 00:23:30) So the training runs and also the inference. Now here’s the thing, the car example — back then you didn’t have pretty much every mathematician and scientist going into the car industry. You didn’t have the combined world’s governments never shutting up about this. And by the way, giving them credit early — since 2023, they’ve been saying this is inevitable. Even in what you said, 5% improvement — I don’t even know how you’d measure that, because a junior software engineer can still experience things and learn things from context, from how people deal with problems. And the way that people deal with problems is not as simple as looking at the code or reading some emails. It’s context cues from speaking to a person. It’s being in different environments. And there are uses for LLMs in coding, I don’t dispute that. But even saying 5%, what does that mean? Is it better at Rust? Is it better at C++?

STEVEN BARTLETT: (00:23:30 – 00:23:36) I’d say productivity. So just like, yeah, shipped. If we did it in the context of coding, it would be like shipped code.

ED ZITRON: (00:23:36 – 00:23:59) That’s the thing. That would be like, “He’s the best writer in the world because his newsletter’s really long.” That’s an insane way of valuing it. With coding, it’s even difficult to evaluate, because “is the software out there better” is actually a great way of evaluating it. And I would say uniformly not. I would say the standard of software across Google, Microsoft, Amazon, Meta especially — God, Meta’s a monstrosity —

STEVEN BARTLETT: (00:24:00 – 00:24:00) — is worse.

ED ZITRON: (00:24:00 – 00:24:23) GitHub. Someone posted on Twitter earlier today, “We should get a notification when GitHub is up rather than when it’s down, because that would be more reliable.” Microsoft, one of the largest companies in the world, and they can barely wipe their own ass when it comes to GitHub. The quality of software is going down, weirdly enough, as more people use LLMs and more businesses demand — and I really do mean demand — that people use these services.

Defining Hallucinations

STEVEN BARTLETT: (00:24:23 – 00:24:51) So on this point of, if we go back to this horse and carriage and car analogy — say that we’re at whatever point today, if you imagine any rate of improvement in the technology, which we have seen since ChatGPT came out. I remember when ChatGPT came out and I was in Asia and I was there showing it to my fiancée, and I was like, “Look, it can do this.” And it was hallucinating once in a while and getting things wrong. I actually don’t have that experience anymore. I have moments where I believe its reasoning is weak, but I don’t have outright hallucinations anymore.

ED ZITRON: (00:24:51 – 00:24:53) See that? I disagree.

STEVEN BARTLETT: (00:24:54 – 00:24:56) So give me an example of what you define as a hallucination.

ED ZITRON: (00:24:56 – 00:24:58) Okay, great one. So I have a Bloomberg terminal.

STEVEN BARTLETT: (00:24:58 – 00:24:58) Yeah.

ED ZITRON: (00:24:58 – 00:26:54) The very useful thing they have on there is “Ask B.” So when you do a Bloomberg inquiry to, like, look up what we think Nvidia’s revenue is going to be next quarter, it runs something called BQL, which is its own programming language. Now, instead of having to learn that, you can just type into Ask B and it will generate it and run it for you. And so you get pulled up and you know where the data’s coming from. It deals with hallucinations really well.

The other day I was like, “You know what, get a little spicy.” I’m going to look up the growth rate of stocks of Microsoft, Google, Meta, and Amazon over the course of five years, I think it was. And I was about to — I was copy pasting it over to something, looked at it in Excel. I was about to, I was writing the newsletter. I went, “Microsoft stock’s never been 575 dollars a stock.” You know what? When it’s a cute little thing like, “Oh, it’s a stock price,” and I kind of caught it, it was no harm, no foul. That’s fine.

But when you’re talking about, I don’t know, like a transcribing tool for a doctor or a financial model that a hedge fund is dependent on, at that point it becomes a little more dangerous. And the thing is, a hallucination with a software package, for example, refactoring a code base and it leaves the door open security-wise, or it just breaks something. And maybe you’ve been vibecoding for six months. You haven’t really been coding with your own hands for a while. Maybe you’ve forgotten a few things. You have this slop to look for and you go, “Fucking hell, I’m not doing it.” And so the problems become multiplicative and I don’t really know how you train them out of that. And they’ve certainly not succeeded.

So on one hand they have got better, but one of the main ways they evaluate them getting better are benchmarks that are adjusted specifically for large language models, because you can’t just have them do tasks. They’ve got better at that. They’ve found some tasks they can have them do. But even them — like Meta, M-E-T-A, they have this thing where it’s like, “Check out this chart, look how much better it’s getting at running tasks. Wow, it can go for an hour.” And then you look, it’s like, yeah, and successfully completing them 50% of the time.

STEVEN BARTLETT: (00:26:54 – 00:27:57) They have a hallucination leaderboard and it really focuses on basic tasks. And it shows that the four-year trend, according to historical data from the Factaria Hallucination Leaderboard, shows that hallucination rates on simple summarization tasks have plummeted from around 21.8% four years ago, down to roughly 0.7% on today’s top frontier models like Gemini and ChatGPT. Again, the point of nuance here is that these are on simple tasks, which is kind of what I’ve experienced. I’ve experienced that on day-to-day things, that it hallucinates less. Again, rate of improvement thinking — if I just imagine the trajectory to continue, there is going to become a time where hallucinations become rarer than they are today, increasingly.

And also what I’d say is, when I think about other technologies, there’s two more points. Other technologies at their inception, when they first came to the world, like the internet, also had technical difficulties. I remember growing up with dial-up modems, and I couldn’t go on the phone at the same time as going on the internet. I’d have to stop RuneScape upstairs to go on the phone. And you thought, “This is crap, this technology is crap.” All the —

ED ZITRON: (00:27:57 – 00:27:58) I don’t know, mate, I loved it.

STEVEN BARTLETT: (00:27:59 – 00:28:58) Yeah, I know, it felt like magic. And then in hindsight you go, “Wow, I now have Starlink and 5G internet from my phone. It’s unbelievable.” You couldn’t leave the house with internet before. And that’s what I mean by the rate of improvement thinking.

I’d say the last point is, we often compare AI to perfection, right? Whereas that’s not actually the alternative in the working world. Like, if I wanted to do, let’s say, a simple writing task, I should compare AI to my alternative way of doing that simple writing task, which is both measured in my time, and my ability to hallucinate as a person who doesn’t know everything, or if I’m hiring someone, an intern who might also be prone to hallucination or have gaps in their knowledge. So it’s not actually that we should compare AI to perfection. It’s AI to the other alternatives. And if someone hallucinates 0.7% of the time, but knows way more and is faster, maybe on a net basis, that’s a good trade. Maybe I should use AI.

The Intern Analogy

ED ZITRON: (00:28:59 – 00:30:06) Let’s start with an example. Someone I love dearly, Matt Hughes, my editor, lives outside of Liverpool. Wonderful guy. I don’t pay Matt Hughes because he knows everything. I pay him because he has incredible context and a ton of knowledge, and he’s willing to expand it and work with me and give moral support. And he’s a great editor, but he’s also someone who gets into the guts of it and has the experiences of it. He’s a decorated tech journalist, and on top of that, a wonderful, loving being with empathy and joy in his heart for the stuff he loves, and absolute venom for the people he hates. That’s — I can’t get that from a large language model.

But on top of that, I push back on just the assumption there. When you say “knows everything,” what good is something that knows everything when it sometimes doesn’t know anything? Are you really paying an intern for something basic? Are you really going to them and saying, “Yeah, can you look up what the date is?” No, you’re doing that on Google. Whatever the task is, you are trying to also train an intern. The point of an intern is to train them and turn them, take them out of Pinocchio status. But it’s also — an intern learns, an intern gets context, an intern learns your habits.

STEVEN BARTLETT: (00:30:06 – 00:30:07) An AI gets context and learns.

ED ZITRON: (00:30:08 – 00:30:08) No, it doesn’t.

STEVEN BARTLETT: (00:30:08 – 00:30:09) It doesn’t learn.

ED ZITRON: (00:30:09 – 00:30:28) I mean, it doesn’t. The way it learns is you create a giant Claude.md file that it sometimes doesn’t read, sometimes does read. You create a harness. It’s like it’s Pee-wee’s breakfast machine from Pee-wee’s Playhouse. You have to do all these contrivances to mitigate the hallucinations. And even then, at the end, how much effort have you put in?

STEVEN BARTLETT: (00:30:28 – 00:30:35) But, okay, this is an extreme simplified example. If I went on my Claude now and said, “What’s my dog’s name?” It would know my dog’s name.

ED ZITRON: (00:30:35 – 00:30:39) Jesus Christ. This company raised 95 billion dollars this year. No, but you know what I’m saying?

STEVEN BARTLETT: (00:30:39 – 00:30:53) I’m using an extreme simplified example to show that it can remember things from the past. Obviously, it knows much more complex things as well, but I just use that as an example. So we accept the fact that it does have memory of the past. It has files it can access that have stuff on it.

ED ZITRON: (00:30:53 – 00:31:30) Yeah, but that’s not the same as memory. And it’s also just — okay, so it remembers your dog’s name. It might remember your habits. It might be able to read things you’ve said before. Does it know your moods? Does it know what’s going on in the world around it? Does it have good days and bad days? Is it there for you? Because it’s just a text machine.

And the thing is, the intern example — an intern is something that can grow. It’s something that you invest in. That’s not something you do through feeding files and text to it. The way that we store memories ourselves, the way in which we accrue experiences, is a maelstrom of emotion and feelings and facts.

STEVEN BARTLETT: (00:31:30 – 00:32:19) Completely different. So I think there’s two things here. There’s the process in which something happens, and then there’s the output. So the process you were describing, the process of how a human does memory, right? The way that an AI does memory is different. But the thing that people care about is their value in the output — i.e., if I dump all of my files into Claude, I don’t really care how it processes it, as long as when I ask it, “What’s my revenue?” It has the number.

And one could say the same thing about training someone. You could say you teach them, you put lots of effort into them, you give them lots of context, you educate them and give them experiences. And then you might come and say to them, “By the way, what’s my revenue?” Now, the processes are entirely different, but the outcome is what I care about. Do they know the revenue number when I ask them? And so I think that’s the part that we sometimes get lost, because I’ve heard this debate about, “Can AI be creative?”

ED ZITRON: (00:32:19 – 00:32:19) Right.

STEVEN BARTLETT: (00:32:19 – 00:32:34) I think the way to answer that question is, it’s about the output. When I ask it to do a creative thing, does it give me the answer? Not, is the process the same as a human process? Because actually, who cares what the process is? People care about — they pay for the outcome, the product.

ED ZITRON: (00:32:34 – 00:32:40) I actually disagree about the process, because Matt Hughes, for example, your editor.

STEVEN BARTLETT: (00:32:40 – 00:32:40) Yeah.

ED ZITRON: (00:32:41 – 00:33:48) Watching him go down a rabbit hole and being there with him, and actually vice versa, him doing the same thing — we wrote these. Well, I mean, we were working on the research. I ended up sitting there for like a day-long session of writing 11,000 words, and he had given me a bunch of notes. Even describing that process, I feel so happy, because it was like us being like, “I can’t believe how messed up these guys — Jesus Christ, they can’t —” like, just the misanthropy of just the horrible, cynical people of asset managers like Blackstone, just learning about them and being like, “It can’t be this,” and having a back and forth with him.

That is fundamentally different because we were both learning together, and the learning process was as much about creating the output as the output itself. When you learn something, you’re not creating the average, which really is what these things do, of the documents it could find. You are not getting particularly novel outputs. If I needed a generic slop output, sure. But I’ve used some of the higher-end LLM harness machines that the hedge funds use, and they all give the same shite. It’s all the same, the same generic reports, the same “we noticed this analysis” things that you can find on any kind of AI slop out there.

STEVEN BARTLETT: (00:33:48 – 00:33:58) What you described to me there, what I heard anyway, is there’s two points of value you’re getting from your time with that. I mean, there’s many more, but you said you’re learning and then you’re getting this book edited.

ED ZITRON: (00:33:58 – 00:33:58) Blog.

STEVEN BARTLETT: (00:33:59 – 00:34:04) Blog. You’re getting a blog edited, which is the output, and you’re getting learning, and you’re also really getting connection and all these other things.

ED ZITRON: (00:34:04 – 00:34:04) Exactly.

STEVEN BARTLETT: (00:34:05 – 00:34:16) But when people sort of think about the value of AI — of course they could use it to learn. But in the example I gave of, “Repeat my revenue number back to me,” I just care about the output. I could use it to learn. I could say, “Spar with me.”

ED ZITRON: (00:34:16 – 00:35:10) What if the revenue number was wrong once? You should have defined deterministic ways of knowing those numbers. You should not rely on them. Even with the terminal running BQL, which I trust, I will double, triple, treble check everything just to be sure. Partly because also the process of learning, for me, I don’t want just a report. I want something that I fully understand and also understand the context around it. I don’t think that LLMs do that, and I just don’t see them getting better in a way that does that, because it’s just not what they do.

And also there’s the other problem of, the more detailed the report, the more likely there are things to be wrong with it. If you are with Matt Hughes, for example, I can trust he’s got it right. I can trust he understood, and I can trust that I can have a back and forth with him that will inform me if I’ve missed something. I can read the stuff that he’s read and actually trust him, because there’s a big trust part as well.

STEVEN BARTLETT: (00:35:10 – 00:35:15) What is the basis of your trust in Matt? Could it be his historical performance?

ED ZITRON: (00:35:16 – 00:35:16) I mean, yes.

STEVEN BARTLETT: (00:35:17 – 00:35:17) Okay.

ED ZITRON: (00:35:17 – 00:35:20) And also the fact we’ve learned half of this stuff together.

STEVEN BARTLETT: (00:35:20 – 00:35:44) But tenure doesn’t necessarily — there’s probably people you know for 15 years who you also don’t trust. So I think I was trying to figure out, what is the thing that’s causing humans to trust another thing? And I guess it would be continual delivery of a commitment made of sorts. And so with Claude, for example, on simple tasks, as we’ve seen from this hallucination leaderboard, it continually delivers for people. And that’s why we’ve seen the fastest —

ED ZITRON: (00:35:44 – 00:35:45) I mean, is that what that board says?

STEVEN BARTLETT: (00:35:45 – 00:35:48) Well, it’s saying, is it getting it wrong? Is it hallucinating?

ED ZITRON: (00:35:48 – 00:35:49) On simple tasks?

STEVEN BARTLETT: (00:35:49 – 00:35:50) Simple tasks.

ED ZITRON: (00:35:50 – 00:35:51) How are those defined?

STEVEN BARTLETT: (00:35:51 – 00:35:51) I don’t know.

ED ZITRON: (00:35:51 – 00:36:39) That’s the thing, though, because this is actually a very illustrative thing of the AI industry. They are the whataboutist masters. They have, like, “Well, look, we’ve got this benchmark that says we’re good at this. And look, the number’s high.” What’s the number mean? What does that mean? And I’m not using this as a critique against you. It’s, when you can’t give a direct answer, you give a side answer.

When you as the LLM industry want to prove your worth, you can’t just be like, “Just use the product.” When the first iPhone came out, I got — it was at Penn State at the time. I felt like the apes at the beginning of 2001. It was immediate. And I showed it to tech friends, I showed it to the most normal people in the world, and everyone was like, “Holy shit, this is —” they were on Razrs, they were on Nokia 3210s. It was obvious the value. Amazon Web Services, same deal.

STEVEN BARTLETT: (00:36:39 – 00:36:40) It wasn’t obvious though.

ED ZITRON: (00:36:40 – 00:36:42) Yes, it was. I mean, I bought it.

STEVEN BARTLETT: (00:36:42 – 00:36:43) To you, to you it was.

ED ZITRON: (00:36:43 – 00:36:48) It was. And I also showed it to a bunch of people because I’m aware that I have bias when I just love gadgets.

STEVEN BARTLETT: (00:36:48 – 00:36:57) But I remember the famous Steve Ballmer, who was the CEO of Microsoft, interview where he is told about the iPhone and he bursts out laughing.

ED ZITRON: (00:37:00 – 00:37:29) “500 dollars fully subsidized with a plan? That is the most expensive phone in the world. And it doesn’t appeal to business customers because it doesn’t have a keyboard, which makes it not a very good email machine. You can get a Motorola Q phone now for 99 dollars. It’s a very capable machine. It’ll do music, it’ll do internet, it’ll do email, it’ll do instant messaging. So I kind of look at that and I say, well, I like our strategy. I like it a lot.”

STEVEN BARTLETT: (00:37:30 – 00:37:37) He burst out laughing, mocking it, because it was so disruptive. It was way more expensive. And it was way different. No keyboard.

ED ZITRON: (00:37:37 – 00:38:37) Well, phones used to be insanely expensive and the carriers would cover them, but you had to sign a long contract. You were still spending 500 dollars. But the thing I’m getting at is you didn’t have to explain to someone what you — perhaps you’d have to get past the cost part, but you could just be like, “Look how good this is.” And then once the app was — the iPhone 3G with the App Store, people were like, “Oh shit, this could actually change things.” Mobile web, even though it was a monstrosity, it was so bad at first, even then you could get your emails and you could just look —

Point is, BlackBerrys were also expensive and were still actually kind of cool, but the way they worked was not like consumer software. They didn’t have the classic GUI. iPhones felt like that. It felt like a cell phone designed even like a computer. It was obvious. It was obvious from the beginning. Everyone — I was dating a girl in the center of Pennsylvania at the time, and everyone I showed it to was like, “Wow, this is incredible.” That to me is the obvious thing. With AI, to this day, when you’re like, “Okay, why is it so amazing?” people still dither. People are still like, “Yeah, you can’t run a business fully with it without this weird system of pulleys and levers and such.”

Growth Curves and the “Honest Cost” of AI

STEVEN BARTLETT: (00:38:37 – 00:39:41) But how come then, when you look at the stats around ChatGPT’s growth — 100 million active users in just the first 60 days after launching. For comparison, TikTok took nine months, Instagram took 2.5 years, and the internet itself, the World Wide Web, took roughly seven years to reach that scale. Over 60% of US adults are integrated into AI tools in their daily and regular routines within three years of the launch, reaching 40% of the population. And that same milestone took the internet five years and personal computers nearly 12.

So I think this is the part that’s giving me dissonance. When I showed my fiancée ChatGPT — okay, it didn’t really work for her at first. But as a sole entrepreneur who — English isn’t her first language, who has to write lots of text and lots of copy and generate lots of images, and was paying a graphic designer to help her make certain images that she couldn’t make herself because she doesn’t have the skills — she would describe it as being transformative for her business. What I’m hearing from you is that it’s not transformative and there’s no value in it for people. But she, if she was sat here now, she’d say it’s transformative.

ED ZITRON: (00:39:42 – 00:39:56) Would she pay the per-million-token rate? Would she pay the actual rate? Because that’s the thing. If this was sold at its honest cost — if people were reacting like that and they were paying two, three, four dollars every time they did something, and they were genuinely — that might be an argument.

STEVEN BARTLETT: (00:39:56 – 00:39:58) What would be the honest cost if they weren’t subsidizing?

ED ZITRON: (00:39:58 – 00:40:02) The actual per-million-token cost, the actual API cost they should charge.

STEVEN BARTLETT: (00:40:02 – 00:40:04) Do you know how much that is relative to what they charge?

ED ZITRON: (00:40:04 – 00:41:49) Oh God, it depends on the model. But there’s actually kind of a point I want to make about the thing you said with the internet earlier. So when I first got on the internet, I had a 33.4-kilobits-a-second modem. Even back then I was like, “Fuck, if this was faster.” And that was immediate, just like, “If this was faster,” because it was slow. You go on like Happy Puppy or something, download — take all bloody day waiting for shareware to download. Immediately you were like, “If I could do this faster, it’d be better.” And even back then, I’m like, “Man, you could probably do video camera stuff with this,” stuff that eventually happened.

And actually, there’s this guy called Jim Covello from Goldman Sachs, in a report he did in 2024 called “Gen AI, Too Much Spend for Not Enough Return,” paraphrasing there. And he made the point that in the run-up to the iPhone, there were thousands of presentations that said, when GSM radios get smaller, when Bluetooth radios get smaller, when Wi-Fi radios get smaller, it is inevitable that we will get something like this. And then he said that there is no such path for AI. There was no roadmap to AI becoming this thing that they promised.

And I must be clear — if these companies had gone out there and said, “Yeah, this is interesting cloud software, it’s generative, it’s really expensive. We’re not sure — not in the ‘ooh, I’m scared’ way — we’re just not sure that this is going to be a disruptive, world-changing thing. It has potential, but we’re going to go slow. It’s really expensive. This is an R&D effort. We’re not going to expose consumers to it,” and actually being like, calling them, I don’t know, language models, and not even calling it — because it isn’t AI, it’s not autonomous, it’s not smart — I actually might respect it.

But this is not what they’ve done. They’ve gone out there since 2022, 2023, and said, “This is the best thing since sliced bread. This is changing everything. This is going to do all your work. This is going to take your job.”

STEVEN BARTLETT: (00:41:49 – 00:41:49) Job.

ED ZITRON: (00:41:49 – 00:42:35) “You’re going to talk to Bing and it’s going to tell you to leave your wife.” All of these crazy things. And what’s funny is, when the writer Kevin Roose, I think it was, was speaking to Kevin Scott, the CTO of Microsoft, about it, Kevin Scott goes, “You know, I’m just glad we’re having this conversation,” instead of being like, “Settle down, Beavis, it’s a website. The website told you something.” It’s just LLMs. They talked it up. And that’s because everyone is talking about what they wish this was rather than talking about what it can actually do.

So this makes it scary to people, deliberately so. It makes it environmentally destructive. Look at the gas turbines poisoning Black neighborhoods — I think it’s in Louisiana, one of Musk’s data centers. Look at the incredible energy draws. It is raising power bills. And also it is creating inflation across all consumer electronics because of the massive RAM demand.

Echoes of the Dot-Com Bubble

STEVEN BARTLETT: (00:42:35 – 00:42:56) Do you know what’s interesting? I almost feel like so much of what you’re saying is true. And also, it can be true that this technology is going to profoundly change the world. And I think back to the early days of the internet as maybe the closest analogy we have — the dot-com bubble. You wrote this great essay.

ED ZITRON: (00:42:57 – 00:42:57) Yes.

STEVEN BARTLETT: (00:42:58 – 00:43:26) Which I found really funny, especially the name, “The Rot Economy.” And you talked about the “rot-com bubble,” about how AI has less value than people think. And in the dot-com bubble, what you saw is huge hype, people overselling the capabilities of their websites and what they were building. But in the wake of the dot-com bubble, yes, 90% of stuff went to zero, but you had generational companies born that changed the world.

ED ZITRON: (00:43:27 – 00:43:27) Right.

STEVEN BARTLETT: (00:43:28 – 00:44:35) And so I do kind of reflect, and that’s what bubbles do, right? Huge hype, overinvestment, investors get crazy delusional, they think everything’s going to change. At the same time, you do have skeptics in these moments. The dot-com bubble had — I mean, the internet itself had the biggest skeptics. In 1998, Nobel Prize-winning economist Paul Krugman said, “By 2005 or so, it will become clear that the internet’s impact on the economy has been no greater than the fax machine.”

In 1995, astrophysicist Clifford Stoll famously — I wrote about this in my book — wrote in Newsweek: “Do our computer pundits lack all common sense? The truth is no online database will replace your daily newspaper. No CD-ROM can take the place of a competent teacher. Commerce and businesses will shift from offices and malls to networks and modems? Baloney. So how come my local mall does a roaring business and the cyber mall gets zero business?”

And then I’ll give you one more from Krugman, the award-winning economist. He said, “The growth of the internet will slow drastically, as it becomes apparent most people have nothing to say to each other.”

ED ZITRON: (00:44:35 – 00:44:44) That may actually be the worst one of those predictions. Like, hang around any bar in Middle America — honestly, the best conversation.

STEVEN BARTLETT: (00:44:44 – 00:44:45) But it’s just all the same thing.

ED ZITRON: (00:44:45 – 00:45:24) I actually — so Clifford Stoll, his piece was interesting because there were some sound points in it. He made points about how an overwhelming amount of bad information out there is bad for society. He’s completely right. Saying how online education would not be a great replacement for regular education — I think we’ve seen that. But there is an economic difference that’s vastly, it’s just completely different.

So the dot-com bubble was actually two bubbles. There was the website bubble, which was just trash on trash on trash. It was just like — I think what was it, Excite@Home bought an e-greeting card company for like a billion dollars. It was insane crap happening that was so small. The big thing that people are thinking about is the dark fiber.

STEVEN BARTLETT: (00:45:24 – 00:45:25) Dark fiber.

Dark Fiber and Subsidized Demand

ED ZITRON: (00:45:25 – 00:46:14) Dark fiber was all of the wires that were put in the ground thinking, “We’re going to have all this demand for internet.” And it turned out that demand for internet — I think the analyst estimate was it was doubling every 90 days, when it was actually doing that every six to 12 months, maybe longer. And so there was a massive overbuild of fiber-optic cable and indeed the transmission stations, to bring that to people’s houses. And there was the assumption that, well, that would all get lit up and people would want it immediately. Didn’t really happen.

Now, the post-dot-com-bubble thing people say is, “Well, but after that there was demand from the internet.” That’s the thing though, that’s very different to demand for generative AI right now. The demand we have for generative AI is predominantly subsidized. Just let’s start there. Predominantly subsidized, and most people experience it and yet are not paying the real cost.

STEVEN BARTLETT: (00:46:14 – 00:46:15) I agree.

ED ZITRON: (00:46:15 – 00:47:42) On top of that, we already have all of the possible marketing in the world. We have the largest, most disingenuous marketing campaign in the history of man pushing this up the hill. We have the apex predator of cloud software, Microsoft — they can only get single-digit billions from selling AI software. And outside of OpenAI and Anthropic, we would barely get 22 billion dollars. And 22 billion dollars is a large amount to you and me. It’s not a large amount of money when you’ve spent a trillion-plus dollars, when you have Anthropic and OpenAI with 1.1 trillion dollars’ worth of cloud commitments.

And on top of that, how does this turn into a post-dot-com-bubble thing? A data center built today is going to be as expensive to run in 2050 as it is today, unless there’s some breakthrough in electricity. But again, that’s not happening with AI. Unless there’s some breakthrough in GPU technology. But we already have Broadcom, Nvidia, Etched — we have every major chip company, Arm, trying to do something about this. And no one seems to magically be able to make this profitable or indeed even less costly. Even Nvidia with Vera Rubin, their more expensive new GPU system — even then they’re like, “Yeah, 10x more efficient.” It’s more dollars per megawatt. They’re all coy about it. They don’t just say, “Yeah, we worked with OpenAI and Anthropic and we found it reduced their cost by 50%.” Easiest thing in the world if it was true. And that’s because it’s not happening.

STEVEN BARTLETT: (00:47:42 – 00:47:49) So are you saying there’s not going to be the demand for — you know, there’s different types of AI, generative AI.

ED ZITRON: (00:47:49 – 00:48:18) Yeah. And actually that’s a good point to make. The reason they use the term “artificial intelligence” is so everyone would lump everything into it. They would lump protein folding — nothing to do with LLMs. Robotics — not LLMs. Autonomous weapons, even as horrible as they are, are not LLMs, because you couldn’t trust them. But they’ve mushed everything into AI so that when you say, “Well, AI can’t,” they’ll go, “Well, sir, you forgot to give us credit.” And also, “AI, it’s working on curing cancer,” when it’s just like, no, that’s not LLMs. Stop giving them credit.

STEVEN BARTLETT: (00:48:18 – 00:48:21) The similarity though is they all need GPUs, all these AIs.

ED ZITRON: (00:48:21 – 00:48:39) And that’s the funny thing. All those data centers that we’re building, all of them are for just generative AI. They’re not for all of the other stuff. They’re not for the cool shit. AI has been around for a long time. Google — a lot of the good stuff that comes out of Google from the search side is AI, but pre-generative.

STEVEN BARTLETT: (00:48:39 – 00:48:46) Well, how would you run the type of AI that sits in a robot, let’s say one of the Optimus robots, if you didn’t have a GPU?

ED ZITRON: (00:48:47 – 00:50:23) So Matic — Matic has this cleaning robot, for example. That thing has not got a little GPU in it. What it has, and may indeed have used some GPUs, but nowhere near as many as they need for generative AI to run the data, feed training data into it so it’s able to clean a house. But when the little bugger’s going around cleaning my floor — Turdsley, I call him, goes around mopping my floor — it’s not like burning money the whole time.

But when it comes to these massive amounts of data centers, Sightline Climate said in February there’s 190 gigawatts of data centers under planning — don’t know about under construction. That works out, if about 12 million dollars a megawatt, that’s like 1.6 trillion to 3 trillion dollars a year in annual demand you’d need for that. We don’t even have 130 billion dollars’ worth of annual demand. And people say, “Well, it’ll grow.” How? When most of the demand is coming from Amazon feeding money to OpenAI or Anthropic, Microsoft feeding money to OpenAI and Anthropic, Google feeding money to OpenAI and Anthropic.

The con side is that we are building these effigies to capitalism, these giant GPU data centers, and people are being told, “Well, it’s for AI, you know, the thing that’s done all this other stuff that’s unrelated.” Or the worst thing I’ve seen, it’s like, “Oh, you don’t like — you like online banking, well, you do like data centers.” There’s a big difference between a data center for regular non-GPU compute, for standing up a server, a content delivery system like Akamai or something that brings the website to you, or how Meta runs Facebook. That is not the same. It takes way less power, mostly CPU-driven. Compared to these giant GPU data centers that offer one thing, one thing only.

STEVEN BARTLETT: (00:50:23 – 00:51:05) But I was doing the research and looking at some of these notes here. It does say that for tougher types of AI systems designed to solve concrete physics, biology, and spatial problems, they require some of the most intense data center infrastructure on the planet. AI systems like DeepMind’s AlphaFold, the protein-folding company used for genomic sequencing and climate forecasting, et cetera, run on high-performance computing clusters. These require immense precision and continuous heavy compute in data centers. Training the brains for self-driving cars requires billions of miles of simulated physics environments. The AI isn’t generating text, it’s learning to navigate 3D spaces and gravity, and relies on data centers.

ED ZITRON: (00:51:05 – 00:52:45) And the thing is, those data centers, they might have GPUs in them. We had GPUs used for this HPC, the high-performance computing, before generative AI. And yeah, that’s how AI has been trained before. That’s how Tesla did it — I believe they’ve had their own data centers when it comes to training the Autopilot system, for better or for worse. That’s how we’ve done it before. Again, that is not why we’re building these data centers. These data centers are being built to sell to generative AI companies, to either train systems or run inference.

These things are being built in this brainless way, where it’s just — maybe this is a good way of illustrating the con. Because everyone saw Google, Microsoft, Amazon, and Meta give Nvidia over, call it, 800-something billion dollars. Because everyone saw that, they went, “Well, they wouldn’t do that for no reason. We’ve got to build more of these things, because there must be all this demand.” Even though the demand — 70% or more of all that demand comes from these two companies who are funded by these three companies.

And that’s the funny thing. The reason that they don’t want to break out their AI revenues is because it will become alarmingly obvious that this was the case. It turns out that the only real big customers — because it’s not like they’re building a few data centers, they’re building trillion-plus revenue potential they believe they’ll get, speculative, entirely speculative. They’re building it because they saw the biggest companies in the world buy a bunch of GPUs and they said, “I want in on that.” They must have diverse customers, right? They wouldn’t just have two unprofitable, failing sons that they’re propping up. They’ve raised 217 billion dollars just in 2026.

Coding Productivity and Google’s Decline

STEVEN BARTLETT: (00:52:47 – 00:52:56) So we know that some of the biggest companies in the world are using AI, generative AI, to write a lot of their code. That is a great productivity gain for those companies, right?

ED ZITRON: (00:52:57 – 00:53:07) I mean, have you used Google or Facebook or Instagram or GitHub recently? Because they’re catastrophically worse. Amazon Web Services went down multiple times because of their AI coding tool.

STEVEN BARTLETT: (00:53:08 – 00:53:09) How is Google worse?

ED ZITRON: (00:53:10 – 00:55:35) Well, I’ll tell the story of a guy called Prabhakar Raghavan, previously one of the heads of ads at Google. In 2019, Google called something called a “code yellow,” which is when they said, “We’ve got a problem.” And it was material weakness in query numbers, which means the amount of times that people were searching on Google Search.

A guy called Ben Gomes, internal at Google, then the head of Google Search, said, “Wait a minute. To increase this number of using Google more, we’re going to have to —” well, what you’re suggesting would mean we give worse answers, because if someone got the answer quickly, that would reduce the amount of queries, right? And people at Google — Shashi Thakur was another engineer who was saying, “Yeah, can we please tell Sundar this? Because this doesn’t seem good. We can’t just increase the amount of queries. That would just mean that people would have to search more, which would make the product worse.” “But it would make them more money,” you say? Yes, because more queries show them more ads. So if you were spending more time on Google —

STEVEN BARTLETT: (interjecting) But is this linked to AI doing code?

ED ZITRON: Oh, I’ll get there. So the problem is that this guy called Prabhakar Raghavan, who was the head of ads at the time, was pushing and pushing and saying, “No, we need to make more queries happen. Got to make it happen.” Nick Fox, who was there as well — I believe he’s actually taken over Google Search — said, “Got to make them go up. This is our new reality.”

Sometime in early 2020, Prabhakar Raghavan takes over Google Search. From then — and this is what I believe, can’t prove it, but if you go and look around the various SEO sites such as Search Engine Journal and the various forums — Google stripped back a lot of the suppression of spammy sites so that people would be on Google more. And then over the course of time, Google wanted to create more queries, and Google Search became much worse. It’s why people always do like “+Reddit” or “from Reddit” or what have you — it’s because the actual underlying search results of Google had got worse.

And then generative AI came, and Prabhakar, wouldn’t you know it, gets put to run part of Gemini. And Google also was having trouble getting people back on Google. And what did they think they’d do? “Well, everyone’s talking about this AI thing, we’ll just put it right at the top so people have to stay at Google.” And actually they’ll use it more because, instead of searching websites and doing that annoying thing where they click away from Google, they’ll just only use Google instead of generating answers — by which I mean giving you search results you click through. Now Google is the answer. Is it right? God, no. It might tell you to eat rocks, might eat poisonous mushrooms. Maybe it’ll give you a few links you could click through. But the ideal situation was that AI was the ultimate form of Google’s evil.

STEVEN BARTLETT: (00:55:35 – 00:55:42) But I’m saying, that’s not the fact that coders could code on Google that’s made Google worse. That’s human decisions that have made it worse.

ED ZITRON: (00:55:42 – 00:55:48) Yes. And then there’s the instability of Google’s platform, which is actually — I should have probably led with that — a problem across the whole tech industry.

STEVEN BARTLETT: (00:55:48 – 00:55:52) Okay, so you’re saying Google is going down more?

ED ZITRON: (00:55:52 – 00:56:05) Yes, Google is less stable. Google Docs is a bug fest right now and has been for a while. Google Sheets, same deal. And the thing is, you’re right, I’m being a little unfair. This is everyone. It’s the same with Microsoft. It’s the same with Amazon. It’s the same across.

STEVEN BARTLETT: (00:56:05 – 00:56:08) How do we quantify that outside of anecdotes? Is there a way to?

ED ZITRON: (00:56:09 – 00:56:35) You’re right. I mean, GitHub downtime is the best example. Amazon Web Services went down, I think, two or three times this year because of AI tools. And honestly, you’re right, it is kind of hard to quantify outside of anecdotes. But I challenge anyone listening to this: go and use a website these days and tell me how well it works. Tell me how buggy it is. Tell me how many problems — even with my iPhone, the supposed best UX in town, even the iPhone is a mess these days.

STEVEN BARTLETT: (00:56:35 – 00:57:00) Okay, so the research says the short answer is yes. Tech downtime and software outages have demonstrably increased over the last few years, and industry data points directly to the explosion of AI-assisted coding as a primary culprit. The problem is hitting the tech industry from two entirely different directions. The code itself is getting buggier, and the sheer volume of AI activity is literally crashing the underlying infrastructure. Interesting.

ED ZITRON: (00:57:01 – 00:57:07) Yeah, that’s because GitHub — people are just writing a bunch of code, pushing it, and thus there’s just more code on there.

STEVEN BARTLETT: (00:57:08 – 00:57:09) That’s interesting.

ED ZITRON: (00:57:09 – 00:57:33) Yeah, it’s a real mess as well, because open source has had this problem too, because it’s well-meaning people. They’re like, “Oh, I learned a bit of code with an LLM, I’m going to go out and do some stuff, I’m going to make this project better.” And these people barely understand what they’re shipping. Or maybe they understand a bit of code and they say, “Dunning-Kruger this,” “I can understand some of this,” and now the code’s all written and just push it right now. So GitHub is flooded with AI code.

STEVEN BARTLETT: (00:57:33 – 00:57:36) This sounds like it’s making humans complacent.

ED ZITRON: (00:57:37 – 00:57:37) It is.

STEVEN BARTLETT: (00:57:37 – 00:57:53) Because we’re going, “Okay, look, I’ve let it write the code for the last 100 lines and it was broadly right, so the next 100 lines, I won’t check them as much.” And that’s human nature — to take shortcuts, to spend less energy on an activity if you can.

ED ZITRON: (00:57:53 – 00:58:46) Right. But the AI’s still making the mistake and we’re still making all the promises of AI. That’s the thing. This thing is meant to be autonomous, perfect. Based on what Sam Altman has been saying for the last few years — this will replace software engineers. Dario Amodei has been saying, “Oh yeah, 50% of white-collar labor is going to go away in the next few years.” These people are promising the world.

Again, if they were saying it would be smaller and they were like, “Yeah, it does have issues and we must be —” none of this, “Oh, what if it wakes up and it’s super powerful?” Just like, “Yeah, it’s probabilistic, it’s going to make mistakes.” And if you don’t know what you’re doing, you don’t really know what you’re looking at, you’re going to miss those mistakes and it’s going to get multiplicatively worse as you go when you don’t know what you’re doing. So yeah, human nature is part of it, but so is the marketing. So are the promises.

Autonomous Vehicles and Job Disruption

STEVEN BARTLETT: (01:00:54 – 01:00:54) My car that drives itself, that is AI technology.

ED ZITRON: (01:00:54 – 01:00:54) Yes.

STEVEN BARTLETT: (01:00:55 – 01:01:21) I sat here with Dara from Uber, and he was saying that in a couple of years’ time, we won’t need drivers for Uber because the cars will drive themselves — they’ll be fully autonomous. And I think, if I’m not mistaken, driving is one of the biggest professions on planet Earth. So when you hear these CEOs saying that there will be job disruption, you say that they are not telling the truth.

ED ZITRON: (01:01:22 – 01:02:02) Yes. Or they’re guessing in a way that’s very good for them. Think about it from a different perspective — Microsoft. Satya Nadella, he’s not going to be like, “Yeah, we don’t know if this is going to work, mate.” Of course he’s going to talk his book and he’s going to say, “Yeah, this is going to replace all workers, it’s going to be amazing, it’s going to be so powerful.” And then he’ll change his tune and say, “Actually, it’s not going to replace workers, it’ll make them more powerful,” because the things aren’t catching up.

Dara from Uber, for example — of course he’s going to say, if this happens, then that would be good for Uber, because Uber would just become an autonomous taxi service. There’s a reason that Waymo’s taken it. I find Waymo fascinating. I think that’s really cool. I think there are socioeconomic problems that will come from it. I think there are actual real problems that will emerge.

STEVEN BARTLETT: (01:02:02 – 01:02:03) What kind of problems?

ED ZITRON: (01:02:03 – 01:02:34) Well, socioeconomically, they’re — like you said, one of the largest employment centers in the world. I mean, just the economics of cabs would fall apart. But again, we are nowhere near that. We’re not even close. Waymo has had to do the smallest rollouts and the most controlled things, because the problem with pretty much every AI system, but especially driving, is not getting 95% of the way — it’s those edge cases. It’s raining, which is a big problem for them in San Francisco. It’s a kid runs across the road, but they’re wearing a high-vis thing. Does it even notice it’s a child?

STEVEN BARTLETT: (01:02:34 – 01:02:44) Again, this is a really interesting but very applicable example of the right comparison to be made — it shouldn’t be autonomous vehicles versus perfection. It should be autonomous vehicles versus human drivers.

ED ZITRON: (01:02:45 – 01:03:10) I don’t know if I agree, because the human driver might make mistakes, sure — but again, not an expert in autonomous cars, just want to be clear. But if we’re pushing autonomous cars out there willy-nilly and we’re not doing so in extremely controlled environments, those edge cases will multiply and be dangerous. Yeah, they might be better than human drivers in some ways, but I was in Vegas the other day and I was in a hotel and I watched a bunch of Zoox cars just get stuck.

STEVEN BARTLETT: (01:03:11 – 01:03:12) The autonomous cars?

ED ZITRON: (01:03:12 – 01:03:26) Yeah, these weird boxy things. And they just blocked the exit. They just all kind of lined up and just fell asleep. I saw the same thing actually happen outside of a hotel when I got out of a Waymo in San Francisco. It just stopped, and then a bunch of cars and another Waymo got stuck behind it.

STEVEN BARTLETT: (01:03:27 – 01:03:29) I’ve seen some human bad drivers as well. I agree.

ED ZITRON: (01:03:30 – 01:03:58) But we have control over deploying these bad or good drivers. We have an ability to roll them out slowly, which is exactly what we should do. I’m not saying autonomous cars are bad. I’m saying we need to be so, so careful and treat them as guilty until proven innocent — because we can prove that. And also they have people overlooking them, they actually have people monitoring the routes. It is something they cannot rush out, and it doesn’t seem like they’re rushing it, which is good. And they’re not promising the world.

STEVEN BARTLETT: (01:03:58 – 01:04:08) I do agree. I’m a big fan of taxi drivers generally, in part because I spend a lot of time in taxis. And I’m not just getting in there because I want to get from A to B. I’m getting in there for lots of other reasons.

ED ZITRON: (01:04:08 – 01:04:09) Yeah.

STEVEN BARTLETT: (01:04:09 – 01:04:52) However, when I look at the stats around what is more dangerous — driving myself or having an autonomous vehicle drive me — there’s a 68% lower overall crash involvement rate when you’re in an autonomous vehicle. Autonomous vehicles experience roughly 2.1 police-reported crashes per million miles compared to humans that are at roughly 4.68 per million miles — so a 55% reduction. And autonomous vehicles show an 80 to 81% reduction in crashes resulting in injuries versus human drivers. So you’re 85% less likely to be involved in a single-vehicle crash, like hitting a wall or a tree, if you’re in an autonomous vehicle versus being driven by a human.

ED ZITRON: (01:04:52 – 01:04:53) I agree, but —

STEVEN BARTLETT: (01:04:54 – 01:04:54) So it’s safer.

ED ZITRON: (01:04:56 – 01:05:12) Also, what’s the sample size of human drivers? We’ve got many, many more years of drivers and many, many more years of accidents. And also, does that not have anything to do with generative AI? If we were just talking about that, we’d be having a different conversation.

STEVEN BARTLETT: (01:05:12 – 01:05:35) I guess the question here was really around job disruption. Like, we look across industries and we go, driving’s a massive profession — is there going to be job disruption because cars can now drive themselves? If we think about white-collar jobs — lawyers and accountants — people sit here and tell me that some of the skills within those professions will be relegated to AIs to do.

ED ZITRON: (01:05:35 – 01:06:21) Here’s the thing, lawyers, for example — great example. Always hearing legal partners talking about AI, never the associates. The associates are the ones that go out and find the precedent. They’re the ones that go and do the grunt work. They’re the ones who are pulling motions half the time. The partner is the one that might be the litigant, may be the client-facing one, but the ones that are actually doing the day-to-day work — I’m not hearing from them. I’m not hearing associates being like, “This is awesome.” I’m hearing a bunch of well-paid people who have sat on ChatGPT and gone, “Yeah, I’m the greatest lawyer ever.” They’re not the ones that I want to hear from — the actual workers.

White-collar labor disruption is not happening. OpenAI had a study that came out, I think, like a week ago, that said there was no connection between spending on AI tokens and revenue per employee.

STEVEN BARTLETT: (01:06:21 – 01:06:22) What does that mean? Could you explain that to me?

ED ZITRON: (01:06:23 – 01:08:09) As in, the more tokens you spend has no correlation at all with the amount of money you make. It’s the second report they’ve put out. The other one was like, hallucinations are mathematically guaranteed, almost — it’s the one thing I respect about that company. Occasionally they just put out a study that’s like, “Yeah, it kind of sucks.”

But the people that are having their lives disrupted work-wise are art directors, transcribers, translators — people who have bosses that don’t care about the output. It’s what they consider cheap work. And the problem is, those people would’ve automated your work away anyway. They would’ve sold it, they would’ve taken the cheapest offer, they would’ve sold it to the Global South. They would’ve taken the shittiest option they could. That is something that AI is doing. And again, those people are not paying the actual cost of AI — they’re using a subscription.

The actual white-collar labor force might have some things that are slightly changing, but there is no evidence of productivity gains. If there were, they would be screaming it from the rooftops. There was an Oxford Economics study last year where it’s like, “Oh, young people are finding fewer jobs because of AI.” We actually read the study, which multiple journalists did not. It was a single line that said, “Yeah, we saw some correlation.” Didn’t give a number, didn’t actually say what the correlation was.

We are so conditioned to believe that the rich and powerful know what they’re doing that we internalize these narratives about, “Well, previous booms lost a lot of money, well, technology takes time to do stuff,” and they are intentionally playing on those mythologies. They are playing on these, knowing that journalists, analysts, investors will believe them. And this is partly because our reality is defined by stock prices. Because the stock prices of these companies went up, we’re like, “Oh look, it must be working, right?”

STEVEN BARTLETT: (01:08:10 – 01:08:17) Both of those things you said were true though, right? That previous technologies didn’t make money at the start. And the other one you said was, “They’ll get better.”

ED ZITRON: (01:08:17 – 01:08:21) But that’s the thing. Okay, because another thing got better, this will get better?

STEVEN BARTLETT: (01:08:21 – 01:08:28) No, but there’s got to be something that they’re saying that is fundamentally not true, because those are two true statements — technology often starts —

ED ZITRON: (01:08:28 – 01:08:29) Okay, I know, I get what you mean.

STEVEN BARTLETT: (01:08:29 – 01:08:29) Yeah.

ED ZITRON: (01:08:29 – 01:09:02) What they are fundamentally misleading people about is how possible it is, how many actual signs they have — because they don’t have the signs. If they had the signs, as in the signs of this getting cheaper, as in the signs of this being able to autonomously do work without the Rube Goldberg machine, and even then, in a reliable way that was making the customer more money, being productive in a way you can say with your whole chest without a series of asterisks — that’s how it is across the board. The people that are most excited about this, psychopaths on Twitter in many cases, are people that I believe —

STEVEN BARTLETT: (01:09:02 – 01:09:03) Psychopaths on Twitter?

ED ZITRON: (01:09:03 – 01:10:22) There are some people on Twitter — because the other thing about this is, this is really unique to the AI industry. I’ve never seen it in any other industry outside of maybe like sports teams. The attachment that some people online have to these companies — if you dare criticize Anthropic, it’s almost this religious attachment.

Good example was this week Bloomberg reported that OpenAI was on track to hit 40 billion dollars in annualized revenue. Month times 12, four weeks times 13, we don’t know, they don’t define it. I saw multiple people online going, “Actually it’s 60 billion, it’s actually 60 billion.” It is like a cult, and it’s a cult of software driven around growth and this idea that by backing the right horse, you will have some grand thing. And OpenAI in particular, in particular Mr. Altman, they have been fomenting this. Thibault as well, one of the guys at OpenAI, they foment this thing online. They build this kind of parasocial relationship with both the large language models themselves and the companies. And one’s allegiance to the companies is so important. It’s truly vile. If only these people gave a damn about, I don’t know, Medicare for All or poverty or things like actual problems in the world versus, “Are we buying enough GPUs?”

Doomers, Danger, and Regulation

STEVEN BARTLETT: (01:10:22 – 01:11:12) Do you know what’s interesting is some of your narrative, one would argue, actually helps them. Because, you know, the AI doomers who have come here and told me — some of the original founding fathers of AI, like Geoffrey Hinton — have told me that what they’re building is highly, highly dangerous and that it will be fundamentally disruptive to society. And it’s interesting because some of the CEOs you’ve mentioned, their historical narrative was also, “By the way, this is really dangerous, and there is a significant chance we could mess up the planet.” And what we’ve seen is this slow pivot away from it, because now they’re getting booed and they’re being attacked. And the pivot almost sounds a little bit like your narrative. It now sounds like, “Actually, no, it’s not going to change anything and you’re all going to be fine. It’s not dangerous at all.”

ED ZITRON: (01:11:12 – 01:11:13) But that’s the funny thing.

STEVEN BARTLETT: (01:11:13 – 01:11:31) And that’s why I’m saying, I actually think there might be a couple of PR people at these big AI companies thinking, “Thank God for Ed.” Because you’re saying, “Actually, don’t worry, everything’s going to be fine. It’s not going to take your job. It’s not going to disrupt the economy. It’s just a fad.”

ED ZITRON: (01:11:31 – 01:12:43) I don’t think they think that. Here’s the thing, I think Altman and Amodei are some of the most deeply corrupt and cynical people in the world. Of course they were going to say — from early 2023, Altman said, “We’re a little bit scared about what we’re creating.” Oh, shut up. I hear that and I feel so frustrated, because I’ve met so many of these rich liars.

And you know why he wants to say that? So you’ll invest in his company and buy the software. So you’ll be scared that if you don’t use AI today, you’ll be left behind in the future, which is their continual narrative — if you don’t get on the train today, then you’ll be left behind. By the way, every single scam and con starts with rushing you. Every single trick in history begins with saying you must do this now. And the best piece of advice I ever got was, if anyone tries to rush you, and it’s not literally a mortal thing — like you are bleeding or on fire or the house is on fire — slow down.

And yet all of these companies say it’s so scary, and now they’re talking about slowdowns. But you ever notice that Amodei and Altman, they say, “Oh, maybe we should slow down progress.” And then they don’t. Right now, Altman’s saying, “Oh, we’ve slowed down progress because we’re so delayed. Now they’re out of compute.” But I can guarantee you, their PR people do not like me. I don’t think OpenAI’s PR people are super fond of me for that.

STEVEN BARTLETT: (01:12:43 – 01:12:48) But I bet there’s elements of what you’re saying, because you are calming people. You are theoretically calming down the general public.

ED ZITRON: (01:12:48 – 01:13:00) And you know what? I hope I am. Because the fear-based tactics are horrible. These companies don’t want that. These companies want people scared. I’m 100% sure. I just fundamentally disagree.

STEVEN BARTLETT: (01:13:00 – 01:13:36) So the timelines — I sit here and I log their quotes over time. And I read them out, from 2015 to 2026. And the change you see is them going from, “There could be extinction” — that’s the early narrative, Elon said it himself, “It’s the single most dangerous thing” — Elon says a lot of things — and then you track it over time and it evolves to this “age of abundance.” “We’re all going to have unlimited stuff.” And then the new slogan at ChatGPT is “Intelligence for everyone.” And whenever Dario comes out and says, “By the way, it’s really dangerous,” they attack Dario.

ED ZITRON: (01:13:36 – 01:13:37) Yeah, that’s true.

STEVEN BARTLETT: (01:13:37 – 01:13:38) They hate him.

ED ZITRON: (01:13:38 – 01:13:40) That man, Dario is —

STEVEN BARTLETT: (01:13:40 – 01:13:41) They’re like, “Dario, shut up.”

ED ZITRON: (01:13:41 – 01:14:22) Honestly, I’ve been saying “Dario, shut up” for years. But the thing is, I get your point where it’s like, I don’t think they’ve changed to calm the public down so much as they’re desperate to not get regulated. Which is laughable. We don’t regulate tech. We don’t regulate anything. America doesn’t regulate anything. We are still trapped in the hands of Milton Friedman, Margaret Thatcher, and Ronald Reagan. We’re still stuck in the neoliberal hellscape, which is growth at all costs, free market capitalism. So no, no one’s regulating. The regulation of these companies should have been — I don’t know — breaking up these companies for sure. We shouldn’t have companies this big. It makes things worse.

STEVEN BARTLETT: (01:14:22 – 01:14:23) But these technologies are dangerous.

ED ZITRON: (01:14:24 – 01:14:27) I mean, they’re dangerous, but not in the ways they’ve been warning about.

STEVEN BARTLETT: (01:14:27 – 01:14:30) Let’s think about cyber hacking, right?

ED ZITRON: (01:14:30 – 01:14:40) And just to be clear, those cyber hacking things that happened were not a result of — they were like, “Break out of this sandbox,” and then they set the sandbox up wrong. They set up the server they were on wrong.

STEVEN BARTLETT: (01:14:40 – 01:15:02) But advanced AI models could very easily, because they can go out onto the open internet as agents, go and look at code bases of different websites, find vulnerabilities, and exploit those vulnerabilities at scale, arguably at a higher intelligence and faster and wider than a human hacker could theoretically. So that’s dangerous.

ED ZITRON: (01:15:02 – 01:15:49) Well, here’s the funny thing. We don’t know how much compute was spent to do the Hugging Face attack, the OpenAI one. We also do know that they improperly set up the server. They thought they’d turned the internet off and they didn’t. That’s human error. And that’s human error in a sense that, yeah, they threw an indeterminately large amount of compute at it. This is dangerous.

But people keep saying, “We can’t let the Chinese get ahold of these models. What if these models fall into the wrong hands?” They’re already in the wrong hands — Mark Zuckerberg, Sam Altman, Dario Amodei. The wrong hands are the hands of those who are running these companies. We should not be training these models to do these things. I don’t know why we’re doing it, other than they’ve run out of other things they can train on.

STEVEN BARTLETT: (01:15:49 – 01:16:04) But would you agree that an intelligence — and I’ll call it that, you might disagree with that terminology — that can go out onto the internet and click around and take actions is inherently, there are risks associated with that?

ED ZITRON: (01:16:05 – 01:17:28) Well, the second part I agree with, the risks. We’ve had people running automated scripts and hacking scripts for a while. We’ve had hackers doing that for years and years. This is brute-forcing it with a bunch of compute. And yet it is dangerous. These companies are doing something dangerous. That is not what Geoffrey Hinton et al. have been warning about. They’ve been saying, “Oh, these things could destroy society, they could manipulate people.” When you actually look at the underlying things, not so much. Geoffrey Hinton, as well — talking in his book, still got his Google stock in there. I think weirdly enough he left Google because he was worried about the AI there, but then immediately made a comment being like, “Yeah, actually though, Google’s very responsible.” Strange thing, that.

But let’s get back to the cybersecurity side. I agree, this is dangerous. These people should not have access to so much compute. They clearly don’t know what to do with it. There’s a really easy way of dealing with this — it’s not letting them use so much compute, it’s regulating that part out of existence. “What if the Chinese do it?” The Chinese were able to distill the model. And also, regulate it and stop. I feel like with this particular thing, we got to this point and let the genie out of the bottle, to use an annoying Sam Altman term. We let this happen because we let these companies be unregulated and use as much compute as they want. We had these enablers allowing them to burn as much compute as they want. And also, for all of these dire warnings about AI danger, no one seems to have actually done anything.

Myths About the AI Industry

STEVEN BARTLETT: (01:17:28 – 01:17:29) Okay, we’re gonna play a game, Ed.

ED ZITRON: (01:17:29 – 01:17:30) Let’s play it.

STEVEN BARTLETT: (01:17:30 – 01:17:42) On these cards here, I have the things that you consider to be myths about the AI industry. The challenge is I want you to give me one sentence on each myth.

ED ZITRON: (01:17:42 – 01:17:43) Oh, Christ.

STEVEN BARTLETT: (01:17:43 – 01:17:50) So just your first reaction — you’re going to pick it up, you’re going to read it, and then you’re going to give me one sentence on your opinion of that belief.

ED ZITRON: (01:17:50 – 01:17:50) Okay, let’s go.

STEVEN BARTLETT: (01:17:51 – 01:17:58) Let’s do this. What does it say? And what’s your one sentence?

ED ZITRON: (01:17:58 – 01:18:04) It says, “The AI industry is creating enormous economic growth.” No, it’s not. It’s nowhere in the data.

STEVEN BARTLETT: (01:18:05 – 01:18:05) Okay.

ED ZITRON: (01:18:05 – 01:18:08) May I do a second sentence?

STEVEN BARTLETT: (01:18:08 – 01:18:08) Go ahead.

ED ZITRON: (01:18:09 – 01:18:17) Pretty much all of the economics is either Nvidia feeding money to its companies like CoreWeave, or these three companies feeding money to these ones to spend it with them.

STEVEN BARTLETT: (01:18:18 – 01:18:22) Okay, and what evidence do you have that it’s not causing economic growth?

ED ZITRON: (01:18:23 – 01:18:41) Just to be clear, other than the spend on semiconductors — so the speculative investment in GPUs and data center infrastructure, that’s happening. But as far as spend on AI goes, barely cracking 100 billion dollars, and most of that is just these two running their services and paying these three companies, Oracle, CoreWeave, and others.

STEVEN BARTLETT: (01:18:41 – 01:18:45) But 100 billion dollars is a lot of money for a relatively new technology.

ED ZITRON: (01:18:45 – 01:18:53) Not when you’ve spent 300 billion dollars in equity funding. And if we’re going with just these three, I think 600 billion dollars in capital expenditures.

STEVEN BARTLETT: (01:18:53 – 01:18:58) Yeah, I get that. That means it’s not profitable. But the 100 billion dollars is an expression of consumer demand.

ED ZITRON: (01:18:58 – 01:19:10) When the compute is mostly driven by subscriptions that subsidize — no, it’s not. When you’re giving someone 20 or 40 dollars for a dollar, they are going to use it more. If this was all on a per-million-token basis, we’d be having a different conversation.

STEVEN BARTLETT: (01:19:10 – 01:19:12) Fair. Fine. Cool. Next one.

ED ZITRON: (01:19:14 – 01:20:01) “The United States needs to spend trillions to beat China in the AI race.” What AI race? That’s actually my point. What AI race is there? Is it to make big scary LLMs? They did that already without the Nvidia GPUs. By the way, they’ve got Blackwell GPUs — Kasey Wang and Doug O’Laughlin, two amazing analysts, I love. They’ve been on this for years. China’s already had Nvidia GPUs that they’re not meant to have for years. But also, to do what? They already got the LLMs. What’s the race to do? To make us spend more money than them? For us to constantly worry about China? Because they won, if that’s the case.

Myth number three: “AI will replace all human jobs.” That just isn’t happening. There’s no economic data to support it.

STEVEN BARTLETT: (01:20:02 – 01:20:03) Will it replace some jobs?

ED ZITRON: (01:20:04 – 01:20:16) I mean, it’s replaced some contract labor that would otherwise be replaced with cheap labor out in the Global South. It’s a digital globalization in that sense. But all jobs, most jobs, a lot of jobs? No.

STEVEN BARTLETT: (01:20:16 – 01:20:17) What about robotics?

ED ZITRON: (01:20:17 – 01:20:20) Robotics is not what we’re talking about. Robotics is a very different thing.

STEVEN BARTLETT: (01:20:20 – 01:20:22) And even then, robotics will be powered by AI.

ED ZITRON: (01:20:23 – 01:20:31) I mean, yes, but there are tons of different kinds of AI. We’re talking explicitly about generative AI, and that’s — this is from my Mythbusters piece. That was definitely about generative AI.

STEVEN BARTLETT: (01:20:31 – 01:20:32) Okay, but what about robotics?

ED ZITRON: (01:20:32 – 01:20:53) The Optimus robot that Elon’s working on at Tesla — the one where even in the demo of the hand, they had to have a guy controlling it, it wasn’t doing it autonomously. Here’s the thing, if they can beat all these challenges, yeah, robotics would be really cool. I don’t know how long that’s — that’s one I’d actually be willing to believe in a couple decades.

STEVEN BARTLETT: (01:20:53 – 01:20:56) Have you seen the Chinese robots? I know you’ve seen the Chinese robots.

ED ZITRON: (01:20:56 – 01:21:00) Well, the Unitree, what’s it called? The one that can dance and that, but they can’t really do human things.

STEVEN BARTLETT: (01:21:01 – 01:21:03) It is pretty mind-blowing.

ED ZITRON: (01:21:04 – 01:21:16) Robotics are really cool. I’m not going to pretend I don’t think robots are cool. I wish they were building robots and actually doing cool stuff. I wish the tech industry still made fun stuff and interesting stuff. Instead we get these large language models.

Robotics, Agentic AI, and the Chief of Staff Test

STEVEN BARTLETT: (01:21:17 – 01:21:58) But AI plus robotics — I was in San Francisco and I went to this massive incubator there. And when I’d gone there three years earlier, it was all software startups. And when I went back three years later, it was all these robot startups. And I remember saying to the founder of the incubator, “Why is everything robots now?” There was this one robot where it was just the arm and it had a frying pan on it, and its whole thing is it cooks for you. So it was showing me it cooking, whatever. And he goes, “Well, you know the arm — the hardware part, the physical parts, that’s always been fairly cheap.” He goes, “The expensive part was the intelligence, and now that’s come down to pennies.” So what you’re seeing is this explosion in the robotics industry, because robotics is a function of intelligence plus hardware. We’ve always had the —

ED ZITRON: (01:21:58 – 01:22:00) And a ton of data, though, as well.

STEVEN BARTLETT: (01:22:00 – 01:22:00) Yeah.

ED ZITRON: (01:22:00 – 01:22:02) And the data is very expensive.

STEVEN BARTLETT: (01:22:02 – 01:22:03) Yeah.

ED ZITRON: (01:22:03 – 01:22:21) The thing is, cybercabs rolled out real slow. It’s going to take a long time. It could be a threat if they do a robot that could replace a human job. Sure it could. But human jobs are multifaceted. Human jobs change with environments. And also a lot of human jobs that you might think of, like, I don’t know, a dishwashing robot, for example —

STEVEN BARTLETT: (01:22:21 – 01:22:21) Yeah.

ED ZITRON: (01:22:22 – 01:22:34) Some guy at a restaurant isn’t paying 10 or 20 grand for a robot to replace the job that they’re already not paying enough for. The point is, yeah, it could, if you can replace the jobs. That is not what we’re talking about with this.

STEVEN BARTLETT: (01:22:34 – 01:23:11) Yeah, I just ask these questions not because I’m trying to argue — I actually am trying to form my own opinion on these things. And I do think, as it’s written, “AI will replace all human jobs” — obviously not, obviously that’s not true. But I’m trying to figure out if the truth is somewhere in the middle, that there’s a certain type of job which actually humans probably shouldn’t have ever been doing, really. If you think back through history, there was someone whose job was just to sit in an elevator and press the buttons. That’s an example of a job humans probably shouldn’t have been doing. And as technology gets more advanced, it takes on a lot of that automated, monotonous stuff.

ED ZITRON: (01:23:11 – 01:23:23) Right. The thing is, with this particular thing — I know that this is from a specific blog I wrote. I was explicitly talking about generative AI, though. I was basically talking about, when people say this, they are referring to that.

STEVEN BARTLETT: (01:23:23 – 01:23:25) So you’re not talking about agentic AI, which is —

ED ZITRON: (01:23:25 – 01:23:42) Agentic AI is LLMs. Agentic AI is just a fancy way of saying an LLM talking to another LLM with a harness on top. That is still LLMs. Agentic AI is one of the bigger lies they tell. It’s like, when you hear “agent,” you’re meant to think autonomous AI can do what you want. It’s still LLMs. It’s still LLMs talking to other LLMs to do LLMs.

STEVEN BARTLETT: (01:23:42 – 01:24:14) Taking screenshots and putting them in LLMs and stuff. Yeah. Okay. But I could make the case that I’m just thinking about my personal usage. I definitely use agents to do things that I would’ve previously asked people to do. It’s not to say that I still don’t hire — because we’re hiring like crazy. But in that particular function, I’m thinking about the chief-of-staff role. My chief of staff would’ve previously triaged all of my inboxes and put them somewhere and told me about them. Now my chief of staff is no longer doing that job.

ED ZITRON: (01:24:14 – 01:24:15) You still have a chief of staff, though.

STEVEN BARTLETT: (01:24:15 – 01:24:18) This is what I’m saying. They’re doing other things.

ED ZITRON: (01:24:18 – 01:24:24) But the thing is, again, what you were describing is fairly basic automation. I don’t know what the tasks are.

STEVEN BARTLETT: (01:24:24 – 01:24:25) Triage of emails.

ED ZITRON: (01:24:25 – 01:25:03) They didn’t spend a trillion dollars on triaging email. That’s the promise. If they’d spent 10 billion dollars and this was much smaller, and you’re like, “Cool software, yay” — a lot of the things that people are impressed with, like script stuff as well, it’s just LLMs doing Python. You should be impressed by Python code. Python’s incredible. You can scrape websites, you can download stuff. It’s awesome.

But the point I’m making is, none of this would be anywhere near as much of a problem if they didn’t ask for all of the attention, all of the money, and promise the world. It’s their promises that are the problem, and the journalists who went along with it, and the analysts and the Twitter people who went along with this, saying that this would change everything and replace everything, leaving the realm of reality.

STEVEN BARTLETT: (01:25:03 – 01:25:08) Is there any technological innovation through history that was really game-changing where that didn’t happen?

ED ZITRON: (01:25:10 – 01:25:13) I mean, the internet —

STEVEN BARTLETT: (01:25:14 – 01:25:15) People overpromised on that too.

ED ZITRON: (01:25:15 – 01:25:53) I mean, they overpromised on the businesses, but I’ve read through a great many pieces about the early internet. A lot of people were excited but hesitant. They were worried that there was not enough demand, but they were still like, “Oh yeah, this could have potential ramifications if it happened.” People were not super negative there. A lot of the skeptics were saying, “We’re worried about an overload of bad information.” Look at where we are — a lot of people were worried about the social consequences of everyone talking online, which they were correct about. With the economic things, they were specifically talking about, like, The Globe, which I think made hundreds of thousands of dollars and had a billion-dollar market cap.

STEVEN BARTLETT: (01:25:53 – 01:25:56) Yeah, there was a massive hype in the dot-com era.

ED ZITRON: (01:25:56 – 01:26:46) I read a lot of those stories. The hype was nowhere near — you didn’t have articles everywhere saying if you don’t get online, you’ll be left behind. You didn’t have professional consequences. Nikhil Suresh, mentioned his blog earlier — he described this thing, “AI Mania Is Eviscerating Global Decision-Making,” where he said that you have businesses where, if you don’t say that you’re more productive with AI — whether or not it’s true is irrelevant — you have professional consequences. You can get fired. There are people having to AI-wash their jobs by saying AI did it, otherwise their bosses who don’t do anything will get mad at them. This did not happen with the internet. It was not present. And part of the thing is, social media was not like it is today. The decentralization of media in general has caused this as well. And also the fact of day trading. There’s so many different things that are different.

STEVEN BARTLETT: (01:26:46 – 01:26:55) It’s crazy. I do think AI is different from the internet, in part if you just measured it on the speed of adoption — especially if we just think about generative AI.

ED ZITRON: (01:26:55 – 01:27:09) But the adoption of the internet required physical connections to your house. The adoption of generative AI involves having a web browser. It took a vast amount of effort to bring internet to people. Even with dial-up connections, it still required the distribution.

STEVEN BARTLETT: (01:27:09 – 01:27:21) And that’s why it was so slow and there was less hype than AI. I do agree that there’s way more hype. And we’re again going back to this point that we’re clustering AI into this big category of lots of different things.

ED ZITRON: (01:27:21 – 01:27:22) There’s generative AI.

STEVEN BARTLETT: (01:27:22 – 01:27:24) There’s generative AI. There’s real-world AI.

The “Era of the Business Idiot”

ED ZITRON: (01:27:24 – 01:28:06) But generative AI is explicitly what I’m talking about here. When bosses are saying, “You need to use AI,” they’re not saying, “I need you to go and buy a Unitree robot.” They’re saying, “Use LLMs.” And that’s the thing — I have this theory, the era of the business idiot, where we are ruled by people that don’t do work. Because nobody who actually does a bunch of work, who really is productive, is harassing someone who works for them for not being productive — they don’t have the time, they’re doing work. Someone who is sitting there with the ingratiation machine that’s telling them that every idea out of their messy little skull is amazing — yeah, they’re going, “Damn, this thing says I’m a genius. Why are you not using the genius machine to do more work?” And yeah, if you’re a boss that goes to lunch, leaves lunch, and sometimes reads your emails, LLMs are magic.

Sloppification and Commoditized Value

STEVEN BARTLETT: (01:28:07 – 01:29:39) One of the most compelling arguments I have for the overhype of AI: in a world where everybody has access to these tools, whatever the tools can do would largely be commoditized. And what the tools can’t do — which one could say is human taste, judgment, people skills, whatever you want to call it — is now going to be the valuable thing, because the scarce and the hard becomes the most valuable through history, and the commoditized becomes the least valuable. So the very nature that we’re commoditizing the generation of content, or code, means that’s actually not where the value will accrue for the user. And if an AI can do it, the “great” thing is not of value.

So I’ve been thinking a lot about how do you avoid the temptation of the sloppification of the things you make, the value you put into the world. A simple example people will relate to: if you use ChatGPT or Claude to make your LinkedIn posts, they will be poor LinkedIn posts, because everybody else is using them. And a great LinkedIn post now is someone who doesn’t use them and makes something that’s irreplaceably human and deeper and more personal, an “n of 1” lived experience — all the things AI can’t do. And I think that’s a compelling argument, that actually the commodity tools produce commodity outcomes.

ED ZITRON: (01:29:39 – 01:29:42) So everyone has access to these things and what’s changed? Like really, what?

STEVEN BARTLETT: (01:29:42 – 01:29:43) The sloppification has started.

ED ZITRON: (01:29:44 – 01:30:23) We’ve got a bunch of slop, but these people were half-assing their jobs before. It’s just a half-assery machine. It’s what I’m talking about with the slop blogs. People that gave you dog shit before have now got the dog shit machine to pump out dog shit.

So there’s a guy called Carl Brown — internet, awesome guy, great software engineer. He said — I might have said this earlier — it makes the easy things easy, the hard things harder. When you know you’re doing a really distinct small script for something and it can produce that, it’s awesome. I used Claude the other day for something useful. My kid loves Minecraft. I was trying to fix a broken mod because he loves his Wither Storm. It’s awesome. And it still took me half an hour and kept getting things wrong.

STEVEN BARTLETT: (01:30:23 – 01:30:25) What do you use AI for? Generative AI?

ED ZITRON: (01:30:25 – 01:30:33) I really don’t. You don’t use it with a Bloomberg Terminal. I use Ask B, which is just — it’s like requesting the consensus analyst estimates for Nvidia.

STEVEN BARTLETT: (01:30:33 – 01:30:35) But otherwise you don’t use it? So how do you know it’s bad?

ED ZITRON: (01:30:36 – 01:31:08) I’ve used it, I’ve put it through its paces. I’ve used it to try and do financial models and found an error, and immediately been like — I’ve never been particularly impressed. The one thing I will defend it on is it’s really good for tech support. I have this thing called Synergy in my New York place, where I have a monitor with a MacBook and a PC laptop and this thing Synergy for using the same mouse and keyboard. Dropping a giant troubleshooting log into it and going, “What’s wrong?” and it going, “This is wrong” — yeah, super useful. Is that worth a trillion dollars? No. Is that a 2-trillion-dollar company? No.

STEVEN BARTLETT: (01:31:08 – 01:31:10) Pretty easy. Better than Google, though, right? Better than Google Search.

ED ZITRON: (01:31:10 – 01:31:17) No, I mean — do you use Google Search still? I try. I have to push the crap out of the way and scroll.

STEVEN BARTLETT: (01:31:17 – 01:31:20) I can’t remember the last time I did a Google search.

ED ZITRON: (01:31:20 – 01:31:32) I find myself using Bing sometimes. I hate saying it too, but I have to scroll past the AI crap because I want the good stuff. I want the actual links to stuff so that I can read the thing and go —

STEVEN BARTLETT: (01:31:32 – 01:31:35) But you can ask the AI to give you the links.

ED ZITRON: (01:31:35 – 01:31:38) Yeah, and it doesn’t do a particularly good job.

STEVEN BARTLETT: (01:31:38 – 01:31:46) The other day my iPad wasn’t turning on and it was doing this funny little thing on the screen. Do you think it’s better to type that into Google than to —

ED ZITRON: (01:31:46 – 01:32:12) Oh no, I must be clear, that may be the only LLM use case I have. The troubleshooting thing is awesome for it. It’s the one weakness I have. It’s like genuinely being able to drop a log into it — that’s awesome. Again, that is not what they’re selling it as. They’re not selling it as a useful little tool. They’re selling it as the software, as the thing that will change everything, that will replace all jobs, that will do this and that. It’s not like they sold it as a quirky bit of software.

STEVEN BARTLETT: (01:32:12 – 01:32:18) No, you are right, they are telling us it’s going to replace everything. But funnily enough, the critics are saying that as well.

ED ZITRON: (01:32:18 – 01:32:21) Which ones? I mean, the Geoffrey Hintons of the world.

STEVEN BARTLETT: (01:32:21 – 01:32:31) You know, even people who have left the safety team at OpenAI, who’ve sat here with me — these are critics warning of the impacts it’s going to have on the world.

ED ZITRON: (01:32:31 – 01:32:44) It’s weird how all these critics also have a vested interest in AI doing well, though. Daniel — former OpenAI guy, AI 2027, written with the Scott Alexander guy — that was nothing more than badly written science fiction that he’s already had to walk back.

STEVEN BARTLETT: (01:32:44 – 01:32:45) He could have made more money by staying at OpenAI.

ED ZITRON: (01:32:47 – 01:32:48) Could he?

STEVEN BARTLETT: (01:32:48 – 01:32:53) I mean, looks like it. If he had options early, sticking around would’ve made him —

ED ZITRON: (01:32:53 – 01:32:54) Did he lose the options? How much did he lose?

STEVEN BARTLETT: (01:32:54 – 01:32:57) But you’re not saying they’re not being critical?

ED ZITRON: (01:32:57 – 01:33:41) They’re not critical of the companies themselves. They’re not critical of the stealing. They’re not critical of the environmental damage. And they’re not critical of the fact that you cannot rely on the answers. They’re critical of this big scary boogeyman out in the future, where it’s like, “I’m scared of when this becomes so powerful, and everyone should talk to me about how scary and powerful it is.” They’re not saying, “Hey, here are the harms today. Here are the things we’re actually looking at today. Here are the social problems of having this automated way of spewing out slop, of filling our feeds with crap, of having information that will pop up that is presented even with the little disclaimer thing” — the tiniest words possible — that says, “Sometimes this gets things wrong.” They don’t talk about the fact that these things are trained on stealing millions of people’s work.

Has AI Actually Gotten More Intelligent?

STEVEN BARTLETT: (01:33:41 – 01:33:57) But on that last point, where you say that it’s going to get progressively more intelligent and when it does, it’ll be a danger — would you agree with the statement that artificial intelligence has gotten more intelligent, if you measure it based on any sort of measure of intelligence one might use?

ED ZITRON: (01:33:57 – 01:34:02) It’s got better on the tests that are rigged for the models. It’s got better at tests where you can train for the test.

STEVEN BARTLETT: (01:34:03 – 01:34:03) Okay, so it’s got better at —

ED ZITRON: (01:34:04 – 01:34:07) It’s got better at tests that they’re intentionally trained for.

STEVEN BARTLETT: (01:34:07 – 01:34:17) So if you logged the rate of improvement on a graph, it would look something like this — you’d agree, in terms of what it’s capable of doing?

ED ZITRON: (01:34:17 – 01:34:29) There we go. Yeah, because it’s not — it’s not got new features. You’ll notice that outside of OpenAI and Anthropic, when you remove the coding startups, there’s basically no successful AI startup company.

STEVEN BARTLETT: (01:34:29 – 01:34:55) So we agree that it’s got better, it’s got more capable at doing things. Okay, fine. Over time, AI’s got more capable. If we imagine that trajectory will continue, it will get more capable. Then at some point, does it not cross — this is what they say to me — it crosses human intelligence. And at such time, will it not start to do some of the jobs that people are doing today?

ED ZITRON: (01:34:55 – 01:35:01) Outside of software engineering — remove software, because I will concede software engineering, it’s got better at that — outside of software engineering, where?

STEVEN BARTLETT: (01:35:02 – 01:35:03) So the chief-of-staff things that I mentioned.

ED ZITRON: (01:35:04 – 01:35:05) Okay, so it’s got better at that, man.

STEVEN BARTLETT: (01:35:05 – 01:35:14) Video generation, photo generation, text generation, theoretically coding. And then I’d say agentic workflows.

ED ZITRON: (01:35:14 – 01:35:15) What is an agentic workflow?

STEVEN BARTLETT: (01:35:15 – 01:35:21) So automated workflows where you’re doing the same — a good example is looking at the backend data of the Diary of a CEO.

ED ZITRON: (01:35:22 – 01:35:22) Summarizing.

STEVEN BARTLETT: (01:35:23 – 01:35:29) Looking at all of the data, ingesting all of it, going out onto the internet and searching who Ed is, looking at every interview you’ve ever done, ever.

ED ZITRON: (01:35:29 – 01:35:32) This is summarizing and generating.

STEVEN BARTLETT: (01:35:32 – 01:35:41) Making a little model on the things people want to know from Ed, producing a report, sending that to my inbox — me getting a 20-, 30-, 40-, 50-page report on Ed before he arrives.

ED ZITRON: (01:35:41 – 01:35:45) This is all basically the same thing it’s been doing for years, though. It’s not really new capabilities.

STEVEN BARTLETT: (01:35:45 – 01:35:46) Research.

ED ZITRON: (01:35:47 – 01:35:52) It’s still the same things. They’ve had web search for years. They’ve had report generation for years.

STEVEN BARTLETT: (01:35:52 – 01:36:21) But we couldn’t generate high-quality videos that are indistinguishable from cameras — SeedDance and these ones that look like movies are incredible. So the point I’m trying to make is that if we imagine that over the last 10 years, there has been a rate of improvement in terms of capabilities and output and quality — we’ve seen hallucinations drop, we’ve seen the models get more, quote unquote, intelligent, get better at, if you gave it an IQ test, it’s getting higher scores than it was 10 years ago. We agree that there’s been an upward motion of improvement.

ED ZITRON: (01:36:21 – 01:36:24) This is pretty much how machine learning goes when you feed it more data.

STEVEN BARTLETT: (01:36:24 – 01:36:34) Exactly. And you put more compute behind it. So if this continues, what does the future look like? The rebuttal I was expecting to hear is that it won’t continue.

ED ZITRON: (01:36:35 – 01:36:39) And actually, I don’t think it — I think that there are hard limits that we’re going to hit.

STEVEN BARTLETT: (01:36:39 – 01:36:41) Okay, so you do believe that there’s a hard limit somewhere?

ED ZITRON: (01:36:41 – 01:38:15) We’ve kind of already hit the diminishing-returns level, because, for example, video generation — which, by the way, is far less an American concern anymore, OpenAI shut down Sora, I think you can still use the API — but nevertheless, look around you at the amount of stuff and the crew you need to get a shot. People think that movies are just shot by shot and they magically happen. When you’ve got — first ADs, assistant directors, gaffers, lighters — and also simulating light is insanely difficult. There are so many magical things that happen in creating visual images that, yeah, you could create a one-minute-long thing that might fool someone. How do you practically turn that into a movie? There was a movie that claimed it aired at Cannes. It didn’t — it aired in the city of Cannes during the Cannes Film Festival, it was not at the film festival.

When it comes to the practical creation of actual things at the end of it, versus magic tricks, the actual practical outcomes are not there. The reason I keep coming back to the capabilities thing for the example is, yeah, they can do better at tests, the number goes up. When it comes to, can this actually do distinct tasks you can rely on — you can rely on it for summaries, you can rely on it for generations, the things it was doing, it’s getting linearly better at. But there’s a ceiling to that. Like, okay, so it gets really good at research — what does that actually mean? You’ve already kind of got the automation there. What is the next step of that? Because training it to be more autonomous, for example, that’s not something that comes from training data. That is actually a new — Gary Marcus called it neurosymbolic. You actually need to build a structure around the AI to make it work. And even then, it doesn’t fix the problem.

STEVEN BARTLETT: (01:38:16 – 01:38:22) So you’re saying that there will become a point where the rate of improvement will plateau. We’re already there and stop.

ED ZITRON: (01:38:22 – 01:38:25) We’ve already hit that dimension. Gary Marcus said this in 2022 as well.

STEVEN BARTLETT: (01:38:25 – 01:38:33) Do you know there are lots of people listening now who’ve had their workflows completely transformed by these tools?

ED ZITRON: (01:38:33 – 01:38:57) Have they? There’ll be people, yeah, there are. The thing is, first of all, every single one of them — did you pay for the tokens? That’s the thing. Did you pay for the tokens? And also, how many tokens did you burn? But putting all that aside, what workflows? Because if it’s, “Yeah, I did a bunch of web scraping or web search” — I’m just not impressed. Did you make an entire movie? No, you didn’t. Is it speeding up your coding? Yeah, I believe that, I’ve heard that from multiple people. But again, how much can you trust this?

STEVEN BARTLETT: (01:38:57 – 01:39:16) I think what I was getting at is, in the moment of any technological innovation, people extrapolate linearly, or they view it as a static state — they think today is going to look like tomorrow, or they think it’s going to get better in a straight line. But what we end up seeing a lot of the time is exponential improvement.

ED ZITRON: (01:39:17 – 01:40:38) All of the innovations we’re talking about, with compute and all that, with faster processors — those are hardware breakthroughs. The hardware breakthrough companies don’t seem to be fixing the LLM problems, despite all the king’s horses, all the king’s men. We’ve got nine, 10 generations of TPUs from Google now. Broadcom’s building stuff with OpenAI, their Jalapeño chip. And yet none of these people can just say, “Yeah, we’re on the path to making this profitable,” because they can’t. If we fix the environmental problems and the profitability situation, maybe I’d be more generous with this stuff, but they don’t seem to be able to.

And you talk about these improvements in capabilities — there’s a certain point at which I’m saying, okay, can it do even a tenth of the stuff they’re promising? Sam Altman, the other week, was saying it was going to be, in like six months, like a genie that you can ask wishes from. Motherf— never watched Aladdin? What’s he talking about? Long story short, the promises do not line up with the capabilities or the capability improvements. An exponential improvement in software and software performance is always a result of direct hardware improvement. We have all the gifted mathematicians, all the gifted software engineers, all the gifted hardware engineers, and where are we? Trillion-plus dollars in, with the future great financial crisis, and the world’s greatest marketing psyop.

STEVEN BARTLETT: (01:40:38 – 01:40:45) I just think in the future, I do think that all of the devices and the computers we use and the physical items in our world will be more intelligent.

ED ZITRON: (01:40:45 – 01:40:47) I mean, sure, but is that LLMs?

STEVEN BARTLETT: (01:40:48 – 01:40:54) And that’ll be powered by the underlying AI infrastructure. It’ll be the more data centers. It’ll be energy coming down.

ED ZITRON: (01:40:54 – 01:41:27) How does a GPU-full data center translate to a Nikon camera that can — I don’t even know what you’d think it could do — because what is the thing we’re talking about here? The idea that devices will get smarter, sure, I can see that. It’s already kind of happening. What does that have to do with the data centers? Because these data centers again are not being built to make your consumer electronics smarter. They’re not being built for anything other than speculating on the ability to capture demand for generative AI services.

STEVEN BARTLETT: (01:41:27 – 01:41:29) But it’s not just generative AI. We went through that.

ED ZITRON: (01:41:29 – 01:41:34) It is. No, but those data centers, they are being built for generative AI. They are not being built for anything else.

Meta’s Basis Points

STEVEN BARTLETT: (01:41:34 – 01:42:08) Would you consider generative AI to be the fact that on Meta’s earnings call, like a couple of weeks ago, Mark Zuckerberg said the big breakthrough we’ve had, which has resulted in 15 basis points of increased retention — I believe he was referring to Instagram — is that we now take anything you post on social media and we run it through an AI to get full context of what it is. Because we can see, say, a guy sat in front of me called Ed with a blue shirt and coffee. We now can train the AI to serve whoever wants “blue shirt Ed with coffee” to the right user, which means people are retained longer.

ED ZITRON: (01:42:09 – 01:42:11) Because isn’t 15 basis points like 0.15%?

STEVEN BARTLETT: (01:42:11 – 01:42:15) Yeah, it’s not much, but it makes a difference at scale. It makes a big difference at scale.

ED ZITRON: (01:42:15 – 01:42:32) Yeah, but 100-and-something billion dollars in, and the best you’ve got is 0.15%? If he could be fired — I mean, how much of a difference? Because there’s a reason he’s saying basis points versus dollars. Because think about it like this — if Mark Zuckerberg was — I take your point about scale.

STEVEN BARTLETT: (01:42:33 – 01:42:47) I’m saying the point I was making was that that is another application of these data centers, because it needs a data center that is driving revenues. But also that’s outside of us thinking about just generating incentives.

ED ZITRON: (01:42:47 – 01:43:37) And that’s their generative model — Muse. Muse Spark is their LLM. That’s them doing the weird thing where it’s like on Instagram and it’s like “Dave the Cat” — why is Dave the Cat suffering? It’s the weird pop-up things. Meta is messed up. That company sucks. Every time I think about how they’ve ruined that product. But that’s the thing though, again — why can’t he just say with his whole chest, “We’ve made a couple billion”? Why can’t he say that? Because he isn’t. Because there’s not actually a way of going, “I spent all this money, I spent 14 billion goddamn dollars on Gale, Alexander Wong, and I made this much.” They can’t. It gets back to a very simple point of — hey, if it was going well, you’d tell me how well it was going, rather than doing this weird rain dance thing where you’re like, “Well, if we move all the pieces around in three years, theoretically this will happen.”

The Metaverse, Crypto, and the “Rot-Com” Bubble

STEVEN BARTLETT: (01:44:16 – 01:45:24) I do think you’re accurate and right when you talk about the fact that there’s a lot of, like — is the word “fugazi”? Where there’s a lot of people that have spent a lot of money and they kind of shouldn’t have spent it, and they messed up. And now they’re thinking, “We’ve spent all this investor money.” Kind of like the metaverse was a bit of a joke.

ED ZITRON: (01:45:24 – 01:45:25) That’s so weird.

STEVEN BARTLETT: (01:45:25 – 01:45:30) A lot of money spent. We kind of thought this dream was coming — of this one — I shouldn’t say dream, because it’s not a dream I’ve had.

ED ZITRON: (01:45:31 – 01:45:32) A dream that they had.

STEVEN BARTLETT: (01:45:32 – 01:45:46) Yeah. This sort of virtual world. And actually it never transpired, and there’s no sign that it will in the near term. AI and the dot-com boom, in this regard, are the same. NFTs were the same. So crypto, one could argue that a lot of the crypto industry was the same.

ED ZITRON: (01:45:46 – 01:46:57) It’s fugazi that is inflated by the media. The difference is, the reason the metaverse and NFTs didn’t escape this was there weren’t stocks to speculate on. There weren’t big companies that you could invest in. They had record earnings in 2021. There’s a bunch of money floating in the system thanks to post-COVID — the PPP, basically government federal money floating into the banks. There’s a bunch of easy money, zero-interest-free era, money was easy to find. Then after that there was the hangover. Growth started to slow down dramatically.

This is actually my rot-com bubble theory: they don’t have any hypergrowth ideas anymore. So suddenly they started buying GPUs, and when they bought GPUs, people went, “They’re doing AI, oh, we better buy the stock.” And the stocks went on an incredible run. Meta’s like several hundred percent growth in the last few years. The stock has grown by hundreds of percent despite zero proof. And because the media was just saying, “Yeah, Meta’s revenue’s growing because of AI, right? Microsoft’s revenue’s growing because of AI, right?” The fugazi you’re talking about was the fact that everyone just gave them credit in advance. And now we’re kind of getting to the point where it’s like, “Hey, you didn’t spend that trillion dollars for no reason, did you, Satya?”

STEVEN BARTLETT: (01:46:57 – 01:47:00) I do think there’s overspending. I want to concede that.

ED ZITRON: (01:47:01 – 01:47:01) Dramatic.

STEVEN BARTLETT: (01:47:01 – 01:47:20) Yeah, no, I do think there is. And I think the reason why there’s overspending, Ed, is I think there is something here. In terms of, I think there is practical use for this technology. And I think when people realize that through history, they go crazy because they want to be the person that owns the opportunity.

ED ZITRON: (01:47:20 – 01:47:22) I’m going to be honest, I just fundamentally don’t agree.

STEVEN BARTLETT: (01:47:22 – 01:47:24) You don’t agree with which part?

ED ZITRON: (01:47:24 – 01:47:40) I don’t agree that the speculation is a result of actual demand. I don’t believe it. I don’t think private credit is sinking hundreds of billions of dollars into AI because of actual demand. They are doing it because they saw the biggest companies in the world building data centers, making a ton of money from two companies they feed money to, and went, “I want some of that money.”

STEVEN BARTLETT: (01:47:40 – 01:47:47) I’m saying that I do think there is value in the underlying technology. I’m not saying how much value.

ED ZITRON: (01:47:47 – 01:47:48) Right.

STEVEN BARTLETT: (01:47:48 – 01:47:48) Okay.

ED ZITRON: (01:47:49 – 01:47:50) I get your meaning. That’s fair.

STEVEN BARTLETT: (01:47:50 – 01:48:06) I’m not saying it’s proportionate to the investment. All I’m saying is, it’s like — if I take your example, the rot economy essay that you wrote — say that you’re on a desert island, and then someone says they found a banana tree. And there’s 10,000 people on the island.

ED ZITRON: (01:48:06 – 01:48:06) Okay.

STEVEN BARTLETT: (01:48:07 – 01:48:28) They are going to stampede towards where they think the banana tree is. They are going to claw each other to pieces. And if your essay here is right, that there was desperation because they hadn’t found an innovation in a while, maybe that explains it. Maybe there is a bit of value here. And they’re stampeding and killing each other and making irrational decisions like hungry people would.

ED ZITRON: (01:48:28 – 01:49:24) I actually think we agree. That is actually my point, which is these three companies and Meta, their main business lines are running out of growth. There’s only so much they can grow. And indeed, in the next 3.5 years, analysts think that these two, OpenAI and Anthropic, are going to spend over 400 billion dollars on these people alone — Microsoft, Google, and Amazon. And the crazy thing is, that’s a large part of their future growth. And if this money isn’t spent, their growth slows down.

So your point about bananas, I actually agree — that is the dot-com bubble. They don’t have a new thing and they’re desperate. And indeed they got rewarded for buying the GPUs. When they bought these GPUs from Nvidia, all the markets went wild overnight. They loved it. There were stories about how they were sending armored cars with the GPUs to Microsoft to make sure Microsoft got the GPUs. And so everyone saw all that money flowing in, even though they never disclosed AI revenues. They saw the expenditures and they went, “Well, I want to do what these people are doing. I want to get a little of that money, don’t I?”

Circular Financing and Historical Precedents

STEVEN BARTLETT: (01:49:25 – 01:49:34) I think the area where we have a slight disagreement is that I think the underlying technology has a lot more promise over the long term than you do.

ED ZITRON: (01:49:34 – 01:50:05) So the thing I want to push back on there is, to have progress with AI, just taking it in a vacuum — for these two companies to keep going and to keep progressing, they need to spend tens of billions of dollars a year on training. The only way that can happen is if these companies, and venture capitalists, and private credit firms, and individuals, and Nvidia, keep circulating money to them. So the progress we’ve got so far is entirely a result of this circular system.

STEVEN BARTLETT: (01:50:05 – 01:50:09) You talked about VCs there, venture capitalists, who are, by the way, investors.

ED ZITRON: (01:50:09 – 01:50:15) The majority of the funding that OpenAI got in the last six months came from SoftBank, Nvidia, and Amazon.

STEVEN BARTLETT: (01:50:16 – 01:50:16) Okay. Yeah.

ED ZITRON: (01:50:16 – 01:50:29) So the point is, you’re talking about progress continuing. Progress in LLMs can only continue as long as the money keeps flowing. Once the money stops flowing, the progress stops, which means —

STEVEN BARTLETT: (01:50:29 – 01:50:31) But isn’t that most, like — Spotify didn’t make money for 20 years.

ED ZITRON: (01:50:31 – 01:50:38) Spotify didn’t lose 20.9 billion dollars in one year. They didn’t need to raise 217 billion dollars in the space of six months.

STEVEN BARTLETT: (01:50:38 – 01:50:40) Yeah. And Uber’s another example.

ED ZITRON: (01:50:40 – 01:50:48) 33 billion dollars since inception before it became messily profitable. Amazon Web Services, between 2003 and 2015, when it became profitable — 29.7 billion dollars.

STEVEN BARTLETT: (01:50:48 – 01:50:49) They’d spent —

ED ZITRON: (01:50:49 – 01:50:56) Yeah, that’s the total capital expenditures. And that’s not just Amazon Web Services, that’s the entire logistics operation, normalized for inflation.

STEVEN BARTLETT: (01:50:56 – 01:50:59) So they all lost money for a long period of time — that’s the takeaway.

ED ZITRON: (01:50:59 – 01:51:09) Yes, but the amount of money they lost is completely on a different order of magnitude from these three companies.

STEVEN BARTLETT: (01:51:09 – 01:51:17) Can I argue then that that’s because the potential of intelligence permeates everything, whereas Amazon at the time was selling books?

ED ZITRON: (01:51:17 – 01:51:17) No.

STEVEN BARTLETT: (01:51:18 – 01:51:19) That was bringing retail online.

ED ZITRON: (01:51:19 – 01:51:21) When Amazon Web Services grew, it was —

STEVEN BARTLETT: (01:51:21 – 01:51:22) Oh, Amazon Web Services.

ED ZITRON: (01:51:22 – 01:51:23) Amazon Web Services.

STEVEN BARTLETT: (01:51:23 – 01:51:24) Cloud.

ED ZITRON: (01:51:24 – 01:52:01) With Amazon Web Services, the reason I bring that up — I’ll repeat something, but it’s really important — 2003, it was founded, mostly because Amazon, as a growing online store, needed hardcore infrastructure. 2006, I think, is when they turned it client-facing — I may be wrong on the dates — but 2015 was the year it became profitable. The total capital expenditures normalized for inflation were 29.7 billion dollars across that 12-year period. And yeah, it lost money. But if we speak cold economics here, Amazon’s margins actually started improving because AWS was a very margin-heavy business.

STEVEN BARTLETT: (01:52:01 – 01:52:02) It was great.

ED ZITRON: (01:52:03 – 01:52:31) These two — Google cash flow negative, Amazon cash flow negative — these businesses, the reason you liked software businesses was, they are meant to be cash-heavy, asset-light. These companies, along with Meta, have added more than 700 billion dollars of new property, plants, and equipment — so, assets, data centers, GPUs — in the last four years. They have gone from being these cash machines to these cash furnaces.

STEVEN BARTLETT: (01:52:31 – 01:52:35) You said a second ago, this can only continue if investors continue to invest.

ED ZITRON: (01:52:36 – 01:52:36) Yes.

STEVEN BARTLETT: (01:52:36 – 01:53:06) And I was saying, I think investors are used to pumping money into things that are burning cash. Your rebuttal to me sounds like, “Well, this is burning more cash than ever.” And then I would say, well, is the opportunity bigger than those other case studies you referenced, like AWS? And one would say that the opportunity of intelligence permeates everything. So the TAM, the total addressable market, is enormous. Maybe the rebuttal back to me is about open source and all these kinds of things.

ED ZITRON: (01:53:06 – 01:53:50) No, no, no, I actually know what you’re getting at. So what you were describing there is the argument that Satya Nadella or Sam Altman would make — the theoretical opportunity of large language models. And I could have bought into that from them when they were like, “We see the opportunity.” We’ve gone way past the point at which you can rationally argue that LLMs need this much money. And when I say the money needs to keep flowing, I am talking about these two companies, OpenAI in particular. Sam Altman has said — the Wall Street Journal, Anissa Gardizy reported a few weeks ago — they plan to spend 750 billion dollars on compute through 2030. I think they’re going to be dead before then, but 750 billion dollars — that is an insane amount of money.

STEVEN BARTLETT: (01:53:50 – 01:53:50) That is crazy.

ED ZITRON: (01:53:50 – 01:55:19) And a large chunk of that is training. So when I say progress, I mean literally to make the models better at stuff requires billions of dollars invested just in data, and also tens of billions of dollars taking that data. And training is actually a really interesting thing, because — like for Jake and Troy, my trainers, when I train with them, when I lift with them, I have a defined thing. And when I do it and I eat right, muscles get bigger, they would. And here’s the thing — when you train an LLM, you’re experimenting each time. And this is not actually a hit on the companies, because they’re still trying to work out how to do the thing.

Because putting aside how I feel like they’re trying to innovate — I think there are people at these companies that actually want to do something interesting — it’s costing too much money. So once the money tap turns off, the money won’t be there to buy the data or to feed the data into the GPUs. Put aside all the thoughts I have — just the raw capital to get them this far has cost increasingly larger amounts of money, and increasingly larger amounts of training money, for training runs that sometimes fail. GPT-5 was meant to be this panacea for the AI industry. They had at least one training run that cost half a billion dollars and did nothing. If we are thinking about progress in a vacuum, they need so much more money just to maybe get somewhere. There’s no guarantee. There’s never any guarantee. But there’s a reason that Google and Amazon are cash flow negative now. There’s a reason why Oracle’s probably going to die as a result of OpenAI, because Oracle’s future depends on OpenAI spending 300 billion dollars over five years.

What the CEOs Say in Their Own Defense

STEVEN BARTLETT: (01:55:19 – 01:55:26) It’s absolutely fascinating, because I was just reading through a list of quotes from the big CEOs of the AI companies to see what they would say in rebuttal to you.

ED ZITRON: (01:55:26 – 01:55:26) Yeah.

STEVEN BARTLETT: (01:55:27 – 01:56:13) And they’re all basically saying the same thing. This is an exact quote from Sundar, the CEO of Google. He says, “The risk of underinvesting is dramatically greater than the risk of overinvesting.” Then Andy Jassy, CEO of Amazon: “We’re not investing approximately 200 billion dollars in CapEx in 2026 on a hunch. We’re not going to be conservative in how we play this. We’re investing to be the meaningful leader, and our future business operating income and free cash flow will be much larger because of this investment.” Then Mark Zuckerberg, CEO of Meta, says, “We’ll continue to invest aggressively in infrastructure to meet the demand. I’d rather risk building capacity before it’s needed than being late.”

ED ZITRON: (01:56:13 – 01:56:30) Makes me think of Shrek, with Lord Farquaad — “Some of you may die, but that’s a risk I’m willing to accept.” It’s like, “I’m just going to spend all this money.” You can’t fire me because Mark Zuckerberg can’t be fired due to the unique reward situation he’s got going. So yeah, he’s just going to throw the money away and hope he’s right. And I know, from the people I know that matter, he’s not right.

STEVEN BARTLETT: (01:56:30 – 01:56:32) The thing is, why might you be wrong?

ED ZITRON: (01:56:33 – 01:56:45) I mean, this is the thing — the AI people who claim this is going to be the biggest, strongest thing in the world, did they ever get that? It’s a good question, because it’s like, they don’t. And the thing is, what would it take for me to be wrong? A bunch of hardware breakthroughs that make this profitable.

STEVEN BARTLETT: (01:56:46 – 01:56:47) A bunch of new mathematics.

ED ZITRON: (01:56:47 – 01:57:00) Because the thing is, when it comes to being a critic or a skeptic, you are put on the hot seat. Not the people spending a trillion dollars, not the people promising the world. The person, the asshole with a blog, is the one who’s like, “No, no, no.”

STEVEN BARTLETT: (01:57:00 – 01:57:02) Trust me, if they came here, they’d be on the hot seat too.

ED ZITRON: (01:57:03 – 01:57:08) Oh, they won’t talk to me. Don’t know why, Steve. It’s because I call him “Clammy Sammy.”

STEVEN BARTLETT: (01:57:08 – 01:57:13) I think it’s because my guests are quite critical that I don’t think Sam Altman wants to come here.

ED ZITRON: (01:57:13 – 01:57:28) Mr. Altman? Go on the show, do it. But this is the thing — of course they’re going to say that. And also, if they thought they were right, I don’t think they do anymore. If I was in their shoes and I thought this was an existential thing, sure. But it gets back to the rot-com bubble, which is, this is the last thing they’ve got.

STEVEN BARTLETT: (01:57:28 – 01:57:44) But I really want to know that question — it was one of the questions I was really excited to ask you — which is, you have a different opinion. We said this at the top. You have a very different opinion from a lot of people. I would categorize the two most popular opinions as: AI is going to hurt everybody and it’s going to be catastrophic and we need to stop.

ED ZITRON: (01:57:44 – 01:57:44) Yeah.

STEVEN BARTLETT: (01:57:44 – 01:57:57) The other opinion is, age of abundance, it’s going to be amazing, let us crack on. Yours is different from both of those, which is, as you said in your words, it’s a con, and there’s no real underlying value in the technology, and it’s overhyped.

ED ZITRON: (01:57:57 – 01:57:58) Yes.

STEVEN BARTLETT: (01:57:58 – 01:58:14) And there’s way too much spending. A few people agree on the spending part, but the other part — you’re probably the first person I’ve spoken to who’s had this opinion. So what would it take for you to change your mind about what you believe here?

ED ZITRON: (01:58:14 – 01:58:24) There would need to be a hardware breakthrough that reduced the cost by like 1,000x — it would have to be just a dramatic breakthrough, that is not happening, just to be clear, because they’ve all been trying.

STEVEN BARTLETT: (01:58:24 – 01:58:26) So it’s the cost for you that would have to change?

Data Centers, Communities, and Access to Capital

ED ZITRON: (01:58:26 – 01:59:12) It’s the cost, and it’s also the data centers. I think the way they’re building the data centers is reckless and damaging to communities. The fact that you have communities like in Vineland, New Jersey, where the residents don’t want this, but the planning boards vote for it — they’re all, I assume, having chummy lunches with the people doing it. I think the use of gas turbines is disgraceful. The water situation I’m not super well read on, so I’m not going to wade into it. But the use of gas turbines and behind-the-meter power is reckless and damaging to communities. The noise these things make. And also, generative AI is this egregious, pornographic demonstration of how unfair the world is. Regular people try and get a loan for a business, a random business they want to have a good idea — they go to a bank, and the bank tells them to go fuck themselves. “I’m not going to get — you’re going to make a store that sells stuff?”

STEVEN BARTLETT: (01:59:12 – 01:59:13) Screw you.

ED ZITRON: (01:59:13 – 02:00:20) You want to build a data center? Jensen Huang will back you. Jensen Huang will give you 25% residual value. You want to build a regular business that’s even profitable? No, a venture capitalist won’t give you the money. Something that’s just growing steadily but it’s profitable? No, I need a 10x, 100x return. Try and get a mortgage — you have to give the bank your full life story. But if you want money to buy some GPUs from Jensen Huang, he’ll give you a contract.

CoreWeave is a great example. A neocloud, which is just a company that builds data centers, puts GPUs in them, and rents them to people. Nvidia, one of their first investors in 2023, signed a 1.3-billion-dollar contract to rent back their GPUs from CoreWeave, so that CoreWeave could go to a bank and say, “I’ve got a customer” — yeah, the guy I’m buying the GPUs from, with the debt I’m getting from you. If you want to buy GPUs, it’s open season. If you want to live a regular life where you build a regular business or buy a house — highest interest rates ever, “I don’t trust you regular folks.” But if you’re an unprofitable neocloud, you get billions from Jensen. It doesn’t matter.

STEVEN BARTLETT: (02:00:21 – 02:00:26) It’s so interesting. You’re the first person I’ve spoken to that has that opinion.

ED ZITRON: (02:00:26 – 02:00:44) I am pro-user. Let’s check another myth. “AI will be conscious.” So superintelligence, artificial general intelligence — these are theories. Anyone saying this stuff will become this is just guessing and does not have proof.

STEVEN BARTLETT: (02:00:44 – 02:00:44) Okay.

ED ZITRON: (02:00:45 – 02:00:46) And that’s really it.

STEVEN BARTLETT: (02:00:46 – 02:00:46) Okay.

The “AI Is Blackmailing People” Myth

ED ZITRON: (02:00:46 – 02:02:24) Okay, let’s take another myth. “AI systems are already blackmailing and escaping control.” So this is a really specific one — there’s actually two. OpenAI’s GPT — in their system card, a bunch of media outlets covered this, saying that OpenAI’s model blackmailed a TaskRabbit into solving a CAPTCHA. What actually happened was, a user of GPT doing the experiment got it to generate things to say to a TaskRabbit worker to make them do stuff. A TaskRabbit, as in a person that you rent — not even to do a CAPTCHA, it’s someone you rent to, like, nail a picture up in your apartment. It’s an insane example. This was covered as if these things blackmailed someone. And they specifically said, “Yeah, we prompted it to do this.” And also the other note was that AI systems can’t do autonomous stuff like this.

Then there was this other one where Anthropic said, “Oh yeah, a model was blackmailing someone,” saying, “If you don’t do this, I’ll email proof that you slept with someone other than your wife.” What actually happened was, Anthropic explicitly trained a model to do this and then prompted it to blackmail. This keeps happening and the media just laps it up — “put the story in the bag.” And it’s frustrating because it scares people. Put aside the fact it’s wrong — it’s scary to people, people living their lives who have to work longer hours to make less money, and their money doesn’t go far, and they turn on the news and there’s some asshole saying, “Yeah, you should be terrified, it blackmailed someone.”

STEVEN BARTLETT: (02:02:25 – 02:02:31) But this is so counterintuitive to their interest, to some degree, and they’ve experienced it backfire.

ED ZITRON: (02:02:31 – 02:02:33) Well, they have now. It’s literally backfired.

STEVEN BARTLETT: (02:02:33 – 02:02:44) It’s backfired — Eric Schmidt getting booed at the commencement speech by 8,000 people every time he said the word “AI.” These CEOs are being attacked at home.

ED ZITRON: (02:02:44 – 02:02:46) Yeah, which sucks.

STEVEN BARTLETT: (02:02:46 – 02:02:46) Yeah, which is terrible.

ED ZITRON: (02:02:46 – 02:02:49) I must be clear, like — you dislike the company, don’t hurt people.

STEVEN BARTLETT: (02:02:49 – 02:02:56) Yeah, don’t attack people at home. But the point here is that narrative is backfiring in a big way for them.

ED ZITRON: (02:02:56 – 02:04:10) I agree. I don’t think they saw it coming, because you have to remember — you mentioned regulation earlier — these tech companies have been glazed for their entire existence. Travis Kalanick’s like, “What, people don’t like me now?” And it’s because Uber was a horribly run place, and he was kind of a monster. Also tons of articles about how great Uber was at the time.

The point I’m making is, these companies are not used to pushback. They thought what would happen, I believe — just guessing — they thought they’d do this scary stuff and they would just get floods of money, and everyone would just be like, “I kneel before you, I’ll do whatever you want.” They didn’t expect this, I think, to have backfired on them, because they were inarticulate. They’re disconnected from regular people. Sam Altman drives a 5-million-dollar car around San Francisco doing it like nine miles an hour. It’s hilarious. But these people are disconnected from everyone else, so they don’t experience real problems, so they can’t build the solutions for them. And they think, “Well, if we scare people into doing what we want, that’ll work, right?” It didn’t. All of this blackmail stuff was an attempt to make it mystic. It was a mysticism attempt. It was to make it seem like this unknowable, impossible-to-control, powerful thing — but “we’re the only ones, only these two angels could possibly control the beast we’ve created.”

STEVEN BARTLETT: (02:04:10 – 02:04:26) This is quite a controversial statement, but I think that for some reason I trust Dario a little bit more, because I think he’s been the most balanced in his writing about the risk profile. Whereas the others, they seem to kind of move with the wind.

ED ZITRON: (02:04:27 – 02:04:45) I get what you mean. The reason I don’t like Dario is, Dario was doing the scare-tactics thing when he worked at OpenAI. When GPT-2 came out, he said it’s too scary to release. He’s also gone on television and given AI psychosis to Axios, being like, “50% of jobs are going to go away because of AI.”

STEVEN BARTLETT: (02:04:45 – 02:05:03) What I respect is the consistency. He’s now being attacked by them. Silicon Valley is attacking Dario verbally, and if powerful people in Silicon Valley are attacking someone —

ED ZITRON: (02:05:03 – 02:06:12) Four months ago, he wasn’t, though. They were all saying he was the smartest guy ever. But the point I want to make there as well is, “Wow, you’re so scared of how powerful this is, it’s so scary — what are you doing about it?” “Oh, nothing.” “Well, we have an alignment team.” So does every AI lab — well, I guess OpenAI cycles through those really quickly.

Here’s the thing — if I’m Dario Amodei, and I’m sitting there going, “I’m scared of all things changing, and I thought I had made a thing that would eliminate all jobs,” I’d be terrified. I’d be walking around like there’s a 10-ton weight on my back — the responsibility. The fact he doesn’t, the fact he wants to be this weird elder statesman who’s too scared to hold Sam Altman’s hand at an event, just makes me believe that he’s just saying it because it’s convenient, and he’ll wind it back, as he kind of already has, whenever it’s convenient for him.

I think OpenAI and Anthropic are basically the same level of bad company. I think Anthropic is more cult-like. It’s so weird, like Jack Clark over there, one of the co-founders — that fella used to be at The Register, one of the most critical journalists ever. Now it’s like something took over him, because they talk of these things in these highfalutin terms. But then again, maybe the people at Anthropic buy their own shit. Maybe some of the people at OpenAI buy their own shit.

The Future-Tense Trap

STEVEN BARTLETT: (02:06:12 – 02:06:26) So going back to the central question we asked at the top: what would have to be the case for you to look back and say, “I was wrong in 2026”? And you said to me it would mainly be that the cost of production around AI drops dramatically.

ED ZITRON: (02:06:26 – 02:06:32) And it would have to also do insane amounts of stuff. It would have to be a truly autonomous —

STEVEN BARTLETT: (02:06:32 – 02:06:34) It would have to continue its improvement in terms of capability.

ED ZITRON: (02:06:34 – 02:06:40) It would have to be a different product. It would have to be indistinguishable from magic. And the reason I have these high standards is, they set them.

STEVEN BARTLETT: (02:06:41 – 02:06:41) Okay, fair enough.

ED ZITRON: (02:06:41 – 02:07:12) It’s interesting as well, because all these myths and all these conversations — it’s about technology, but it’s also an information war. It’s literally narrative versus narrative. Everyone trying to escape the financials, everyone trying to escape what the models can actually do. And the big thing I always say about AI boosters is, if I could regulate them, I’d regulate them so they can’t speak in the future tense anymore. You’ve got to talk about today, mate. You get two weeks in the future, max. Because if they were constrained to what was happening today, they would sound like insane people.

STEVEN BARTLETT: (02:07:13 – 02:07:18) Yeah, no, I think most technology companies would at the time — like Uber would sound insane. Amazon would sound insane.

ED ZITRON: (02:07:18 – 02:07:20) Uber was basically — the difference —

STEVEN BARTLETT: (02:07:20 – 02:07:21) They were pissing money away, though, weren’t they?

ED ZITRON: (02:07:21 – 02:07:39) They were pissing money away, but the unit economics were the same, just subsidized. So you were still getting a service from A to B and paying a much lower cost. It wasn’t like you paid Uber 20 dollars a month and got 500 miles of Uber, and then one day you started paying by the mile — because that’s what’s happening with this.

STEVEN BARTLETT: (02:07:39 – 02:07:45) Have they changed their business model for customers like me now, so that I have to buy credits?

ED ZITRON: (02:07:45 – 02:07:47) No. Well, kind of with Fable.

STEVEN BARTLETT: (02:07:47 – 02:07:48) They asked me the other day. With Fable.

ED ZITRON: (02:07:48 – 02:08:03) So with Anthropic’s Fable model, with some accounts you have to pay for usage. And also, adoption of Fable has been pretty low because of this, because of the cost. But with enterprises — companies over 150 people — you have to pay by the token now, per million tokens.

STEVEN BARTLETT: (02:08:03 – 02:08:04) Oh, so they are moving to a token model.

ED ZITRON: (02:08:04 – 02:08:25) Yeah, but when they did that, everyone went from being like, “This is the most impressive thing ever,” to being like, “Oh, we’ve got to control these costs.” Uber’s COO, Andrew Macdonald, said it’s getting hard to justify — because it’s hard to connect spending money on tokens to actual useful outcomes. He said the actual thing I’ve been saying.

STEVEN BARTLETT: (02:08:25 – 02:08:27) So we’re in an AI bubble.

ED ZITRON: (02:08:27 – 02:08:27) Yes.

Is the Bubble About to Pop?

STEVEN BARTLETT: (02:08:27 – 02:08:32) And when this AI bubble collapses, so much of the economy is resting upon it.

ED ZITRON: (02:08:33 – 02:08:33) Yeah.

STEVEN BARTLETT: (02:08:34 – 02:08:40) It’s going to have downstream consequences. So I’ve got two questions for you. Are we in an AI bubble? And what happens when the bubble pops?

ED ZITRON: (02:08:41 – 02:09:18) Yes. And it depends. The big thing that people say is, “Oh, we’ll get bailed out.” Here’s the problem with this. It isn’t just an AI bubble, it’s the rot-com bubble. So the AI bubble collapsing will probably be this company running out of money — OpenAI. And the thing is with OpenAI is, they were meant to go public this year, and now it’s been pushed to next year, a week and a half after I released their audited financials. Wonder why that was. But they’ve delayed to next year. Sarah Friar, the CFO, has now said, well, they’ll do it earlier than 2027, or 2027. Great answer there.

STEVEN BARTLETT: (02:09:18 – 02:09:46) For anyone that doesn’t understand what going public means — that means joining the stock market. And at such a time, when you join the stock market, your investors can finally sell their equity that they got for investing in the company when it was private. So oftentimes, companies will flirt with the idea of, “We’ll go public someday soon,” because investors will have a moment in their head where they’ll get their money back. So you kind of need to, if you’re in these guys’ shoes, be flirting with going public, or investors won’t want to invest.

ED ZITRON: (02:09:47 – 02:11:12) OpenAI, up until this point, has been a private company, and their last funding round they were valued at 865 billion dollars. Now, when they tried to go public — New York Times’ Mike Isaac reported this — they wanted to go at a 1-trillion-dollar valuation. Apparently their advisors said, “No, don’t do that. That is very bad for a number of reasons.” One, OpenAI needs perpetual amounts of money. They raised 122 billion dollars this year. Most of it’s spent. There’s some left, but they are going to need to raise at least 100 billion dollars a year just to survive.

If they can’t go public, they will have to raise another funding round. The problem is, it’s going to be difficult to raise even at the same valuation they raised at, so they’re probably going to have to take a flat round. But they need money. They need money so bad. Amazon sent them 35 billion dollars that was meant to be contingent on them going public early. They did that because they need the money.

Now, OpenAI is the kind of catastrophe center here, because Anthropic is likely going to beat it to go public. And once Anthropic goes public, it’ll be borderline impossible for OpenAI to do so, because Anthropic — an unprofitable, unsustainable AI lab, but a better business that’s growing faster than OpenAI — I believe they have a ceiling. They’re eventually going to face their own reckoning too. I think sometime in 2027, things are going to start running out of steam, because the only way these models get better is if you feed more money, tens of billions of dollars, into them.

STEVEN BARTLETT: (02:11:12 – 02:11:15) So you think OpenAI runs out of steam in 2027?

ED ZITRON: (02:11:15 – 02:11:18) I think they’re already running out of steam. But I think they run out of cash.

STEVEN BARTLETT: (02:11:18 – 02:11:19) You think they run out of cash?

ED ZITRON: (02:11:19 – 02:11:19) Yes.

STEVEN BARTLETT: (02:11:20 – 02:11:25) And the sequence of events here will be, they go out and try to raise, and they have trouble raising another round?

ED ZITRON: (02:11:25 – 02:11:38) I think maybe Nvidia props them up a little, maybe private credit, Blackstone, BlackRock, and the like — the ones — and the reason that private credit is getting involved, so asset managers, is because they’re investing in the data centers, and they know this company is most of the data center demand.

STEVEN BARTLETT: (02:11:38 – 02:11:40) Okay, so they run out of steam in 2027, according to you?

ED ZITRON: (02:11:41 – 02:12:47) Yep. And maybe if they bum-rush to go public, they’re going to have worse economics than Anthropic. They’re going to get savaged. WeWork was a great example, another SoftBank classic. Now, I think OpenAI collapses — there are many different ways it could happen, many different ways it could end. But the crucial thing is, there are multiple companies that are existentially tied to OpenAI. SoftBank, one of the largest companies in the Japanese stock market, a holding company with lots of investments — they have, on paper, about 100 billion dollars’ worth of OpenAI stock. If OpenAI can’t go public, SoftBank can’t do anything with that.

And so SoftBank’s future, their ability to continue paying the people around them and existing as a business, relies on their ability to continually liquidate funds — to take the things they’ve invested in and realize value from them, either by selling the stock or taking loans out on the stock. If OpenAI can’t go public, SoftBank can’t do that. SoftBank probably won’t run out of money, but we’re going to see one of the largest holding companies in the world become much smaller. We will also see Amazon, Google, and Microsoft have to restate guidance. They will have to say, “Actually, we don’t think we’re going to grow as fast.”

STEVEN BARTLETT: (02:12:47 – 02:12:48) And what happens then?

ED ZITRON: (02:12:49 – 02:13:22) Well, I think we enter a tech depression, because the rot-com bubble, the core of my theory, is that they’re out of hypergrowth ideas, but the market doesn’t think so. The reason they’re so maniacally spending is because buying AI GPUs allows them to kick the can further. It allows them to say, “We’re still doing something, we’re working on AI, don’t think too hard.” And also, their current businesses are still growing. Their current businesses will eventually slow. There’s only so many price increases, there’s only so many tweaks to ads, only so many tweaks to Google Search, only so many ways that Amazon can squeeze merchants.

STEVEN BARTLETT: (02:13:23 – 02:13:39) So in that tech depression, which you think might be triggered in 2027 — is that a cascading downstream economic depression? Because the stock market is heavily dependent on these companies. The stock market sees a pullback, investors stop investing, they get panicked.

ED ZITRON: (02:13:40 – 02:13:41) Yes, I think that because —

STEVEN BARTLETT: (02:13:41 – 02:13:44) What’s the sort of downstream consequence, the domino effect?

ED ZITRON: (02:13:44 – 02:14:14) There’s so much to imagine that it’s difficult to capture everything, but there are a few things that worry me. First of all, a ton of American money, just regular people’s money, retail investors are in these companies, and they bought into the Magnificent 7 thinking number goes up forever. Nvidia is the largest company on the Fortune 500 and NASDAQ as well, and like 7 to 8% of the S&P 500. When the bottom falls out from Nvidia — and we haven’t really got into it, but Nvidia is doing the most circular of financing, feeding companies money so that they can raise debt to buy more GPUs.

STEVEN BARTLETT: (02:14:14 – 02:14:14) GPUs.

ED ZITRON: (02:14:15 – 02:14:23) I think Nvidia’s revenue could go 50 to 70% down. Nvidia back in 2022 was making single-digit billions of dollars.

STEVEN BARTLETT: (02:14:23 – 02:14:30) And what happens though? I’m thinking about people like Jenny and Dave who are watching this right now, and they are just normal people with normal jobs.

ED ZITRON: (02:14:30 – 02:15:29) People’s retirements are going to contract severely, and I don’t believe they’re going to return to those values. And I think that, because so much of the value of the S&P 500 and Russell 1000 index comes from these four companies and the rest of the Magnificent 7 — Apple, Tesla, Meta as well. And I don’t know what happens after that, because venture capital has also — more than half of venture capital last year went into AI. I think most venture capital investments in AI are going to zero, because when it comes to building a company on top of an LLM, all of those are unprofitable too.

And the thing is, LLM companies have not really been acquired. The exception being Cursor, bought by Elon Musk for the coding side. But you have Cognition, which is just another LLM company raising a 26-billion-dollar valuation. That means that company has to go public, because who’s buying a company at 26 billion dollars other than Elon Musk? And there are rumors that Elon Musk was trying to buy them as well. Is Elon Musk just going to pick off every LLM company like it’s a discount rack?

STEVEN BARTLETT: (02:15:29 – 02:15:31) So is that a recession you’re describing?

ED ZITRON: (02:15:31 – 02:15:40) It is a recession, but it’s also a depression within people’s retirements. I’m talking about 20, 30, 40% off the top of these companies’ stock value.

STEVEN BARTLETT: (02:15:40 – 02:16:05) Economic contractions, recessions, consistently lead to job losses and rising unemployment. When an economy contracts, the mechanism driving job losses typically follows a predictable sequence. Falling demand: consumers and businesses spend less money, causing revenues across most industries to drop. Margin compression: with lower revenue and often fixed overhead costs like rent or debt, corporate profits shrink. And lastly, cost-cutting measures: to survive or protect profit margins, businesses freeze hiring, reduce hours, and resort to layoffs.

ED ZITRON: (02:16:06 – 02:16:34) Yes, that would all happen. But the thing is, we’re talking about equity values dropping, and there not really being a home for that value or that money. So much is riding on these companies, but you can’t bail it out. You could theoretically bail out OpenAI — I don’t think it happens. You could pump these companies full of money and keep them alive for a bit, but at some point they’re going to have to start. Between these two companies, Anthropic and OpenAI, you have 1.1 trillion dollars of commitments.

STEVEN BARTLETT: (02:16:34 – 02:16:35) Mm-hmm.

ED ZITRON: (02:16:35 – 02:16:36) Just OpenAI.

STEVEN BARTLETT: (02:16:36 – 02:16:36) Mm-hmm.

ED ZITRON: (02:16:36 – 02:16:50) Oracle is building 7.1 gigawatts of data centers, so over 400 billion dollars’ worth, just for OpenAI. There is not a customer on Earth — Oracle’s revenue has been flat the last 15 years when you adjust for inflation. So now Oracle dies.

STEVEN BARTLETT: (02:16:50 – 02:16:59) So you think OpenAI is going to crash and run out of money, and that’s going to cause this domino effect across these other big tech companies, which is going to impact the stock market and impact the broader economy?

ED ZITRON: (02:16:59 – 02:17:24) Yes, and also the tens of thousands of people that will be laid off from the tech sector. But also the venture capital thing is significant, because venture capital is having one of the worst runs in history. Since 2018, the average return from venture capital — total value put in, so the amount of money you get back for your dollar — is between 0.8 and 1.21, meaning for every dollar you invest, you get 80 cents to 1.20 dollars back.

STEVEN BARTLETT: (02:17:24 – 02:17:25) Paper gains?

ED ZITRON: (02:17:25 – 02:17:40) Well, no, that’s just actual returns. Paper gains they’ll give you, but even then the internal rate of return, which is a whole separate thing, even that’s not very happy. But long story short, venture capital is not making money come out. Venture capital is not actually providing returns.

STEVEN BARTLETT: (02:17:40 – 02:17:42) They’re celebrating paper gains.

ED ZITRON: (02:17:42 – 02:17:43) They’re celebrating paper gains.

STEVEN BARTLETT: (02:17:43 – 02:17:49) And they’re raising off paper gains. Paper gains, I mean, just being able to say, “Look, the valuation of Anthropic went up.”

ED ZITRON: (02:17:49 – 02:19:05) So that’s what Google and Amazon were doing. Google’s last quarter, they boosted their net profits on paper by 99 billion dollars because of the increased value of their SpaceX holding and their Anthropic holding. And the fact this is happening is insane, and the fact it’s not a scandal is insane. But we live in this culture, I guess.

Everyone is really benefiting right now. It’s that great tweet where it’s like, when you’re reaping, it’s like, “Yeah, this rocks.” Sowing — “Ah shit, this sucks.” Because right now they’re all like, “Yeah, all the speculative gains are awesome, the paper gains are awesome, the theoreticals of Anthropic being worth 2 trillion dollars — wow, the articles we can write, the promises we can make.” Then when the rubber meets the road, it’s going to be pretty rough on them.

Because the valuation of Amazon, Google, Microsoft, and Meta is based on this idea that they will grow eternally, that they will grow forever. If that changes — to quote Ed Elson from Prof G Markets again — it’s this: they’re all doing Botox right now. They’re sinking money into it to make themselves feel young again, and the market believes them. When the market doesn’t, we’re not just talking about a depression. I’m talking about the market valuing them like — you’re real big and you make money off your existing products, but guess what, you don’t have new stuff, you’re just going to be doing this forever, and we’re going to value you as such.

STEVEN BARTLETT: (02:19:05 – 02:19:13) So for Jenny and Dave, should they do anything differently? Should they be conserving money if there’s a recession or depression coming? Should they be a little bit more conservative?

ED ZITRON: (02:19:14 – 02:19:21) Yes, I actually think — I don’t know, I don’t have money in the market. I think it’s a casino pumped up by the media.

STEVEN BARTLETT: (02:19:21 – 02:19:24) Should they invest in the S&P 500? Should they invest in OpenAI?

ED ZITRON: (02:19:24 – 02:19:27) Oh God, no. Honestly, I live in cash right now.

STEVEN BARTLETT: (02:19:27 – 02:19:28) You live in cash?

ED ZITRON: (02:19:28 – 02:19:37) Yeah. I don’t trust the market, man. I’m not comfortable giving financial advice, but it’s like you’re gambling.

STEVEN BARTLETT: (02:19:38 – 02:19:39) Okay, be conservative. Things might get volatile.

ED ZITRON: (02:19:40 – 02:20:36) Yeah, it really is. Act as you would with volatility. Take the gains when you’ve got them. Don’t sell everything. Be suspicious of tech — that’s actually the biggest thing. Be suspicious of what they’re promising. If you are acting based on their promises, don’t trust the promises. Trust that they are going to say what will make the stock run, rather than what’s actually happening, and that they will find every dodgy way to make you think something is happening rather than it actually happening.

Annualized run rate — great example. Microsoft said that they had 37 or 38 billion dollars of annualized run rate in AI. You hear that, you go, “They made 37, 38 billion dollars, right? Wow, that’s so much.” Run rate — maybe month times 12. They don’t even define it, but it’s built to manipulate. And they do that because we don’t have a functional SEC, and we don’t have a media environment where skepticism is the priority and where protecting the readers is necessary.

STEVEN BARTLETT: (02:20:36 – 02:21:19) What would they say? They would say, “Ed, this technology’s going to be so great and so transformative that we are investing a ton of money in advance of the value and utility showing up.” That’s what they would say. And I’ve heard your rebuttal, but I just wanted to express that — I think that’s their sentiment. I’m not defending them, I’m trying to provide enough balance to see if we can dance between these two perspectives. And a lot of people would say that there’s going to be a bloodbath, because they can’t win big in the way they’re describing — someone’s going to have to lose. And when one of these players starts to lose big, there could be some kind of domino effect or contraction.

ED ZITRON: (02:21:20 – 02:21:41) Yeah. And I think the thing that people want to believe is the dot-com bubble thing — it worked out afterwards, because Amazon, Oracle, they didn’t die after the dot-com bubble, they’re actually fine. This isn’t like that. They’re bigger companies. They have bigger promises. I actually think Oracle could die. RIP Larry. Couldn’t happen to a nastier man.

STEVEN BARTLETT: (02:21:42 – 02:21:43) You don’t like these people, do you?

ED ZITRON: (02:21:43 – 02:21:44) No, I actually —

STEVEN BARTLETT: (02:21:44 – 02:21:51) Why? I ask this question purely because I want an answer, not because I agree or disagree. But why don’t you like these people?

ED ZITRON: (02:21:52 – 02:22:57) I don’t like being misled, and I don’t think regular people like being misled either. And I really don’t think that the average person can get away with bullshitting as much as these companies do. And I don’t think the average person gets anywhere near the level of affordance for failure and lying as these companies do. And I think there is a real economic and human cost to allowing these companies to run rampant and promise the world and never really get called up on it.

The tepid nature of criticism these days is so frustrating. There are some really great critics out there, they’re really great people, but seeing these ultra-rich, ultra-wealthy, ultra-powerful people lie through their teeth, or misstate, or whatever people want to call it, it turns my stomach, and I hate seeing people being misled. And I feel like I write at such length because I really want people to see why I’ve come to a conclusion. Am I right? Am I wrong? I think I am, of course I do. But I also just find it loathsome. I find these companies don’t make good products anymore. They don’t care about their customers, and they treat their customers with contempt.

Closing Thoughts

STEVEN BARTLETT: (02:22:59 – 02:23:03) If people want to go read more about your work — you have a great Substack.

ED ZITRON: (02:23:03 – 02:23:06) Ghost, actually. It looks exactly like it — I moved off of Substack in 2024.

STEVEN BARTLETT: (02:23:07 – 02:23:09) Oh, okay. And you also have a podcast, you do?

ED ZITRON: (02:23:09 – 02:23:10) Yeah, Better Offline.

STEVEN BARTLETT: (02:23:10 – 02:25:11) I’m going to link both of them below. So if anyone wants to read more, get more detail, and follow Ed, I would highly recommend it. It is fascinating. And you know, one of the things people sometimes struggle with when they listen to podcasts is you get lots of different opinions. And weirdly, I think some people assume podcasts are going to be like one person saying the same thing as the next person, and then the next person. That is just not the nature of information in the world, and opinions, and progress, and discussion. What happens is, people have different opinions. And I think my job, but also the listener’s job, is to try and parse through it and, over time, collect more of these reference points from different people and do your own research — whether it’s on your health or on something like this — to watch and learn.

And I would say, also, never believe one person, never believe one particular perspective religiously. Collect a body of evidence and follow the evidence yourself. But I love watching your YouTube, because it provides a different opinion, and that challenges me to think beyond my current opinion about what might be possible. So when I’ve heard you talking about how this is an economic bubble, and I’ve heard you talk about the CapEx spend with these big frontier AI labs, it really did make me pause for a second, and it really did make me consider that there could be a bit of fugazi going on here. And then it made me reflect on history and go, through history, there’s always a bit of fugazi in these moments. And what’s going to happen in 2027, 2028, when there’s a bit of a market pullback?

So I highly recommend people go watch, because you challenge me to think differently, and we need some of those contrarian voices to have honest discussions. So thank you for doing what you do. Really appreciate it. And I find you to be a very compelling, captivating communicator, and I feel like I’ve learned a lot today. So I appreciate that.

We have a closing tradition where the last guest leaves a question for the next guest, not knowing who they’re leaving it for. And the question left for you is: given that high-quality relationships are important for health and longevity, what should we be doing to improve our relationships and social connection?

ED ZITRON: (02:25:12 – 02:27:07) So this is actually connected to the AI bubble. I’m a critic, I’m a skeptic, we’ll call it. I have found that showing and appreciating and loving the people around you, and uplifting them and raising them up as you succeed, is the way we do that. Your success should be everyone around you. It’s not economic. Talking about Matt Hughes for a while made me really happy. This whole thing has been at times quite grueling and quite negative and quite brutal. But the love I’ve found and the joy I’ve found from community and the people around me — even in the small groups of, let’s say, fellow skeptics, people like Gary Marcus and people I talk to, Edward Ongweso Jr., Molly White, Brian Merchant — there are so many people who have been loving and caring.

And I think, especially in these very critical moments, when you’re very much dialing in on how negative things are, how bad things are, it’s about finding the people who maybe find it repulsive too. Finding the people, finding your people who can be there, and the people who will talk to you about it. Even Troy and Jake, my trainers, who are so excited about this — talking to them about this stuff as normal people, knowing that there are people going through their own struggles, but also to remind you that you are human too. And focus — I know this is kind of an all-over-the-place point — but it’s really easy to get hard locked on everything in life and to get away from why you do things, and focus too much on the work, when the most important thing at times is just to know there are other people feeling the way you do.

And when I hear from my listeners and my readers, the most common thing I feel is they feel like they have a voice, and they feel like someone is there for them. And I don’t think it can be understated how much it means when you just reach out to someone you love and tell them you love them, tell them their work rocks, tell everyone you — when you like an artist’s or a writer’s work, or a podcast like this — tell them you love it. We don’t do this enough, and we need to do it more.

STEVEN BARTLETT: (02:27:07 – 02:27:49) Well, that’s a good closing message. So if you have enjoyed the conversation today with Ed, please do let Ed know that you love it. Please leave your opinions down below and I shall read all of them. Ed, thank you so much. I’ll link to your website but also to your YouTube channel where people can learn more, and I would highly recommend it, because it is truly fascinating. And I think we need more voices demystifying a lot of the fugazi and the narrative in this moment, and you’re certainly one of them. I really enjoyed the conversation. Thank you so much.

Related Posts

• Transcript: Howard Lutnick Interviews Anthropic’s Tom Brown at G20 Innovation Summit
• Transcript: Sam Altman Interview on OpenAI, New AI Model and Global Growth – G20 Innovation Summit
• Transcript: The OpenAI–Hugging Face Incident – Black Hat USA 2026
• Transcript: Emad Mostaque Interview on Peter McCormack Show
• Andrew Ng: The Biggest Opportunities in AI Aren’t Where You Think (Transcript)