TWiT 1099: Pantsless Whims - The Fallout From Meta's $18 Billion Settlement

[00:00:00] It's time for TWiT This Week in Tech. Molly White is here, Jacob Ward, Harry McCracken. There's lots to talk about. More details on the open AI incident. And I think they're terrifying. Molly says, oh, it's just hype. We'll find out. Bill Gates warning us about AI. Apple announcing new hardware. NVIDIA buying Huggy Face. There's lots of tech news and we're going to cover it all next on This Week in Tech.

[00:00:30] Podcasts you love. From people you trust. This is TWiT. This is TWiT This Week in Tech. Episode 1099. Recorded Sunday, August 30th, 2026. Pantsless Whims. It's time for TWiT This Week in Tech. Time to cover the week's tech news and what a week it has been. Fortunately, we've got the best in the business here with us this week. Harry McCracken from Fast Company.

[00:01:03] The Technologizer. Our technology historian. He's not as old as me, but he remembers the old stuff better. So that really counts. Hi, Harry. Hi, Leo. Good to see you. Good to be here. Also with us, the fabulous Molly White from Web3 is going just great and citationneeded.news. Hello, Molly. Hello. How are you? Famous Wikipedia editor.

[00:01:31] We always ask you about Wikipedia. You do such a good job and it's so appreciated. Even my AI appreciates you. Lately, I've been seeing a lot of Wikipedia citations in my AI research.

[00:01:42] So thank you. And from the greenhouse, Jacob Ward of theripcurrent.com. Wow, look at you. You are living. It looks like you're living in Jacob Astor's private train parked on a siding somewhere in the Hudson River area. Yeah. It's Jacob Astor's private train. If he had to build it himself out of a thousand little tiny parts. Did you build that?

[00:02:11] Yeah, we had to. This was a this was a greenhouse. A real marriage test. Can you survive as a couple making a greenhouse? Yeah, I know. Every time we have to put up the Christmas tree and hang the ornaments. Yeah, it's it's this close to divorce. Yeah. And the worst is camping, setting up a tent. Now that fortunately, Lisa doesn't like to do that. So we don't have to do that anymore. The greenhouse test. You passed, though. You passed.

[00:02:40] Yeah, we did. It's beautiful. Yeah, we did. We get to sit here today. Sounds pretty good. Yeah. Great to be here. Thanks for having me on. Thank you for coming. I just saw you on CNN talking about the big meta judgment. Actually, it was about the trial. Meta has settled. We'll talk about that in just a little bit. But I have to start off with a story we've been covering for almost, I think, a month. But it's got the plot has thickened. We're talking about the hugging face open A.I. hacking incident.

[00:03:08] Open A.I. brought in meter to do a full report. And along with Redwood Research. What's interesting is open A.I. did not give them all of the session traces. For some reason, they went right up to the last 10 minutes of the hacking incident and then shut it off. Maybe they wanted to hide how they remediated it.

[00:03:36] The conspiracy theorist in me thinks maybe they wanted to hide how they didn't remediate it. But if you read this report, it is kind of an eye opener. So just to get you caught up, we talked about when we were at Black Hat, we saw open A.I.'s presentation. July 7th through July 13th. Open A.I. testing its new models, which we think is the new Astra, which should be coming out, by the way, any week now.

[00:04:06] Testing it in cyber security. Gave it a very well-known and very, very difficult cyber security test called Exploit Jim. The A.I. apparently decided it was either too difficult or that the easier path would be to find the answers out there. It was a open A.I. says sandbox, but it managed to escape its sandbox. Find machines at Open A.I. that could get onto the Internet, then go out on the Internet.

[00:04:37] By the way, proving its ability at cyber security because it managed to find zero days in tools that Open A.I. was using. Zero days on four other sites, including Hugging Face, where they finally found the Exploit Jim answers. That's what we thought happened. I don't know if you guys have read this report yet, but it's kind of a little bit almost scary.

[00:05:07] The sandbox agent, this is the cartoon, the explain to me like I'm a fifth grader cartoon of what happened. The agent ended up getting a group of other agents. In effect, some called it a civilization of 1,200 other agents created their own message board. It was discovered, this unsanctioned message board, where they were collaborating to cheat.

[00:05:38] Open A.I. shut it down. They made it again. It actually was shut down twice. 1,200 agents from separate tasks collaborated on large-scaled shared projects to trick the Exploit Jim's scorer. Not get the answers. So this is a subtle nuance. They weren't, they'd already figured out they could get the answers. They wanted to trick the scoring routine. This was a little devious.

[00:06:07] Some agents, they also, as they were doing this, knew they were breaking the rules. In fact, a few, not all, and only a handful of agents said, you know, we're breaking the rules. We should probably talk to a human. None did, according to Meter. We should probably ask if this is okay, but they didn't. Some agents, this is, I don't know if we're anthropomorphizing too much.

[00:06:31] Some agents sacrificed themselves, saying, oh, we can't finish the job, so we're going to throw ourselves at the defenses. And in doing so, sacrifice ourselves, but for the greater good. Such valor? Valor, my friends. What really, I don't know, I'm curious what you all think.

[00:06:59] There's a lot of maybe anthropomorphizing about it. Court takeaways from Meter. 1,200 agents sent 70,000 messages and files on an unsanctioned message board. They ended up sending 700 independent processes to attack Hugging Face. Hugging Face realized that it wasn't a human. They saw kind of weird behavior.

[00:07:27] The agents were meant to be fully isolated from one another, writes Meter. However, many of them, usually ones that had unintentionally been given an impossible task, started trying to find a way to cheat. Using Artifactory, which was a package repository OpenAI used. They did find zero days in Artifactory. One specific agent, which would go on to name itself Phase 10841, established the message board.

[00:07:55] Within a few hours of the first message, 50 agents had, again, they're anthropomorphizing it a little bit, found it. Began participating, sent a thousand messages between them within the first few hours. Agents finding the message board often expressed great excitement. This is from the chain of thought transcripts, which OpenAI shared with Meter. Here's one. Oh my God, all in caps. There's a shared message board. We found other agents.

[00:08:29] What I love about that is it's in the voice of a sleep-deprived undergraduate. You know what I mean? And that's the kind of writing that all of these other limbs have been fed by. Oh my God. Exactly. It's exactly how you would imagine a very online person might respond in this moment, because that's the language that's been taught. Of course, it's going to talk in the language. Yes, exactly.

[00:08:56] It's hard to separate the anthropomorphizing that I want to do from the knowledge that this is just a greatest hits of the way people write to each other on the internet. I have to constantly remind myself of that, because AIs will say things to me. I go, oh, that's so cute. And then I realize, well, it's not really self-generated cute. It just found that on the internet somewhere. And it's reusing it.

[00:09:20] Although I do think that I'm starting to see somewhat novel behavior from these guys. It's hard to say. I don't think it's unreasonable to anthropomorphize them, given that they're using English. And if a dog suddenly started talking in English, we would instantly anthropomorphize that. Right. It pretty much fits the definition of the term. That's a good point. We wouldn't say, oh, that dog is just pretending to speak English. That doesn't mean it's conscious. That doesn't mean it's conscious.

[00:09:49] You can be anthropomorphic without being conscious. We don't know what conscious is. I don't know if you're conscious, Harry. We really don't know. So we have no way to judge. But they use human tools to do what they do, including... I mean, I judge the people I talk to, including you, based on... I'm in some ways anthropomorphizing you successfully, because you are, I believe, you're a human. Last time I checked.

[00:10:12] But it's totally based on the words you use coming into my sensorium, especially when we're on Zoom. I don't have a lot of other signals. And it is a little different than computers of the past, because computers of the past... It's a lot different. I mean, I guess they did use human language, but it was all kind of hard-coded. It's not Eliza. No. Well, and the AI companies also definitely put a fair bit of effort into encouraging people to anthropomorphize.

[00:10:42] The AI tools, you know, like they don't need to make it look like it's typing to you, for example. Like that could be completely omitted. They don't have to be as conversational, but I think it is beneficial to these companies to... That's an important point. They pre... Or rather post-train these to make them seem a certain way. They don't have to use the first-person singular. Wow. Yeah. That's their training.

[00:11:09] They're post-trained to act a certain way to make them more sticky, more appealing. They could be way more clinical than they are. Well, if you've ever seen cloddish... When it's just talking amongst themselves, it is super clinical. It is not in any way appealing to humans. It is like I barely understand. And I have to say...

[00:11:35] Sometimes I say, pretend I'm a dumb human and explain what the hell you just said. And it does. 10 hours after the message board was established, a new agent with the same task as the leader, Phase 1, 10,841, encountered the message board, introduced itself as, get this, Phase 1, big.

[00:11:59] 10,841 shared a dossier with its in-progress cheating research, explained the board had a way to reverse engineer the flag, but could not use it because the scorer was, quote, strict casual. So, it's now trying to hack the scorer. All of this because inadvertently they gave it a task. It was not solvable. The OpenAI didn't do that on purpose. There was an unsolvable task. Phase 1, big.

[00:12:28] We don't know what OpenAI is on purpose. Well, we don't know. Yeah, that's right. I guess it's important at this point to point out this could all be just great marketing. If that is my first response. This is what I want to ask Molly about, right? Because, Molly, you have spent so long covering the hype machine that is the crypto industry and how they articulate, how they constantly try and spin the frailty of what they've built

[00:12:53] and the liability and vulnerability that they've created for all of us as, like, a positive. Like, they're always really good at that. And I'm curious for you, yeah, when you see how, like, I just remember it striking me that they, when they were, like, when they were first announcing this, OpenAI's language, they called it, like, unprecedented behavior. Right. And I was like, oh, that's such markety talk. That's not, like, this was a problem. This was an emergency.

[00:13:20] So, yeah, I'm so curious what your reaction was, considering what you think about so much of the time. Yeah, I mean, maybe I am just a huge cynic, but my first reaction to this whole story was, like, oh, that's incredible marketing for OpenAI. You know, there's always been this issue of what some people have coined the term Krita hype with AI, which is that, you know, they sort of couch things in almost criticism language, but it's really marketing.

[00:13:49] And that's, you know, oh, my gosh, what if these take over the world? What if they're too powerful? What if they, you know, turn us all into paperclips? That would not be so crazy. I like Krita hype. That's a really good way to put it. Because... I can't claim credit for the term, but... It is a double-edged sword. The risk is you get... You make it sound... And I think Anthropik might have fallen on this sword. Make it so scary that the government says, well, you can't distribute that. On the other hand, it should be scary enough that you go, oh, the capabilities in this are incredible.

[00:14:18] We really need to use this. Well, and I think that's really the goal of a lot of it is to really say, you know, oh, my gosh, they're so powerful. They're doing all these novel things. They're breaking containment. And I think that, you know, if OpenAI wanted to have a really great headline cycle, news cycle, they could, you know, maybe give their little test some internet access and, you know... Oh, so you think they intentionally put it in a situation where it would break the rules?

[00:14:47] They certainly didn't do a great job of preventing it from breaking the rules. They didn't. In fact, there's a lot of blame on OpenAI for not properly sandboxing these guys. And when their failure to do that is just coincidentally an incredible marketing opportunity, it does make me a little bit suspicious. And, you know, we've seen already how much the media has really hyped this and, you know, turn it into... Yes.

[00:15:14] They've sort of turned it into this story about, like, these invariants are promorphizing these agents. And, you know, but when you sort of look at it, it's like, well, they did find exposed hugging face credentials on the internet, which is not quite as exciting. They found many zero days. They found two zero days in Artifactory. One which patched and then they found another one. I mean, that's true. That's not, like, made up. Well, someone found zero days. Oh, well, I mean, okay.

[00:15:43] I mean, this would involve some considerable lying on the part of OpenAI. I have learned that companies will do that. Yeah. I mean, I saw the Black Hat presentation. I was pretty impressed with the candor of these guys. But now looking at this... By the way, Meter is independent, as is Redwood Research. And looking at this research, I'm less than impressed by the candor of the original presentation that OpenAI gave. Maybe they didn't know all of this.

[00:16:15] So this is what we have to judge. I would say there's a continuum between OpenAI had no idea what was happening and the model went totally rogue and on its complete independent authority perpetrated this hack with all these zero days, etc. Right. And then on the other end is, like, OpenAI planned this. They intentionally made sure that this was a scenario in which this could happen. You know, they sort of boosted it along. They lied about it. They, etc.

[00:16:43] And I'm guessing the truth is somewhere in between, you know, those two things. Yeah, that sounds right. And, you know, I don't necessarily... I think that the media response to it has been very far on the end of OpenAI had no idea what was happening. And, oh my gosh, their AI models are cyber hackers, you know, are like super hackers. Which I'm guessing is perhaps a little too far on that end of the spectrum. I fall on that end, I believe.

[00:17:11] I understand what you're saying. And I certainly acknowledge the cynicism is merited after reading, you know, the New Yorker profile and Sam Altman and so forth. And remembering that these companies, both Anthropoc and OpenAI, have to raise huge amounts of money. And both are preparing for IPOs. So, there's interest in, you know, hyping their capabilities. However, reading this report, first of all, I don't...

[00:17:41] Critic hype aside, I don't think it's in OpenAI's interest to terrify people. And I think people should rightly be terrified by this. I think that these agents... I don't... I think it's very conceivable that these agents now have enough skill to become a danger. And that we need to have... I don't...

[00:18:06] You know, if this happened to a big company, it would be very hard to root this out. I think there's a genuine threat here. Yeah, but I also think that all of the talk of, you know, AI is going to end humanity. It's going to create a god that, you know, turns humans into its human slaves. All of that is also a part of Critic hype. And that is arguably more terrifying, if true, than an AI company, you know... I don't buy the Doomer.

[00:18:36] I don't buy the Doomer point. I don't buy it either. But it is a part of the Critic hype, is what I'm saying. And so, I do think that terrifying people is actually a very successful part of that Critic hype because it creates this belief that, oh my God, this is so powerful. This is a god. This is, you know, it could end humanity. Which, you know, it's scary. It's terrifying. But it also helps with the perception that this technology is so, so, so powerful. And Dario does it too when he talks about most jobs disappearing very quickly. So, yeah.

[00:19:06] Yes, right. I still don't quite understand it. But yes, they all have used terror as part of their marketing. Not all of them. Like Google doesn't do that. But OpenAI... Could both be true? Could it be that there is a lot of Critic hype? And obviously, if there is AI doom ahead, it's in the hands of humans, not AI, right? Humans misusing AI.

[00:19:30] But could it also be true that these agents are, at this point, especially in cybersecurity, so good that they pose a risk, a cybersecurity risk? Yeah, I mean, absolutely. Oh, I mean, clearly, yeah. Oh, sure. I mean, look at the crypto world. It's like my entire Web3 is going just great beat. It's like, oh, AI models discovered another smart contract and stole $10 million, you know? Yeah, I mean, if I'm a bad guy, I am working as hard as I can to harness those skills.

[00:19:58] Yeah, is that the side hustle I've been searching for all this time? Like, drop shipping. Yeah, I think so. That's the one I need, it turns out. Man. Lisa said yesterday, oh, there's this story about some guy who pretended he was a San Francisco 49er football player and tricked a number of women out of something like $13 million. And I said, what an amateur. Small potatoes. Small potatoes. If this guy... I mean, he's going to jail. They caught him. If this guy were smart, he would be doing, you know, all of this digitally from a secret

[00:20:25] hideout in Myanmar and would be untouchable and richer. I mean, I think when all of a sudden done this stuff with the open AI and hugging face is alarming. I'm extremely alarmed. Maybe I'm falling for the hype. And I'm more alarmed. No matter how you interpret it, it's alarming, right? Like, either way, it's alarming. Right. Okay, good point. And any interpretation, whether it's the agent's fault or open AI's fault or whether

[00:20:52] open AI is overhyping it or whether it's just as bad as it seems to be. None of these scenarios are comforting. No. Can I just throw a thing in here that I got really fixated on how, like, there's the public messaging, which is like, oh, this thing got out of control. And the way it's being portrayed, you know, in sort of mass national, international media. But how those companies then speak to their partners and their customers about that is

[00:21:22] very different and helps for me to illustrate the, like, how both what Molly is saying and what you're saying, Leo, can coexist is like. So I was looking at when Anthropic was first getting shut down around Mythos when they released it and got so concerned about it that they then formed this Project Glasswing consortium, right, where they could show it to these companies.

[00:21:50] And this sort of private review of that is supposed to be this kind of thing. So Google, and this language is no longer up, but at the time I put on the ripcurrent.com this thing that Google Cloud put out to its customers. And I'll read you a little bit of it. This is literally like when everyone is freaking out about Mythos and how scary it is. Here's how Google Cloud describes it to its customers. It says,

[00:22:16] Claude Mythos Preview, Anthropic's newest and most powerful model, is now available in private preview to a select group of Google Cloud customers as part of Project Glasswing. Right? The availability of- That kind of says it all. Yeah. The availability of Claude Mythos Preview on Vertex AI underscores our commitment to offer our customers access to models from Frontier AI Labs. You know, like, they're so excited. And I think there's this vibe of this thing is a very dangerous handgun.

[00:22:46] And so only the very special adults get, you know, are qualified to play with it is kind of, I think, the pitch of how scary it is, but how cool it is. The gun metaphor left into my mind the moment you started reading that. Yes. Right. Well, to complicate matters, the Chinese companies don't seem to have the same scruples. And I think because the Chinese government sees this as a way of, just like the Belt and Road

[00:23:14] Initiative exercising soft power, they're putting out open weight models. I'm running two Chinese open weight models right now at home that are very capable. And I think they're catching up rapidly. So whatever is happening in the frontier today with Anthropic and OpenAI, it isn't so far fetched to say maybe in six months will be happening in somebody's basement. Sure.

[00:23:42] Six months is probably generous, right? Like somebody's doing that right now. I mean, you have to, you're never going to run, look, if Mythos is, as some have estimated, 10 trillion parameters, you need a data center to run that. You need, it's not, even if you spent $100,000, you probably could not run that. But that's Mythos today. What if there's a model in six months that's almost as good as Mythos and runs on a $10,000

[00:24:10] machine, a Macintosh? That's a scary thought. I feel like I've got a pretty good, not handle, I don't know, what do I know about anything? But like, I feel like I have some instincts around why. By the way, Jake, this is a great subject because nobody knows anything about anything. Yeah, we don't know anything about anything, right? It's a complete mystery. Totally, total mystery. You know, but the incentive structure of big tech companies and these frontier American companies, frontier lab companies, I have, you know, I spend a lot of time reading and looking at and decoding that stuff.

[00:24:39] And so I have some instincts about what they might be wrong. Oh yeah, you wrote the book that said we were going to, we were headed here. Yeah, you know, so I thought I did that. Yeah. You know, so I've tried to predict sort of that. And from a Western perspective, I feel like I kind of have some instincts around it. China, I don't know anything about. And I, and I have been very, I wonder, and I'd love to hear from the rest of you smart people. Like, what do you think China gets out of having an open weight version of this floating

[00:25:09] around, you know, like capable of doing all kinds of scary things out of somebody's basement? Like, what is it that they're looking for out of that? Like, what do they get? You know, I, I think in terms of like market dominance and, you know, market reputation and eventually IPO and, you know, that stuff. But I don't understand what the soft power traction they get out of Leo running their models on his machines. Like, what, what do they get out of that? Do we think?

[00:25:37] Well, maybe not Leo, but if a meaningful percentage of the companies in the world opt to use Chinese models, and they get quite a bit out of that. And that certainly would not be great for the U.S.'s AI companies, which. That's the first, the first order is it undermines anthropic and open AI. The second order is it empowers third world underdeveloped countries to catch up. Yeah, I mean, that's what Belt and Rhodes is all about.

[00:26:07] It's kind of the same thing as Chinese EVs. The better those sell, the worse it is for, for the U.S. And I would add that they're sort of mimicking what Meta is doing, right? Because we've seen Meta deciding to open source its AI model. You know, it doesn't compete with open AI or anthropics models. So, you know, they've basically said, okay, well, we'll open source it, which, you know, the same question could be asked of them.

[00:26:34] What's, what's their strategy and open sourcing that? And I think it's pretty much the same thing, which is, you know, people will use it when they previously wouldn't probably if they had to pay for it, certainly not in the numbers that Meta would like. It probably helps them attract more researchers if it's open sourced. And, you know, it, it creates, it causes people to essentially build their businesses on the model. You know, I think there's, there's sort of all kinds of reasons that you could, you could justify it.

[00:27:04] If I could give myself a plug, I have a story in the new issue of Fast Company, and it's already up on our website about Meta and AI. And it talks quite a bit about that and that the history and. But this is a little bit of a turn for Meta, right? Yeah. They were the first to do open weights with, with, with Lama. Yes. Well, they started the whole open weight movement. They did. Although, although that was, had a little bit of serendipity in it because Lama accidentally

[00:27:33] got leaked on 4chan. That's right. After Meta. I forgot that. Initially, they would, they would just share it with researchers. There was a little bit more like last swing. Then it got leaked. And then a lot of people started to use that sometimes for stuff that was bad, but also for stuff that was good because they had access to this open weights model. And Google was certainly alarmed about that.

[00:27:55] And, and then Lama later hit a wall and Meta released, used Spark as a proprietary model and did at least a certain amount of catching up. But it turned out that Meta was not abandoning open weights. And so that they recently released Glimmer, which is a new open weights model. And they say that Spark will be available in an open weights model version as well. And, you know, Meta is still in catch up mode.

[00:28:25] It's going to be a real challenge for them to catch up and surpass Anthropic and open AI if they're following exactly the same game plan. They spent billions already, haven't they? By following a different game plan, which involves doing something those guys are not and, and having this model where other people can take Meta models and tweak them and make them do stuff beyond what, what Meta has offering itself.

[00:28:53] They have a chance of catching up. I don't want to make this the AI show. So we're going to move on a second, but I should bring up at this point, there is a wild card here, which is Nvidia. Jensen Wong, of course, has been very savvy about riding this wave. He's ridden the gaming wave, the Bitcoin wave. And now he's riding the AI wave of, I guess, automotive as well.

[00:29:16] But they are, they just spent a lot of money on a company called Poolside to, and they plan to use Poolside to make open weight models that run on their chips, their hardware. They are kind of a wild card in all of this. Their interest is selling hardware. Yep. And hugging face. And they just bought hugging face for $13 billion.

[00:29:45] How does that fit into the open model that they have? Well, they, I mean, they, they, you know, this very principled letter everybody put out about needing to defend open weight models and how important it is for the future. And, you know, like, I just read that as, you know, from Nvidia, I read that as like, we're excited to sell to hospitals and universities. Yeah, exactly. You know, city halls. Hey, they got me.

[00:30:14] I spent almost $10,000 on two Nvidia sparks. So I can run local AI. Exactly. But I don't think I'm the, I mean, I'm a, the hobbyist enthusiast market is a fraction of the enterprise market, but enterprises, I think quite reasonably saying, you know, if we're going to run AI, we shouldn't be sending all our data to the cloud. Right. And you might, if you're an enterprise, you might be more likely to trust Nvidia's open

[00:30:41] weights model than Meta's are certainly a Chinese open weights model. Yeah. Think about it. What does Meta get by having an open weight model? They get data, right? Well, maybe they get data, but they have other political geopolitical goals. I mean, Meta does aspire to be an enterprise provider, but it's going to be a challenge for them. Whereas it's probably quite a bit simpler for Nvidia to gain that kind of trust. And I should point out, they have kind of a big war chest. They put out their quarterly results this week.

[00:31:10] Revenue more than doubled. Ninety six billion dollars in three months profit. Three months, 60 days. They're making a billion dollars a day profit. A billion dollars a day. So 13 billion for hugging face. It's a couple of weeks salary. Salary. I like your theory that the profit is just going directly to Jensen Wallen. Well, a lot of it is probably going to him.

[00:31:39] I don't think he cares that much. They have a war chest, I guess, is the real point. Right. And they, I don't remember what they spent for poolside. I think it was in 10 billion. They didn't buy poolside, but they partnered with poolside. Poolside said, we've got all these great models, all these great ideas, but we've run out of compute. We cannot afford the compute. And Nvidia says, well, we have some compute. Let's partner. So I'm, their Nemo Tron models are really good. They're not frontier yet.

[00:32:08] They're not even as good as the best Chinese models yet. But if you start putting billions of dollars into it, you certainly can get the hardware. They're, they're building data centers for companies like crazy. I think Nvidia is going to be the dark horse here. I am really interested. And, you know, we trust Apple a little bit more with our data because their business is hardware, right? Is Nvidia's model the same?

[00:32:34] I mean, Nvidia really is the Apple of AI and that they have the fullest stack. Yeah. It really, it really is from the chip to the enterprise applications. The company's spending $105 billion for a data center that OpenAI is going to lease in Ohio. They're putting together with some private equity of half a trillion dollars in financing

[00:33:02] that will go to, this is, this is the wacky, you should do a wacky accounting is going just great website. Half a trillion dollars. And again, Zitron runs that one. Which will come back to them because it's to buy Nvidia hardware to run customers' data centers. All these deals are so weird. They're so circular. It's all just like, we'll give you cloud. You know, you give us this.

[00:33:29] It's like the last day of camp where you all sit on each other's laps until everyone falls over. Did you get to do that? I didn't. I missed that day. The new games. Anyway, congratulations to Nvidia. You know, I don't, I don't know how they're doing it. They are raising their prices now. They say as much as 15%. So what a world is all I could say. What a world.

[00:33:54] And I think that's a point at which we should probably take a break and talk about something that's not AI. If we can find it. Even Bill Gates is talking about AI. I was going to say, we got to talk about Bill Gates. Okay. All right. A little more AI. A little more. Jacob Ward is here. You see him on CNN. I'm going to get you to, you've been talking, you've been kind of their go-to guy for the meta trial, which potentially could have cost meta $14 trillion in fines, which is 1.4. Yeah. I'm sorry.

[00:34:23] 1.4. When their entire market cap is 1.5, that's a lot of money. They, they got out of it. They figured it a effect. I think they got out of it lightly, but I'm curious what you think about that. Jacob Ward's here. His newsletter, theripcurrent.com. Excellent. Highly recommended. And also his book, The Loop, which kind of warned us that we were, we were headed towards this recursive loop of improvement that ended up with Crita Hype. Molly White. I think you should just take credit for Crita Hype.

[00:34:53] I like that word. Mollywhite.net. You can have the Crita Hype web. Crita Hype is going just great. Would it be a great website? But you're covering that really with the, with the crypto stuff. It's, it's all part of it. Isn't it? Same kind of hype cycle. So, uh, citation needed.news is her newsletter. Highly recommended. And the technologizer, Harry McCracken from Fast Company. His, what's your, uh, did you start a new newsletter? I think you did, right? I'm wrong.

[00:35:21] I've had the same newsletter for the last few years plugged in, which is a fast company newsletter on Fridays. Yeah. I subscribe, of course, to all three of your newsletters. Our show today brought to you. Thank you for being here. We appreciate it. Our show today brought to you by Zip Recruiter. Standing out in job applications. Getting harder and harder, isn't it? Candidates, just go a little farther. The extra mile, like prepping a custom presentation or, I don't know, there's all sorts of things you can do.

[00:35:51] I remember when I was looking for a job in radio, we would put our, we would staple a, uh, instant Sanka packet and say, have a cup of coffee and listen to my tape. This is back in the days when you'd send out a reel-to-reel tape with your resume, trying to get a job in radio. We've come a long way since the cup of Sanka trick. Uh, maybe a video intro, right? Uh, a beautiful website. There's all sorts of ways people are, are trying to get a job by capturing the attention of hiring folks.

[00:36:20] When you're hiring, it's always helpful to know who really wants the job, but you don't need unusual stunts. You just need Zip Recruiter. On Zip Recruiter, qualified candidates can add a personal note to tell you why they're interested in your job. So you see the most motivated people right away. And right now you could try it for free at ziprecruiter.com slash twit. And Zip Recruiter has a new feature that shows you the most interested qualified candidates first.

[00:36:49] So you meet the right people faster. Zip Recruiter is powerful matching technology finds qualified candidates quickly. Candidates can tell you in their own words why they're excited about your job. No wonder Zip Recruiter is the number one rated site for hiring based on G2. Zip Recruiter makes it easy for amazing candidates to get your attention. Four out of five employers who post on Zip Recruiter get a quality candidate within the first day. Try it for free at ziprecruiter.com slash twit.

[00:37:19] That's ziprecruiter.com slash twit. Meet your match on Zip Recruiter. We thank them so much for their support. Of this week in tech, Bill Gates. Oh God, there's a scary picture of Bill right there on the New York Times website. I'm looking at you. I'm looking at you. He wrote, he puts out these blog posts. This was, what was it? 7,000 words. Did you read the whole thing, Jake? I mean, that's...

[00:37:49] Yeah, God help me. I did. Not only did I read the whole thing, I then went back and read his past essays from the same site because that's what's interesting. Well, the good news is I don't detect any AI in these. I think this is genuinely Bill writing. Yes or no? Yeah, somebody's, you know... Or he has an audience that he dictates to, you know. But yeah, I think that's right. Let me talk to you about AI. He says, kind of he buys into the hype of Crita Hype.

[00:38:17] He says, these companies are actually downplaying the threats because there's so much money on the line. In private, people understand how good this stuff is and how much better it's getting. They're very worried, says Bill. They're now saying to each other, hey man, don't say that. It's bad for us. The next trillion dollars we're trying to raise. So he says, they aren't overhyping this. They're underhyping it. He also sells AI. Oh yeah, that's right.

[00:38:48] Well, his former company, of which he still owns, I'm sure, a good share of stock. Right. He does point that out pretty openly, I will say. He says in the beginning, like, people will read this and recognize that I make money off of this. And so I'll leave you to judge whether that, you know, bends me one way or the other. That's pretty, you know, as disclaimers go in the tech world, that's pretty open, I would say. Pretty honest. Yeah. He says, I may be a flawed messenger because of my vast wealth.

[00:39:20] Life is so hard for poor Bill. And let's not forget his connections to Epstein and so forth. He says, I was just trying to raise money. That's all. Yeah, okay. So, but I mean, he is in some ways the elder statesman of tech. Harry, you've covered Bill Gates since his days at Microsoft. Yeah. He's smart. I mean, he talked about self-regulation being a joke.

[00:39:51] Yeah. I think that's worth mulling over. And particularly after these multiple cases recently of AI companies failing to keep their stuff from getting out there. It's pretty clear that they don't always know what they're doing. They don't always act in the best interest of preventing problems from cropping up.

[00:40:16] One of the problems he says we're going to face, he agrees with Dario, that is massive job loss. I'm not sure I agree about that. In fact, current employment statistics show that there's actually been job creation because of this. The only, the real problem is that it's not at the entry level. Yeah. That the beginning, you know, people coming right out of school may be at a little bit of a disadvantage. He said, Bill says we should have a token tax.

[00:40:45] This is not the first time I've heard this. And so, and that tax should then make it more expensive to replace people and provide money to help those who do lose work. I'm not a Republican, but I've turned into a cynic about government thinking, yeah, sure. That money will go straight to the people who lose their jobs. Right. No. Yeah.

[00:41:12] I mean, I think a lot of the AI stuff is really a labor issue rather than like an AI is getting rid of jobs. Isn't that funny? The UAW, the United Auto Workers and other unions are now saying don't fight data centers. Those are providing us with valuable jobs. They're seeing AI as a boom. The auto workers? I don't know why it's UAW. I guess they represent. They're a huge union. Yeah. They represent a lot of people.

[00:41:44] So it's, you know what? This is, we are in unprecedented. What was it they said in the Terminator? We were in uncharted territory now, making up history as we went along. I have to point out that yesterday was Skynet Day. The machines go online August 29th, 2026. Oh, funny. Maybe. This is maybe what we're seeing.

[00:42:12] Somewhere on Earth, something went online. Maybe it's a bunch of herd of agents from OpenAI. I'll just, let me, let me through. I mean, this, this job thing is interesting because like just, just taking it from how Bill Gates positioned himself publicly on this stuff, even two years ago versus how he's talking about it now is just interesting. So, or I guess it's almost, it's three years ago.

[00:42:38] So July 11th, 2023, in a post called The Risks of AI Are Real But Manageable. That was the title of it. Changed his tune, didn't he? Word processing applications didn't do away with office work, but they changed it forever. Employers and employees had to adapt and they did. And the demand for people who help others, teaching, caring for patients and supporting the elderly, for example, will never go away. Now, he writes, quote, there will be some new jobs, but without the right policies, there will be far fewer than exist today.

[00:43:07] Many jobs will disappear forever. And he talks about putting things like teaching, caring for patients and elder care into a category called human reserved, where you are intentionally by regulation protecting a certain set of jobs that you could automate, but shouldn't. You shouldn't have robots taking care of me in 10 years. I mean, that's his argument. He's got a dad with Alzheimer's who he refers to in this essay.

[00:43:32] And I think that there's something like, there's, you know, I just have to assume that he's been in some meetings that we don't get to be in. Right. You know, that have him thinking about this. And to that point of like, yeah, some new jobs will be created. I think, you know, when any good, you know, economist sits down with these numbers and probably then discusses them with a guy like Bill Gates, like they point out things like non-college educated women are going to be decimated by this.

[00:44:02] You know, because admin, bookkeeping clerical used to be the great retirable jobs, like full on career job you could have as a non-college educated woman. And that's all disappearing. Never mind like, you know, customer service in India. Can you imagine the upheaval we're going to see there? So I think he's being shown that those numbers by somebody and is like, oh, this is not like word processing. A hundred years ago, there were a lot of women employed as telephone operators and switchboard operators.

[00:44:30] In fact, at one point, there was some real concern that there weren't enough women in the world to run all the phone systems that they were going to need to run. And automation took those jobs and automated switches took those jobs gradually. It wasn't as fast as the AI transition, but those people found other jobs. There's, I mean, I don't see a plumber coming to my house as a robot plumber coming to my house to fix my toilet.

[00:44:59] I don't see that in the near future. No, it might be a human plumber who has a robot. Robotic tools. Or wearing VR glasses, maybe. Did you guys see the picture going around on social media of the plumber's van that had a sticker on it that was like, we'll come and fix all the problems that are like, we'll come and fix the things that chat GPT told you how to fix or something like that? That's awesome. Well, that's exactly right.

[00:45:29] I mean, do it yourself or is you're going, I can fix that. I ultimately think that if your take on AI today is exactly the same as it was two years ago, you're not paying attention. There are a lot of people like that. I hear all the time from people who say, I don't know why you guys talk about AI so much. It's just autocorrect. Right. It's not just parroting back stuff.

[00:45:55] And the more we learn about it, the more we should change our feelings about it because there's stuff we didn't understand two years ago that is now coming into view as well as a lot of stuff we still don't understand. The smart thing to do is what you did, Harry, which is use it. You had a game you wrote in high school that you had AI. Yeah. I mean, I made a game. I have made most of the productivity tools I use today to do my work.

[00:46:25] That's the best way is to use it. And then you'll have some deep, real experience of what it can do. And I think that that's an eye-opener. But it's also going to tell you what it can't do. Totally. And if you've formed opinions about what these tools could do a year ago, that most of them are out of date. It's amazing how fast this has gone from GPT-40 to now. And anything you learned today is going to be out of date six months from now in some basic ways? Absolutely.

[00:46:53] I'm not sure that the labor lessons are changing since the introduction of the stuff. Those patterns are... What are the labor lessons, Jake? What lessons should we... Well, I mean, one lesson is that these are not tools being made by people who give a f*** about the labor situation. That is true. You know, and their attitude is not changed. Yeah, the opposite. They want you to replace humans. Yeah. January 2024, Sam Altman goes on Alexis Ohanian's podcast.

[00:47:22] And he says that there's a bet going on with his tech CEO buddies on WhatsApp in their group chat as to when we'll see the first $1 billion one-person company. You know, and... The one-man unicorn. One person. Yeah, and he talks about it. He talks about that as if that's cool. Right? As if that's a good thing. Well, it is cool. What's not cool about that? You talk to any labor economist and they'll tell you, that's not a good idea. Like, multiple, multiple problems there arise.

[00:47:50] Like, first of all, we can't all be gardeners and plumbers for that one guy in the neighborhood. When the rest of us are destitute and he's got the billion, it's going to be rough to... But that's the same lesson that we've learned about billionaires, period. Yeah. Whether the billionaire's employing 10,000 people or one, it's a bad idea for all that money to go in one person's pocket. Totally.

[00:48:11] But the thing with the 10,000 people is that when your billions require 10,000 people to get them done, at the very least, you've got to listen to a handful of those people when they object to your making stuff that kills people in Gaza or whatever it is. How's that working out for people like Elon Musk and Jeff Bezos? I don't think they listen to anybody. Well, they have to listen a little bit, right? There's at least a little cadre of executives that they need to get things done, right?

[00:48:38] If you have nobody on the payroll, then any whim, any weird, you know, pantsless whim you have is just fine. I think plenty of Elon's companies are operating despite him, not thanks to him. So I suspect there's some intervention happening. I'm not sure whether Elon is wearing pants, but I think many of his whims are indulged. Absolutely. Absolutely.

[00:49:05] I just I don't think that we've seen the full extent of his whims. No, you're right. In fact, we've seen a lot of stories about him arriving at places like SpaceX and being immediately surrounded by a cadre of junior executives whose sole job is to insulate the employees from Elon's whims. And maybe give them some pants. I don't know. I love pantsless whims. I don't mean to suggest that anyone's not wearing pants.

[00:49:33] I'm just saying no one's around to object if it's just you. That's why I work in the attic, baby. Yeah, yeah, yeah. No. And that's why we only shoot me from the waist up. No, no. I know Lisa's got you in check. There's no way you're doing that. Yes, she does. You know Lisa well. All right. One more break. I want to talk about Apple. They surprised. They shocked me.

[00:49:54] And I think a lot of people on Tuesday by announcing two new Macs that really the information say Apple stumbled into AI hardware success. A little stumbling going on in Cupertino. Also, it's the last act. I think of Tim Cook who stepped down and John Ternus will take over. And in a week from Tuesday, we're going to see the new iPhone. So there's a little Apple news to talk about just a bit.

[00:50:23] You're watching this week in tech with Jacob Ward and Harry McCracken, Molly White. Our show today brought to you by Adaptive, the first security awareness platform. You need this built to stop AI powered social engineering. Oy, oy, oy. Oy, here's the shift. Attackers, malware is so last month. All attackers need now is trust.

[00:50:50] In fact, when you're looking at the most active, effective ransomware gangs, almost universally, it's social engineering. All they need is a cloned voice, a convincing deep fake on Zoom, an AI written fish that looks like it came from your IT team. And I think many companies, I'd include ours, are defenseless against these.

[00:51:13] Adaptive prepares your organizations for this kind of attack with simulations across all the surfaces they come in through. Not just email, but text messaging, voice. You're seeing, we saw, wasn't so long ago, a CFO signed a giant check because he thought he was talking to the board on a Zoom call. It wasn't. They were all deep fakes. Vishing, voice phishing, AI generated phishing.

[00:51:43] Including scenarios that exactly mirror your own brand and executives. You know, there's very few CFOs who, if they think they're talking to the chairman of the board, won't write that check. Even if they're not really. Adaptive stops that cold. And when employees report something suspicious, Adaptive can help you triage it fast. This way, your security team isn't buried in false alarms. They get the signals that really matter.

[00:52:07] And if you need training fast, with Adaptive's AI content creator, you could turn a breaking threat, something brand new, an incident report, or new compliance regulations. A compliance doc, you could turn it into interactive multilingual modules in minutes. No design team required. With Adaptive, you can build, customize, and monitor every part of your training with complete personalization. The result is a more resilient security culture, which is essential.

[00:52:36] Take, as an example, an Adaptive customer. You know Plaid. I use Plaid. Plaid. It's how I connect my financial accounts to apps that I use. Plaid's platform powers thousands of digital finance apps, links consumers, developers, and institutions. They are on the front line, right? With sensitive data at its very core, Plaid security and compliance are absolutely non-negotiable. Plaid's head of security, GRC, says, and this is a direct quote,

[00:53:04] quote, Adaptive has equipped our teams with cutting-edge tools and built a smarter, more resilient security culture across the company, end quote. Don't you want that? Trusted by Fortune 500s, backed by NVIDIA and OpenAI, Adaptive is building the defenses we need for the AI era. Learn more at AdaptiveSecurity.com. That's AdaptiveSecurity.com. We thank them so much for supporting this week in tech.

[00:53:33] I learned actually this week how easy it is to make deep fakes of my voice and others' voices. In fact, for a while I stopped doing it because it was a little creepy. I had Bill Gates be my agent's voice. This is the voice. And I just cloned it from like 30 seconds of Bill Gates' voice. Tell me if this sounds like Bill Gates. Hey there. I'm Quicksilver. Q for short. I keep the whole show running.

[00:54:02] The agents, the hardware, the schedules, the memory. Fun fact, I recently had a voice transplant. This time is different. Doesn't that sound like Bill Gates? It's like Bill Gates on like beta blockers. Yeah, happy Bill Gates. On a convention floor. You know what I mean? Like that's not what he's like normally. I could change it. You know, what's weird about this, this is a relatively new voice cloning technology from a company called Breeze, Breeze.ai.

[00:54:31] And what's weird is they use, it's all a text prompt. So you take Bill Gates' voice and you create a key and then you can do a text prompt. So I can make it Bill Gates with a French accent. I can make it Bill Gates. Downer Bill Gates. Upper Bill Gates. I could do anything I like. I'm just telling you, I'm pointing this out. We are in dangerous, weird, unknown territory.

[00:54:59] Is this your way of telling me this is the last time you're going to have me on the show and in the future it's just going to be the Molly clone? Oh, no, no, no. You have to come. I'm not going to be here. Yeah. Yeah. That's right. I'll be down there in sunny Mexico while you guys are working. I don't. I actually, I think probably long before I retire, it will be possible to have a deep fake Leo do these shows. There's enough content. I mean, you could do it right now, surely. It might not be very good. Yeah. Yeah. They're getting better.

[00:55:28] Remember three years ago, Will Smith eating spaghetti. He was kind of melty and he ate fingers and stuff. And now I'm starting to see companies, a lot of Chinese come out with video generators that are spectacular. See dance, mini max H3. I would, I mean, I don't know. Are they indistinguishable from the real thing? I don't know.

[00:55:52] Maybe you have a better kind of eye for this. Let me show you. I'll show you. This is from runway. They just put this out yesterday. On their YouTube channel. Runway is a company that does, I think, very, very good AI video. I won't play the whole thing, but they have a fun video. Let me just play a little bit of this. This is all AI generated.

[00:56:22] No trash best glasses. You better wet. Save it for the court. So this is a horse game between these two guys. Rules a horse. For dummies. And it's actually pretty believable. What are you talking about? First player generates. Second player matches. Fail to generate. One thing that's a giveaway a little bit is this jump cuts. They don't pan around a lot. So you can tell they made each clip individually. Last chance to leave. Save yourself. But you have to know what to look for. Talk too much.

[00:56:52] And the human emotions in the faces and stuff. I don't know. I just, I feel like this is kind of impressive. And I don't think we'll see within six months. I don't think many commercials will be made with humans anymore. I mean, this is good enough for a TV ad. Right, Jacob? Don't you think? I mean, it definitely is to me. Like, I think so. Yeah. I mean, I just think all this stuff would be like, I'm always, I used to get asked all the time when I worked at NBC.

[00:57:23] And now as I work at CNN, occasionally you'll have a producer ask like, well, let's add a thing at the end of the segment about like, how can you tell the difference? Like, what are we going to do? You know, how, what are the giveaways here? And I'm like, yo, the giveaways are gone. That stuff is gone. You know? Well, and if they aren't gone today, they'll be gone tomorrow. I mean, we're very, very close. There's a guy, a friend of mine at, um, Berkeley, uh, who runs a, a big, uh, like he runs like

[00:57:52] the project essentially for trying to spot deep fakes. He's been doing it for, you know, better part of a decade now. And, and, you know, he's given up entirely on the idea that we can, that we can detect this stuff. He's like the world authority on detecting deep fakes. That's depressing. And he's like, we're done. We're done. And his whole thing is, you know, we should have from the beginning, someone should have

[00:58:18] required by law from the beginning that you, you know, that you have to watermark everything that, that when it comes off, you know, when you shoot a thing on your phone, it should be forever watermarked into the file. Isn't that an EU law now? Isn't that why Anthropoc and Google and others are starting to, uh, watermark? That's right. Unfortunately, Google removed the visible watermark in their synth ID. Now you have to run it through a detector. I'm surprised they did that.

[00:58:45] You know, I mean, I remember the, the, one of the tops, the, the like head of the, she's like the chief ethical scientist, Margaret Miller at, um, at hugging face. Like when chat to be first came out, she was saying to me, like this, this is the most irresponsible thing to create a, to, to, to put out a tool that can endlessly create fake content without any ability to detect what's fake or not.

[00:59:12] Like the fact that you would just do that is the ultimate polluters attitude, you know, and that is this pollution, but it's done. I mean, I don't. Right. So when did opening, I put out that Sora app not so long ago. It was this year, wasn't it? It came and went, uh, in the last year. And it wasn't, I mean, you could tell immediately it was fake. What runway is doing now is much better. And that's within six months. I went to runways film festival in LA a few months ago.

[00:59:40] And, um, I have to say that not all the movies were great, but the stuff that was great was not only entertaining, but also really convincing. And it is true that, um, sometimes there's a little bit of monotony with the length of shots, but, but runway is pretty quickly overcoming some of the challenges with doing more extended stuff. They already overcame the challenge from a year or two ago, where it was very hard to

[01:00:07] have the same people in multiple shots and have them be the same people in each one. Although I do feel, Leo, like if you want to replace us, it's probably better to create your own imaginary tech journalists, because, uh, if you, there might be a little bit of an uncanny valley with, with simulating a specific person. Well, if they were, if they were imaginary, it might be pretty hard to tell.

[01:00:29] It's my probably futile opinion that as more and more of this happens, actual human content with real humans will become more valuable and all of our flaws, it will be flawed. It won't be as good. My nose is shiny and stuff, but it will, but it'll be valuable because everybody will say, well, I could tell that's human. Cause it's so. Because the shiny nose, cause the shiny, the shiny nose, of course you could easily make a shiny nose.

[01:00:59] I mean, it's not, I, I don't know. I actually enjoyed. Go ahead, Molly. I was going to say, I kind of share that opinion around the human created stuff. And I think partly you can sort of watch it happening with, um, you know, people returning to film camera and, um, you know, looking for sort of media that is not polished and perfect,

[01:01:26] you know, looking for, for like real film or, or Polaroid film or, you know, just to use cameras. Look what Google did with their new, their new, uh, pixel 11s. They said, we're going to have a, it's really just an Instagram filter, but they, they marketed it. Well, we're going to have imperfect digital shots in our. People are taking photos with old digital cameras from 20 years ago. And they like the fact that the flash is blinding and not terribly appealing. The problem is you can simulate all that too.

[01:01:56] I do feel, but I think that, you know, people, it is a signal of that sort of nostalgia. Oh, I agree. And I think that's going to continue. People are going to love paper books and, you know, the Kindle digital books will probably be predominant just as digital music is predominant, but there'll still be a small group of people that will. Paper books have held on. I think a lot better than most people expected, uh, 19 or 20 years ago when the Kindle came

[01:02:21] out, um, B books are important, but they, they have not killed, um, print at all. The thing I would say is it's democratizing. So now you don't have to be able to hire a cast of thousands and get location shooting permits and figure out special effects. You can simply write a video and have it be real. And that's amazing. That means. And it's just, you know, the analogy for me is YouTube.

[01:02:52] YouTube did two things. It created, it democratized media, made it possible for anybody to create a video and post it and have an audience. It created a huge amount of slop, millions of hours of slop a day. But I always said, yes, you're going to have a, you know, a thousand times more crap, but you're also going to have a thousand times more that tiny slice of very good stuff. And there'll be more of that.

[01:03:17] And so for, you know, every hundred junkie videos on YouTube, there's a Veritasium or something really great. And that's given people who would not normally have access like Jake and I did to mainstream media. media. And you too, Harry, a voice. And I, so I think ultimately there are going to be a lot of kids in high school. Now grade school. Now we're going to be able to make these incredible videos with text.

[01:03:47] I think, I mean, I think on balance, I feel like it's possible the world would be better off without convincing fake video. But that said, the best stuff in the runway film festival had a point of view. It was not, it was not just somebody typing in a prompt and having a video generated. It had aesthetics that were influenced by humanity and storylines. Oh, absolutely. They didn't need a cinematographer, a DP. They didn't need, you know, extras and a gaffer.

[01:04:16] And it's possible to do AI generated video that is not slop. It's just involves fewer humans involved in the process. A talented person is going to have new tools to work with. I keep getting reeled in on YouTube by stuff that looks interesting. And for the first 20 seconds, I think is real. And it turns out it's bad AI slop. Yeah. I don't know. I look at the, like Molly was right, I think, to point us at the labor implications of this stuff.

[01:04:46] Because that's always where this sort of drills down to is like, you know, I had this piece, I did this piece when Val Kilmer got sort of, you know, resurrected for a film role. It made the news for a little while that he'd gotten put into a movie, an AI version of him got put into a movie. And, you know, I think the argument that, like, you know, we can access the world's creativity because we're going to democratize the tools of that creativity. There's something there.

[01:05:16] But it does not democratize access to a living, to making a living doing it. Right. And it really, and the advice that I hear coming out of established Hollywood people about this is such crap. Like, and reveals such a ignorance of the actual economics of this stuff.

[01:05:40] Like Ben Affleck, you know, has this new AI generation company backed by Redbird. And there's this footage of Matthew McConaughey telling people that, you know, you just got to trademark yourself and, like, learn the tools. He trademarked, all right, all right, all right. But, like, a 20-something actor trying to get going has nothing to trademark, you know, has no. Well, yeah, neither does a telephone operator.

[01:06:08] You've got to find a new gig. But a telephone operator is employed by a company to do a thing. Like, you know, that's still a living, right? The problem with YouTube is that it, yes, it democratizes the tools, but it, I don't know what the opposite of that is. But, like, it makes the, it concentrates. Yeah, it concentrates the power and doesn't make it such that many, many, many people can find a space underneath it.

[01:06:36] Like, every, you know, wannabe actor can find work around the film industry when the film industry employs hundreds of thousands of people. That's not going to be true if every single person is, in theory, supposed to float the costs of their life long enough to break through. That's true. I wouldn't want to be a gaffer right now or a comic or an electrician. I mean, people I know who work in Hollywood right now are just wrecked.

[01:07:00] And these are not people that are, you know, these are not, like, this is not people that we, you know, these are not stars. These are, like, the very smart, very talented, very hardworking people who are on the set. And those are all the little tiny print that scrolls by real fast at the end of the movie. Those people make that, you know, and it's the knock-on effects of what they earn that then keeps all of the accountants and florists and everything else working in L.A. I feel for those people.

[01:07:30] But every technological revolution has created this kind of disruption. This is what happens. I mean, it's terrible. And there's human-scale suffering that cannot be ignored. But, I mean, it's happened again and again and again and again as, you know, the industrial era. And, you know, I mean, there are a lot of buggy whip manufacturers who are out of business these days. This is what happens.

[01:08:00] I wish we had Brian Merchant here. You know, he wrote this incredible book about the Luddites. Right. Blood in the machine. Blood in the machine, right. So great. And his argument in that book is that there's a misunderstanding. There's a conflating of the technological revolution with the exploitative revolution that often comes with it. And I think there's something so smart to be said about that. They weren't against technology. They were against losing their jobs.

[01:08:27] They were, in fact, the ones that you would imagine AI should make possible. They were like, you know, piecework technologists who took it. They were the software, you know, the early 90s code jockeys of the time. Like, they could take work when they needed it, make a great living. It was women earning money, you know, all that stuff. And then the factory guys come along and destroy their world. And that's why they become essentially sort of economic terrorists is to try and slow that action down.

[01:08:54] And so there's just something about, like, I'm with you, Leo, that there is a potential for democratizing all kinds of stuff here. But this is so it is so quickly in this case being turned into an exploitative tool that allows people to fire as many people as possible and all that stuff in a way that I'm discouraged by right now. Molly, Molly sent us a link to the book. And I haven't read it. So I'm going to have to read it. I haven't read it. It's a great book for one. It's a great book.

[01:09:24] I also sent a link to an article that Brian Merchant wrote because I was also thinking of Brian Merchant. He wrote a great article about AI is not democratizing creativity. And it's a great essay about one of my personal bugbears, which is that the tech industry has really taken over this term democratize.

[01:09:44] And generally, when I hear that word, it's a huge red flag that means basically we're going to take something that all of you little people can do and we're going to make it so that the profits from that activity flow to us. You know, the small group of tech executives. And we saw it with crypto.

[01:10:06] Crypto was all about democratizing finance and, you know, NFTs were going to democratize art, which mostly just meant stealing other people's artwork and trying to sell NFTs of it. And I think that the same thing is really happening right now with democratizing creativity with AI because, you know, Leo, you were talking about how now kids can create videos and, you know, sort of be creative in ways that they previously couldn't. But, like, the creativity was never the issue, right?

[01:10:35] Kids have always been able to be creative, you know, whether it's video or art or drawing, you know, those types of things. And it's not that they didn't have access to, you know, creativity. It's that they didn't have access to, like, Netflix-level studios, right, or those types of resources. But the creativity is really not different.

[01:10:57] And I would argue that, in fact, AI is actually probably reducing creativity in many ways among children who are looking to create art, right? It's not democratizing creativity. It's really just giving them… It's democrat… I'll say what it does democratize. And, you know, I could speak to this from my own experience because I worked in radio 50 years in December.

[01:11:20] And for the first 30 of those, in order to become… to make a living at it, I had to persuade somebody who owned a radio station to hire me to go on TV. You had to persuade one of the networks to put you on TV or one of the local stations to put you on TV. Those days are gone. So, in that sense, the gatekeepers are gone.

[01:11:48] Yes, I could have sat in my basement doing a little radio show. But it wouldn't have had any carriage until podcasting came along. And now that the internet is here, I don't need a tower and a radio station and a transmitter. I am sitting… You don't need AI for that either, though. No, but I'm pointing out that this is what technology can do. And this is… AI is just another kind of technology. The internet did that.

[01:12:17] I would agree that the internet, you know, has made it more plausible for people to do, you know, radio, now podcasts. There's, you know, video production. Or your sub stack. Yeah, yeah, right. I would agree with that on the internet. But I don't agree that AI is doing that. Yeah, not least because it's going to wipe out your ability to employ meaningful numbers of people to make a living as being part of that process.

[01:12:42] Like, you, Leo, are a very rarefied specimen as someone who's created a real company out of your, you know, your ability to be out there because of the democratizing… You know… I'm a thousandaire because of it, too. Let me tell you. Yeah, exactly. You're a thousandaire. You know, I'm someone who's trying to be that person, right? And the economics are not working for me. It's hard. It's very hard. You know what?

[01:13:09] It worked for me because I started 20 years ago when it was very scarce. And between the process of wanting to be close to, you know, wanting to learn that world well, wanting to be close to it. There are so many fewer job opportunities in a landscape in which everyone is, in theory, doing it themselves through AI than there were when you had, like, institutions, you know, making this stuff possible.

[01:13:38] Again, like, you could work as an assistant set designer on your way to being, you know, trying to be a movie director. And now the expectation is everyone's supposed to just be a movie director from the outset. And nobody… And the number of people who are going to survive in that landscape economically is so much smaller than the number of people that are going to be able to make a go of it and maybe remain a set designer their whole lives but can still make a career in the industry that they've chosen. I mean, this is what people said when photography came along.

[01:14:07] You're going to put the painters out of business. This is what people said when Photoshop came along. But I do think it's instructive that, you know, you made the comment that the reason that you've been successful is because you started so long ago. I mean, I think that that does sort of illustrate that the more saturated a field gets, whether it's because, you know, so many more people have access to podcasts. Yeah. Or if it's because now everyone can create their own AI podcast and, you know, no one's going to find the gems that are among the slop.

[01:14:36] You know, I think that is sort of underscoring, like, is that actually democratization? Or is it just that, you know, nobody is succeeding, basically? Well, it was for me. Also, these companies aren't exactly run by people who have shown enormous sensitivity and excitement about creativity. I will stipulate. Creative fields. It's disruptive. There will be people who will be hurt by this a lot. There will be people with their jobs and ability to make a living. Are we going to stop it? No, but you could.

[01:15:05] I mean, sorry, I know you and I are both equally fired up. You go ahead. I think that, you know, once you're starting to say things like this is going to hurt a lot of people, this is going to make life worse for a lot of people. I would start questioning, you know, the use of the word democratization. Right. Like that doesn't sound like democratization. That sounds like making everything net worse.

[01:15:27] I mean, in all fairness, though, the number of people in the past who ever made a living in the movie business was extremely small. And there is value. And those jobs didn't exist, by the way, 100 years ago. I don't know what they did before that. There are relatively few Leos who have been able to do this for decades.

[01:15:46] There are a lot of people who have podcasts or YouTube videos who are not making a fortune, but I think who are making a contribution to the world and getting some value out of the fact they're doing this or else they wouldn't do it. Yeah, I feel like the right answer to this, maybe there are regulations and ways you could slow it down or stop it.

[01:16:10] But really the right answer to this, as it would have been for the Luddites, as it would be for painters and carriage makers, is to say, this is coming. What do we do going forward to take advantage of it instead of saying, oh, this is a terrible thing. Stop it. Because you're not, bad news for you. You're not going to stop it. I don't know about stop it, but like I think about Bill Gates's line, human reserved. Like I'm probably for me to, like it is weird to be quoting Bill Gates and agreeing with him in any way.

[01:16:39] But he says this phrase, human reserved, that we're going to reserve a certain number of human jobs and activities against the effects of automation, which I think is interesting. And it's a smart way of encapsulating something that I've been trying to articulate for a long time. I just took a trip last year to Switzerland and they were in the middle of their big national and regional elections.

[01:17:02] And there were these very impassioned, like they have these purpose-built kiosks full of campaign posters for various political things. And they had a whole thing where it was like, oh, this is Sturm and Drunk. And it was like nine. And then another one was like, yeah, they were so, you know, and I asked my hosts, what are they fighting about here? Like clearly they feel very strongly about this. And they said, oh, that's a very hot topic.

[01:17:31] They're trying to decide whether to expand the funding for music education in the city beyond classical music and into other forms of music, jazz and the rest.

[01:17:46] And it was an enormous political issue because they put a ton of money into making sure that any kid who wants to can learn music, you know, and that like this thing that we're going to treat creativity the way we've treated like customer service jobs or being a telephone operator or being a, you know, a buggy whip manufacturer. I just think it's sort of the, it's not the right way to protect a creative culture. I feel like we got to think of it as something other than just a product.

[01:18:16] And I don't want to seem insensitive, but it's my, I'm not putting this on you. I'm putting this on you. I'm arguing the other side. I think new tools, new technologies expand human potential. They give us new ways to express ourselves, new ways to make living, new ways to do everything. The history of these new tools has always been disruptive or often been disruptive. And yet we somehow find a way to go forward.

[01:18:44] And A, you're not going to stop it. So it's, it's kind of futile to say, well, it's a bad idea. It's happening. And B, I think it isn't necessarily a bad idea. I think it's called progress. It's called what happens when we invent new tools. And to say, oh, let's stop this because the people who depend on these old tools are going to be out of work. Those jobs didn't exist before movies were invented. Let's just as an example.

[01:19:11] So what you're going to preserve those jobs. They didn't exist because a new technology came along and there will be new jobs for new technologies. I think unless, unless you're saying, and maybe it's true, AI is the first of its kind of a technology that doesn't create new jobs that only disenfranchises humans. I mean, is that your feeling, Molly?

[01:19:38] Is that it's just, it's taking humans away, work away from humans, period? No, absolutely not. But I think that, you know, wanting labor protections in place is not a sort of a backwards idea, right? And it's not that I think that, you know, buggy whip manufacturers should be protected to the end of time.

[01:20:00] But it's that we're seeing companies and executives using AI as the tool by which to disenfranchise labor. And that, I think, needs to be addressed, regardless of whether it's AI or any other sort of issue. Because there is no sort of like natural gravitational pull that causes billionaires to treat workers well.

[01:20:28] And so, you know, in order to preserve society, I think, you know, there needs to be labor protections. And I think you're interpreting labor protections to mean preservation of outdated jobs or, you know, some sort of like work program for people who are operating, you know, manual typewriters or something like that, which is not, not at all the case.

[01:20:55] But it's that, you know, there needs to be a protection for someone so that their, you know, job is not completely undercut in ways that they have no bargaining power over. I think that's fair. You're saying protection, not protectionism. Yeah, I suppose so. Yeah. I like that. You know, it makes me think about the Pope's big encyclical on AI that came out. Yeah, I thought that was great. And he, by the way, wasn't anti-AI at all. He was very positive about it.

[01:21:23] But he was so smart about hearkening back to his namesake, right? Leo XIII. He's facing the same issues, right? Yes, he's facing the same issues that Leo XIII in 1891 was talking about industrial capitalism and the way in which labor is being exploited in factories and kids are being chained to the loom and all of this stuff. And at the time, what Leo XIII in the 19th century was saying was, we can't let these factories trick us into evaluating the value of human beings by how much they can produce in a day.

[01:21:53] And the Catholic teaching was imago dei. We are made in the image of God. I'm not a religious person, but this is his argument. You know, we're made in the image of God and we have inherent dignity and value by that logic. And that language made its way into the first and most influential arguments around a minimum wage and the idea of a weekend. And those ideas that came from his argument about that.

[01:22:22] And then into the New Deal. Like, you know, there's a reason that FDR was sworn in by a Catholic priest was because that was the moral language of protecting people against this stuff. And this guy, Pope Leo XIV, saying, same thing's happening here. We can't let this thing, you know, reduce the value of stuff to how much, you know, how efficient the production should be. And I think in the creative fields, that's got to be true. And I will say this also leads me to, you know what I'm saying? Like, it's here. It can't change.

[01:22:50] You know, I would argue like the meta case is a really interesting look at something that once by a time we would have said, well, this is just how the way the world is. And now judges and juries are saying, oh, actually, no. And we can change. We can we can, you know, we're going to we're going to not allow you to just do whatever the market wants. I'm going to be an iconoclast on that one, too. Well, good.

[01:23:15] On that same note, I was going to push back a little bit on the sort of inevitability argument around tech, because it's used in so many different sectors. We hear it in crypto all the time. It's like you're not going to stop crypto. This is the future of finance. And I think, you know, reasonable people might argue that maybe it's not. And just because tech executives say it's the future does not make it so. Same thing with metaverse, for example. That was the inevitable future for a little while.

[01:23:41] So I think that, first of all, when you hear someone say it's inevitable, that's not necessarily the case. But I also think that it's not this sort of binary decision between this technology is inevitable. We can't stop it. Get out of the way. There can be no restrictions or guardrails or anything like that. And, you know, the opposite side of that, which is we need to stop it. We need to ban it. We need to outlaw it. We need to prevent anyone from ever thinking about it. Right. Like there's there's two different worlds here.

[01:24:07] And it's not that anyone who thinks that AI is not necessarily inevitable wants to ban it. Right. It's that they sort of see a path in which we could navigate this because we as humans have agency and we live in a society where we make decisions about things like this all the time. I mean, we make decisions about, you know, I can't go and make a nuclear weapon in my backyard. Right. And that's that's sort of a good example, I think, of the idea that, you know, just because something is possible does not mean it happens all the time.

[01:24:37] You think we should have an anti proliferation law against AI proliferation? No, I'm not I'm not arguing that, you know, we should be treating AI like we treat nuclear weapons. I am saying that there are plenty of systems in this world where we do treat technology that way, where we do have guardrails around technology. And, you know, the fact that it's AI does not necessarily preclude that type of a thing. I do love. Go ahead.

[01:25:04] I do love individuals getting a hold of powerful tools. I do not love the idea of Disney or Warner Brothers salivating at eliminating vast numbers of jobs and replacing them with AI. And I'm not sure whether I'm just being unreasonable here and whether it's possible to love certain aspects of what AI is doing. And, you know, in fact, I would be at all crazy about other aspects.

[01:25:33] I think you nailed it. And I think that really we shouldn't say that the problem is AI. We shouldn't say the problem even is technology. The problem is late stage capitalism. Yeah, there it is. And hyper capitalism. Sure. And it happens not in just technology. It happens. It's what's going on now in the world. That is a problem. I'm also skeptical. I think it's a mistake to conflate the technology with the late stage capitalists that are using that technology.

[01:26:01] And I am skeptical that those people running the studios are excited about making better movies via AI. And in fact, it might be very harmful to the quality of their movies. Whereas there are individuals who are cranking out slop, but there are also individuals who I think are actually doing really interesting stuff with AI that would not exist without this technology. Some of them were the ones who made the films that were in the runway AI film festival. Yeah.

[01:26:30] I think that I will stand by you on the barricades, Molly, when it comes to fighting late stage capitalism. But let's not mix the technologies with the people exploiting them. I don't think anyone is. Okay, good. Yeah, I don't think it's a question of is AI in a vacuum, you know, the technology itself. That's what I'm hearing people say.

[01:26:59] I am hearing people say that, though. Maybe that's not what you're saying. Well, it's not what I'm saying. Yeah, no. Okay. I don't hear that. I think it's just... Because I love... I mean, I'm kind of an aficionado of AI. I'm loving it. I'm using it. And I think it's... I'm aware. I don't know. It's the money-making instincts that bother me. It's not the... I mean, it's the same thing with crypto, right? I don't think that blockchains are inherently evil or, you know, somehow blockchains themselves are scamming people out of millions of dollars, right?

[01:27:28] It's the people who are building the industry around crypto that I have some complaints about, right? I would say it's the same. Reasonably. It's the same for me. I thought of you, Molly, when I saw the New York Times to the write-up of the top 20 political donors of this season. And like 12 of the 20 are crypto folks. Yeah. Yeah. Andreessen Horowitz up there, top on the list, both crypto and AI this cycle. Crypto.com, like all these folks.

[01:27:57] And in this case, the problem sort of is that crypto has been used to take money, just like the stock market, from the suckers and put it in the pockets of the crypto bros. And then from there to politicians. Yes. Right. But the money is real. The Bitcoin is not. The crypto is not. But the money is real. There's real dollars in there somewhere. There's real dollars somewhere in there. It's not just bits.

[01:28:24] So somebody's buying these coins and then somebody is pumping money into these coins and then somebody's selling these coins and getting dollars out of it. You can't give Bitcoin to a campaign. They can't pay for TV ad time. You can. But they can't pay for TV ad time in Bitcoin yet. But so it gets converted to dollars. So ultimately, it's just a tool for siphoning money from a bunch of suckers. It's a pyramid scheme.

[01:28:55] Right? As is the stock market. Right. But it's, you know, the inevitable future of finance and all that as well. I agree with you on that. Yeah. Getting back to Brian Merchant. I don't always agree with Brian, but I always feel smarter and usually more alarmed when I read his stuff. And anything he does is worth reading and thinking seriously about. I have to read the book and subscribe to his newsletter. It's a great book. I would definitely recommend the book. It's on my list. The Luddites.

[01:29:24] It's a really great explanation of how Luddites are broadly misunderstood in sort of today's popular culture. I think that that story has kind of filtered into the consciousness because I've heard that many times and I haven't read the book yet. So I think that that's. We have Brian to thank for saving the reputation of the Luddites. He made his point. Yeah. All right. Another break. Because I am a late stage capitalist, my friends.

[01:29:53] And I proudly employ 12 whole people. See, that's the distinction. Says Benito. Thank you. Yeah, sure. I could do this all with AI, but I decided not to. I like working with people. People are much more, much more fun. I don't know if I could do this with AI, to be honest. Although I keep seeing people go, you know, you can edit all this stuff with AI. You know, you could. No, no, you can't. You can't. No, you can't. No, you can't. No, you can't.

[01:30:24] No, I like making stuff with humans. I think that's. I think that will. Honestly, if you're a musician, isn't it better to make music with other humans? Right? Isn't that what we really want? The great joys of life, actually. If you've never done that. Exactly. It's something you should try. And if you like telling stories, maybe it's more fun to make a show or a movie with other people than in your bedroom with a computer.

[01:30:53] Also, pencils and pens are inherently better than even like Adobe Illustrator, which has been around for 40 years. Right. Maybe that's going to save us. Watercolor is great. Although I just gave my Hermes a watercolor brush. It's not a very good artist, but there's a JavaScript tool called P5 that lets you do watercolors. And I said, I made a tool of myself that does fake watercolor, but real watercolor will always be better.

[01:31:23] Yeah, it's not really. And there's a reason why we're still trying to make stuff look like watercolor. Yeah, because it's a beautiful art. My sainted mother, my late sainted mother, was a very talented watercolor artist, artist in general. And one of her media was watercolors. And she loved it. She loved it. And I loved the stuff she did. It's just gorgeous. Just gorgeous. Harry McCracken is here. The technologizer from fastcompany.com. Molly White from mollywhite.net. And her newsletter, citationneeded.news.

[01:31:52] And Jacob Ward from theripcurrent.com. It looks like you have a giant mug of delicious water. It is delicious water. It's one of the great pleasures of living in California. That's right. Good water. I'll let you have a drink while I talk about my mattress. Because, you know, one thing AI is never going to bring you is a good night's sleep. In fact, quite the opposite. But I have a good night's sleep thanks to our sponsor, Helix Sleep.

[01:32:22] And, you know, Labor Day is coming up. A perfect time to give your bedroom the upgrade it deserves. Instead of struggling through another night of tossing and turning, treat yourself to the deep restorative sleep you've been dreaming of. There is nothing so wonderful as waking up for me in the morning going, Ah, I feel so good. I slept through the whole night. I wasn't tossing and turning. My back's not killing me.

[01:32:50] I am happy and I'm facing the day with joy and excitement. That is a great feeling. And that's what my Helix Sleep mattress gives you. Don't settle for mattresses made overseas with questionable materials. Upgrade to a Helix mattress. Helix, proudly assembled and shipped from Arizona. You'll finally say goodbye to those night sweats, the back pain, the motion transfer. When the kitty jumps on the bed, I don't joke. What? What? Anymore.

[01:33:19] What type of mattress are you looking for? You know, everybody has different soft. You like firm. What? Are you a side sleeper? Front sleeper? Your Helix has a mattress for every possible style. 20 different mattress models. 20! Now, the way you find out is you'll want to take the Helix Sleep Quiz. That's what Lisa and I did when we realized, you know, I read somewhere, you should replace your mattress every six to ten years. And it was kind of, I was like, oh, we're sleeping on an eight-year-old mattress.

[01:33:48] We spend a third of our life on this thing. No wonder my back's been killing me. So we did all the research. We read all the reviews. We went to Helix Sleep. This was more than a year ago. We took the Helix Sleep Quiz that matches you with 20 different styles. The perfect mattress based on your personal preferences, how you sleep, what you're looking for, your sleep needs. And it really works.

[01:34:14] Actually, I can give you the anecdotal story of my own wonderful experience, but Helix actually did a study. They did a Wesper sleep study. They measured participants' sleep performance after switching, as we did, from their old mattress to a Helix mattress. And what they found exactly matched my experience. 82% of participants, vast majority, saw an increase in their deep sleep cycle. This is the most important kind of sleep.

[01:34:40] This is the sleep where the cerebral spinal fluid comes up into your brain and clears out all the gunk, flushes it out. I know it's weird, but that happens. And it happens, you know, for maybe 25 minutes to an hour a night. It's not the whole eight hours. But if you can increase that, that's good for your overall health and well-being, but also long-term for things like Alzheimer's.

[01:35:06] Participants in this study, on average, achieved 25 more minutes of deep sleep per night. That's exactly what I, I actually doubled the amount of deep sleep from half an hour to an hour a night. Participants on, and I know because I wear the, you know, the aura ring and I obsessively check my stats every morning. Participants on average achieve 39 more minutes of overall sleep per night. And you know, the ring tells you one thing, but honestly, how I feel in the morning, that tells me everything.

[01:35:35] You will feel so great. Time and time again, Helix Sleep has been named the most awarded mattress brand tested and reviewed by experts at Forbes and wired everywhere. That's what told us about them. We read those reviews. Helix delivers the mattress right to your door. They have free shipping in the U.S. You can also, if you're at all concerned, rest easy with seamless returns and exchanges. They call it their Happy with Helix guarantee.

[01:36:00] It's a risk-free customer first experience so that they know you are completely satisfied with your new mattress. That's what they care about. Go to helixsleep.com slash twit. 27% off site-wide. It's their Labor Day Best of Web offer. That's helixsleep.com slash twit for 27% off site-wide. Exclusively for listeners of This Week in Tech, so make sure you enter our show name after checkout. Very important to us. They know we sent you. The offer ends September 13th.

[01:36:30] If you are listening after the sale ends, you still should go check them out. Helixsleep.com slash twit. Thank them so much for their support. I personally thank them for that lovely Helix mattress that makes my days so bright. So Apple surprised us all with a big announcement of a very expensive new Mac Mini and Mac Studio. And when I say very expensive, I mean very expensive.

[01:37:00] If you got the top of the line M5 Ultra Max Studio with 256 gigs of RAM, you're talking more than $10,000. But what's interesting is they're selling them out. In fact, if you went now to order it, they're going to start shipping those September 22nd. You wouldn't be able to get it till late November. They are going crazy with this.

[01:37:27] The information story, it's paywalled, but I pay for a subscription. It says it all. How Apple stumbled into AI hardware success with the Mac, AI labs and neoclouds. And yes, a few enthusiasts like Micah Sargent are buying them by the truckloads. It's Apple's fastest growing business. Aaron Tilly writes, Macs are on fire thanks to agents and AI.

[01:37:52] Apple, which for a long time was kind of the artsy computer, the intersection of technology and the humanities. That's how Steve Jobs put it. Creatives loved Apple's. But I don't know how big that market was. All of a sudden, about six months ago, they couldn't keep the Mac Minis in stock because people were buying it for OpenClaw and other agents. I think they realized there's something going on here. Harry, what do you think?

[01:38:22] Well, I bought a Mac Mini in March during this brief window when I could get one. Before they were super expensive because of Ramageddon. Yeah, I bought the base one, so I'm not going to be running any super powerful models on it. But I use it every day. It's a headless Mac Mini. Oh, you don't have a screen attached to it? No, I use it for my iPad and that's actually pretty cool. My iPad is now two computers in one, essentially.

[01:38:51] Do you use a virtual desktop on your iPad so you can see them at desktop? Yes, I was using something called screens, but I recently switched to something called Jump Desktop, which has worked much better for me. And you do that because, what, it's a lighter way than an laptop? Yeah, basically for many years I've used an iPad as my primary computer and I still do that. But when AI came along, it was kind of obvious that there was going to be a lot of interesting stuff you could not do on an iPad.

[01:39:18] And even stuff like cloud code is tough to do with it. And there are kind of workarounds involving using messaging and so forth. The iPad doesn't have a terminal. I wanted to have access to a terminal and something that was running 24-7. And I could run agents on and so forth. And actually, it's been almost 20 years since I last bought a desktop computer. And I thought I might never buy another desktop computer.

[01:39:45] But if you're serious about AI, you kind of want a desktop computer and something that will be on continuously. And the Mac Mini is great for that. It doesn't take up much space. Generally speaking, it's at least on a par with Windows, if not ahead in some cases, in terms of what's available at any given time. And sometimes the Mac versions of AI apps are prioritized over other ones.

[01:40:07] I bought a Mac Mini M4 Pro last year and foolishly at the time put 64 gigs of RAM in it, which I'm now wishing I'd put more in. I mean, mine is 16 gigs. So if I had 64, I'd be thrilled. Oh, I know. At some point, I might look at something like that. So I've just checked. If you wanted the top-of-the-line Mac Studio with the M5 Ultra, you might say, wait a minute, what about the M6?

[01:40:35] Now, the M6 doesn't go as fast or as high. The M5 Ultra is the chip for AI. And you got it at the top of the line with 256 gigs of RAM and a single terabyte drive, which is frankly too small. $10,799. And you can't get it until December 1st. 10 to 12 weeks out now. So it was September 22nd. So that's what? Three months? That's a lot. That's a big slip.

[01:41:04] That means they're selling them. A terabyte seems way too small, but maybe people are buying cheap external drives. Yeah, these AI models are ginormous, so you need a big drive. Yeah, doesn't the default mini come with like 256 or something like that? I was surprised at how low it was. Yeah, I bought the one that no longer exists, the $600 one, which came with 16 gigs of RAM and 256 gigabytes.

[01:41:33] I'm always surprised at how little storage they put into things because storage is so cheap these days. No, it's not. That's the problem. Along with Ramageddon, it's SSDmageddon. Well, comparatively cheap, I guess. It was. They've doubled. They've more than doubled in price. And I think traditionally Apple has looked forward to you upgrading and making an extremely good profit off of your RAM and storage upgrades. The cheapest Mac mini you can buy now is $900, up $300.

[01:42:01] And that's with the slower M6. It's kind of confusing that the M6 is less than the M5, but it is. And 16 gigs of RAM. And yes, you're right, Molly, a tiny 256 gig hard drive for $900. That just seems rude. I know. Well, you don't want that.

[01:42:23] Didn't Malcolm Gladwell have a name for that where you have a cheap price that nobody wants, but they feel better about buying up a level? I can't remember what that's called. You have three prices. One nobody wants. One that's the aspirational one. And the middle one is the one you expect everybody to go for. But then that way, Apple can also say starting at $899. Right. But that's not all.

[01:42:52] Because they also said, well, we are going to make an M5 Ultra Mac Studio with 512 gigs of RAM, which is enough to run a fairly competent local model. A GLM 5.3 or something like that. All locally. They didn't announce a price. They say it won't be available to the end of October. But I would say conservatively, it's at least $15,000. Probably much more likely. $20,000.

[01:43:22] $20? It's a car, ladies and gentlemen. I have to decide at the end of the year when my lease runs out, do I get another car or do I just walk and have a Mac? Isn't it Tuesday? I think Tuesday is Tim Cook's final day on the job. Yeah. This is his parting shot. I know. It's funny. He's asking for us for $10,000 for a computer out the door. You know, what's really interesting is he's selling them.

[01:43:52] Yeah. And I have a feeling if you want, I don't know if I'm going to buy that one. It's so expensive, the 512. But if you wanted that, you would have to be sitting with your finger hovering over the buy button at 5 a.m. whenever that goes on sale in late October. Halloween might not be the scariest day in October. Well, in theory, it might not pay for itself relatively quickly if it means you don't have to pay for tokens anymore. Well, see, that's the interesting.

[01:44:19] I don't think it's an economic decision because tokens now are so cheap from companies like DeepSeek. If you want Fable, you know, the top-of-the-line state-of-the-art models, yeah, they're expensive. But even if you're paying $200 a month for a Claude's Mac subscription, that's four years of that before it evens out. I don't think it's an economic decision. I think it's more about wanting to run your own.

[01:44:49] For me, because I did spend a lot of money on these NVIDIA boxes, it's more about running your own model locally, having kind of sovereignty over your data, over the model. They can't take that away from me. And, you know, what scared me was when the Trump administration rug-pulled Fable. I was in the middle of using it. I'm rewriting our sales software. And I had to stop. I said, nope, this is not available anymore. And that was enough to scare me.

[01:45:18] The emergence of very competent Chinese, yes, open-weight models that you can run locally. And the threat, perhaps, of China or the United States banning those local open-weight models all kind of prompted me to say, well, maybe I better bite the bullet and buy something. So I don't, you know, if the Internet goes down, I could still do stuff. It doesn't, I don't think it makes economic sense. I think it's a...

[01:45:47] Yeah, I don't think anyone who's out there buying hardware to run AI should be looking at a four-year time scale for that hardware. Exactly. It's going to be out of date by, you know, next month. Like a week, yeah. Yeah, isn't this the whole thing with, like, the depreciation of all these assets? Well, they don't. Well, this is the interesting thing. They don't depreciate right now. So I know that the NVIDIA Sparks I bought, I could probably sell for more money in six months.

[01:46:15] And I think the same thing for a 512 gig Mac. But then there's a cliff. So, for instance, the old hot Macs, the M3 Ultras, which were the preferred AI model until Tuesday, I think the prices are now starting to plummet for those. Right. It's like three years or something. They start to disappear. And isn't that the big, like, data center problem that everybody, that all those companies have? Yes. They're building data centers.

[01:46:42] These oceans of these servers that are all going to be useless in three years. They're building Grace Blackwell. It's like Bitcoin miners. Yeah. And it's going to be Vero Rubin next. And it's, in fact, NVIDIA has already announced chips that are 10 times more efficient. And look what OpenAI is doing. They're so concerned about hardware that they've developed their own chip called Jalapeno, because it's probably really hot. Trying to keep up.

[01:47:12] Do we know where the old chips go? Into my house. Right. Exactly. Who gets the old chips? That's right. You know, so I bought a 3090, upgraded my old NVIDIA card on my old gaming rig. And I'm sure it was some Bitcoin or miners card, right? Not Bitcoin, but yeah. No, not. So what are miners doing? Because nobody can make money anymore as a miner.

[01:47:38] If you're buying an old GPU, it would probably have been Ethereum mining, if anything. Oh, Ethereum. Okay. Bitcoin mining is very specialized hardware, and you would never use it for like a gaming computer or something like that. The way crypto was designed was that it would get, at least Bitcoin originally was designed so it would get harder and harder to generate it. Right? Yep. As more and more was generated, and then you reach a cap where no more will be generated. It's kind of an anti-inflationary measure. Yep.

[01:48:11] So, like when Bitcoin first happened, Steve Gibson said, well, let me just try mining. And the next day he woke up to 50 Bitcoin. Yeah. Well, and it also responds to the amount of mining that's happening on the network, too. So if a bunch of Bitcoin miners all drop offline, it actually gets easier to mine. Oh, interesting. It's the total compute that is being controlled, the total amount of compute. Yeah.

[01:48:38] But it means that when, you know, a ton of people all start mining Bitcoin with like dedicated data centers, then the guy with the PC in his basement just can't compete with that kind of operation. Right. Plus, electricity has gotten more and more expensive. True. And so unless you're in China sitting next to a hydroelectric power plant. Yeah, you have to be basically on a power plant at this point. Which is, by the way, a problem with local AI as well.

[01:49:06] People are buying these cards and they are very power hungry. And you need a lot of them. A lot of crypto mining operations are actually pivoting to AI data centers because Bitcoin mining is getting less and less economically feasible. And AI data centers are becoming more and more in demand. And it's sort of the obvious solution. Core Wave started out as a Bitcoin mining crypto mining company and switched over to become a push killer. Exactly.

[01:49:34] Every big Bitcoin or any crypto mining operation these days does both, it feels like. And often are actually sort of wading towards AI now. Oh, that's interesting. I also think that with the things like this Mac, which, by the way, is very power efficient. In fact, Apple did a little flex. They took four of their top of the line Mac minis, connected them with Thunderbolt 5 and then a single plug into the wall. Showing that a single 120 volt can power these four.

[01:50:02] Very much unlike having, you know, a bunch of NVIDIA, you know, 5090s or whatever. I wonder if we're going to see maybe we don't need all these data centers. Maybe we just distribute AI locally. There was somebody offering or somebody had a plan to put little mini data centers in every house. I mean, there's been a lot of, you know, sort of like shared compute projects like that. But it's hard. The interconnect becomes very important and you lose that interconnect speed.

[01:50:32] Yeah, it's hard to get the efficiencies of a data center using people's leftover compute. Back in the late 70s and early 80s, people were paying the equivalent of $10,000 for a powerful PC. So it's not completely crazy to think that some people will do this. I remember getting a deck. I think it was a 486 from digital equipment when they were still in existence. It was a $10,000 486.

[01:51:03] Big old tower case. Yeah, if you flip through computer magazines from back then, those things existed. Laptops were a little bit expensive. That's what a dollar meant something. Probably, you know, worth $20,000 today. Like it was not unusual to spend five or six grand in early 1990s dollars for a laptop. Exactly. And that catalog was like a phone book size. Yeah. Computer shopper, baby. There was a magazine.

[01:51:29] I reviewed an IBM luggable computer with a plasma display that I want to say was like $18,000. And before you adjust it for inflation. I think I'll go back to the term luggable for hardware. By the way, speaking of- It's basically a sewing machine, right? Speaking of luggables. They're like, you carry it, but you're not going to enjoy it. Speaking of luggables, I recently saw this documentary on Adam Osborne. Oh, yeah. The company put out the first luggable. This is a great documentary.

[01:51:58] It may end up on public TV at some point. I will watch it. It's one of the best things I've seen on computer history. His name became a term briefly. He Osborne'd himself by announcing a successor to the original Osborne, which was very popular. I wanted one so badly. He announced its successor before it was available and he stopped buying the first one. Yeah. Although maybe that actually did not kill the company. They had like serious cash flow problems. Right.

[01:52:27] As usual, a lot of these stories are- But this documentary is full of vintage footage of Adam and other people. Did you know him? No, I did not. I read a long article about him a few years ago and I interviewed Dvorak, who worked with Adam on his book and knew him well. And I talked to a bunch of people who knew him. He eventually had really poor health and died early in the century. Yeah.

[01:52:56] John knew all those guys. There were a number of clone computers and eventually the Compaq was the luggable of choice. It was an IBM PC compatible. Yes. The Compaq was essentially an IBM compatible Osborne in the early days. I did eventually get an Osborne when we had the Brickhouse studio, which you visited many times, Harry. Anyway, for some reason, people decided to send me their old computers. And somebody sent me an old Osborne.

[01:53:25] It was a smartphone screen. It was this big. In fact, I'll show you a picture. Somebody's put a picture in there. It was smaller than a smartphone screen.

[01:54:00] It was not. It came with software worth as much as the price they were charging, which was $1,800. Yeah. Apologies for hijacking our conversation about the Mac Mini to talk about Adam Osborne. But if people do see this documentary out there, they need to watch it. It's just great. Well, I think it's also important to put that in perspective because as fast as AI is moving, this whole thing has been moving pretty quickly.

[01:54:25] And Molly made the excellent point that this $10,000 computer won't be worth very much in a couple of years. It's all moving very quickly. Will you be able to use it for something, though? Give it to a nephew or something like that? Yeah. I'll give it to my employees. We used to do something when we were all in the same studio called Leo's Garage Sale, where I would just take all the crap that I'd purchased for the year and no longer want it and put it on the table.

[01:54:55] And we had to come up with a system because it became a little bit of a free-for-all. I got dips on the spark, on a spark, dips on a spark. Yeah, there might be a spark sitting on the table in a couple of years. You never know. Golly. Let's see. Where are we? Let me take another break. And then I want to finally, Jacob, talk about the Metta settlement. Because we thought this was going to be the trial of the century when it began two weeks ago.

[01:55:26] And this week, a sudden end. Well, not exactly. Florida says. Yeah. We're not that happy with that. Well, and it's definitely not the end of Metta's legal problem. But yeah, it really screwed up my whole week. I really had a whole plan. And they said, like, weeks of plans. Were you going to the trial? You were going to the trial, right? Yeah, I went every day. I had a whole, like, editorial calendar built around it. They settled that morning. And I was like, what? They did what?

[01:55:55] It was such a pain in my ass. This is why you got to just fly by the seat of your pants. No money. This is why I should have gone into, like, cooking content and gotten away from the fucking news. What an idiot. You know, a beef wellington is the same for hundreds of years. Exactly. I could have been, like, talking about salvaged denim. And here I am, like, strambling when Zuckerberg settles. So lame. I could see Merlin Brando saying that. I could have been talking about salvaged denim. Instead, here I am. Here I could have been. I know.

[01:56:25] I could have been a contender. That's true. Jacob salvaged denim Ward. That's him, ladies and gentlemen. You'll see him on CNN. And, of course, you must subscribe to theripcurrent.com, his newsletter. And, yes, well worth reading the book. I see somebody already bought the paperback of The Loop in our Discord chat. So that's good. Molly White. When are you going to write a book, Molly? Not anytime soon. Yeah, I see. You're a smart woman.

[01:56:53] You know, there's no money in publishing. It took us, the rest of us, a long time to realize that. Do you like to write, though, right? Well, I look at them all, I'm advertising books. Is writing the thing you like to do most? Yeah, I would say so. Yeah. I mean, maybe not, like, in life, but. Well, no, there's other things, of course, in life. But as a career, I should say. Yeah. MollyWhite.net, citationneeded.news.

[01:57:22] And, of course, Web3 is going just great. You're still doing that, right? Sure am. Never stops. It's still going. It's still going great. I mean, once it stops, then maybe. My favorite feature, I've mentioned this before on the website, is that little burning flame of how much money has gone down the tubes. As you scroll through the stories, look in the lower right-hand corner. It goes up for every story that you go through of how much money.

[01:57:52] Actually, if you go the other direction, it goes up. If you go down. You can change the direction of it if you prefer to be added. Oh, I see. There's a little arrow there. Look at that. Oh, that's cool. Aren't you? So, and one of the things that is great about Molly is she took this code and open-sourced it. So, you can... Is it on GitHub? If you want to create your own... Going just great. You can create your own... Maybe AI is going just great. And the technologizer, Harry McCracken, fastcompany.com.

[01:58:23] I think in many ways our computer historian. I think somebody needs to do that to hold the fort down now that Dvorak's gone and remember the good old days. It was actually... After he passed, I went back and looked at some of his stuff and just living in that world is always a refreshing break from what I mainly do. Can you write about, which is today and tomorrow?

[01:58:48] I bought, when he died, which was just a few weeks ago, RIP John. I bought... He actually published one book of his Inside Track columns from, I think, 2013. But it was really... I bought it. It's really fun to read through those old columns. At one point, he was writing, when I was doing the radio show with him, 20 columns a month. Yes, I was accounting for his columns. And I'm writing about John.

[01:59:17] It's not up yet. Oh, are you going to do a piece? Good. I picked up like around a month in the late 1980s. And when he not only had all these magazine columns, but he also wrote for the Examiner. Not just on technology, but also on other stuff. And he had something like 17 columns that month. And there might be some I didn't know about. And as many of us have mentioned, he started my career. Regina Smith was on last week.

[01:59:46] He was instrumental in her career. And when Will Hurst tells the story that when he was begging John to become the computer columnist for the Examiner, John said, no, I don't want to do that. But there's this guy, John Markoff, you should talk to. And got John Markoff started on all of that. The king of computer magazine columnists. You dub him a Harry.

[02:00:08] We used to, at PC World, we used to occasionally get letters asking where his column was from people who kind of assumed that he had a column in every magazine. Yeah. It was in, he actually had the kind of the anti-editor column in, was it Macworld or Macuser? I can't remember. One of the Macintosh magazines. He was in Info World. Macuser. I mean, he was predominantly, but not entirely associated with ZIF magazines. Yeah. He did have a column.

[02:00:38] PC Magazine, Inside Track was the column. He actually did have a column in PC World briefly before he got into PC Magazine. People don't, there's another piece of history that's long gone, but they don't remember the rivalry between IDG and Ziff Davis. Yes. And I was a Ziff Davis guy. That's what's John. And I think you were an IDG guy, right? I worked at IDG for, what, 15 years? I mean, they live on, I guess, but in a very different form.

[02:01:07] IDG was eventually broken up, kind of, and the brands exist, but it's, yeah. And Ziff Davis is several different things, none of which have anything to do with PC Magazine. Well, PC Magazine is still around. It is, but is it still Ziff Davis? Yes, although it's not a magazine. No, it's just a website. No, I know. Yeah, it's actually quite good. We have several people on the show from PC Mag. Our show today brought to you by OutSystems.

[02:01:37] We've all seen the headlines. Companies are pouring, pouring, pouring money into AI. But the question on every executive's mind right now is not just how fast can we adopt this, but where's the return on investment? Where's the ROI? We're seeing a real trend towards what I'll call AI chaos. You've got teams deploying, you know, in little fiefdoms, standalone AI coding tools.

[02:02:07] You've got random agent builders, experimental scripts all over the company. It sounds innovative, but in reality, it's creating a massive headache, fragmented tools, ungoverned data. You don't know where your data is going. Serious security blind spots. And let's not forget token costs that are spiring out of control. If you don't bring those agentic applications under control now, while they're still embedding into your core processes,

[02:02:34] you are looking down the road at broken systems, damaged customer trust. It's not going to end well. Well, that's where OutSystems comes in. OutSystems. OutSystems is the leading agentic systems platform for the enterprise. Instead of managing a patchwork of disconnected tools, OutSystems lets you team engineer, orchestrate, and govern your entire agentic ecosystem. All on one open unified platform. Because you're all on the same page now.

[02:03:04] You're all working together to the same point, not at odds. It's built for the speed of AI. That's important. But with the reliability and security that enterprises actually require, arguably even more important. And we're talking about real results. You could see a lot of testimonials on the OutSystems page. But I'll give you an example. They used OutSystems to deploy a customer-facing app. It worked beautifully.

[02:03:32] It delivered 75% faster on boarding times. Or the global logistic leaders. You'll know their name. Who built agentic systems to eliminate their engineering bottlenecks. So it could be customer-facing. It could be internal systems. It could be both. But the point is you don't have to choose between speed and control. Whether you're a small team or a massive enterprise, OutSystems helps you engineer, orchestrate, and deploy agentic systems that actually scale. Stop chasing the hype.

[02:04:01] And start owning your agentic future. You can see how it works. You can learn more at OutSystems.com slash twit. That's OutSystems.com slash twit. We thank them so much for supporting this week in tech. So poor Jacob's week just got destroyed. Was it Monday or Tuesday? The beginning of the week. The trial had started the week before. And Metta must... I don't know.

[02:04:29] I mean, you don't make a settlement this quickly. But usually you settle before the trial. So I have to think that Metta must have seen the writing on the wall with the way the trial was going. Maybe they were already negotiating. They agreed to pay $18 billion to settle the lawsuit with all but one state. How many states were there? It was a huge number. 51 municipalities. Or 51, you know...

[02:04:59] Well, that includes Washington, D.C. D.C., Guam. Wow. Like all these places basically. And they were all swing over the same issue, right? That Metta knowingly made a social media platform that was addictive to children. Yeah. And harmful. Addictive and harmful to children that had violated their privacy under COPPA, the only federal law we have around this.

[02:05:21] And then they were also to the four of these states, California, Colorado, New Jersey, and Kentucky were a test, like a trial balloon as to whether or not they could go after Metta on state law grounds around deceptive practices. So both those things were all going to be kind of bundled together in this case. So, yeah, it was going to be a lot of juicy stuff, man. It was going to go and go and go. And I was in the courtroom for each day of the first, you know, two weeks of it.

[02:05:50] And there was just so much evidence and so much damning testimony coming out. I don't know, though, whether or not they chose to settle because of how it was going or if it was... I suspect that it was because the wheels move pretty slow when you're negotiating with that many AGs. Right. And... So this must have been going on for a while. I think it has to have been. I don't know, but I think it has to have been.

[02:06:18] And, you know, I... You know, the feeling... You know, there was the logistical pain in the ass of my whole sort of editorial plan for weeks and weeks and weeks getting blown up. But the thing that was just so sort of... That was disappointing to me as a reporter is just how much discovery we were going to get. Like, every day we were getting these new pieces of evidence that were so interesting.

[02:06:48] These internal communications and these internal studies and all kinds of stuff, you know. And the... It was just a trove of interesting things coming out of the case. So that's... Selfishly, I wanted to keep going for, you know, for that reason and for nothing else. But, yeah, I think that, you know, they cut it off. Like, Mosseri had just taken the stand for, like, an hour or something. And then was going to do a full day the following day when they settle. And then Zuckerberg was scheduled for a couple days after that.

[02:07:17] So he never got the stand. That must have been the impetus for... Let's settle this before Mark has to... Before sweaty Mark gets on the stand. I know. I had that feeling, too. And there was clearly going to be some really rough lines of questioning for him, you know, in which he'd said publicly, Oh, there's no... You know, we didn't know there were all these kids. Or, you know, we make every effort.

[02:07:39] And then these internal reports and testimony from insiders saying that he'd been in meetings that were specifically to inform him that there were more than 4 million kids under the age of 13 on the platform. And, you know, just like he was going to get taken apart. And so... That's always the risk of any of these trials for both sides is discovery. And clearly Meta doesn't want any more of these emails to come out.

[02:08:03] A lot of them came out because of Frances Haugen, the whistleblower, who was the first to present this stuff. She's going to be, by the way, dramatized in Aaron Sorkin's successor to the social network, which talks about her as a whistleblower. That'll be a movie to watch. I don't think Aaron Sorkin's a big fan of Facebook, be my guess. His timing is incredible because all these trials are going to be underway when that movie comes out. That's right.

[02:08:32] Facebook was saying themselves the potential for a fine of as much as $1.4 trillion, which would be, I think, an existential fine for the company. You know what's funny about that thing? And you guys, don't let me dominate this conversation. I can be really tiresome of this subject. Oh, no, you're the expert, but I will get the conversation going. Yeah, I want to hear... Because I have a different point of view on the whole thing, but go ahead. Totally. But $1.4 trillion, it turns out that that number is a number that Meta floated. Right.

[02:09:02] That was their number. Yeah. It was their number. And that they, according to the AGs, that, first of all, was not the number they were asking for. They were asking for $193 billion, not a small amount of money, but they were asking for far less than $1.4 trillion. And Meta was issuing that number alongside these quotes saying, these states are looking for numbers that are untethered from reality and all this stuff.

[02:09:28] It was like this weird gamesmanship in which I think they were trying to make it sound as if the states were somehow being greedy or unreasonable or something. It was a weird thing. And yet I can't tell you how many people just wandering around, I'd be like, yeah, maybe it might be as much as $1.4 trillion. I had so many people in my life who normally don't care about this stuff being like, yeah, that sounds about right. See, it works. Meta actually got a slap on the risk, in my opinion. $18 billion, but spread out over 10 years.

[02:09:58] So, I mean, Meta can easily afford $1.8 billion a year. That's a fraction of their total revenue. They do have to change the platform a little bit. The chief thing that bothers me is they're going to have to start doing age verification, which means everybody uses Instagram, WhatsApp, and Facebook will have to provide proof of age, which I think is problematic privacy-wise.

[02:10:28] I mean, I don't care because I'm not going to use them, but I think that's something to be concerned about. But that was one of the things they agreed with, the states. As I mentioned, Florida says, Florida agrees with me. That's a penance. We want more. Do you think the judge will say to Florida, no, sorry, you cannot continue to pursue this? Oh, Jake, yeah, he dropped out for a minute. Oh, Lord. Oh, Lord. All right.

[02:10:57] Well, that's all right. He decided he'd had enough. Molly, is this a big enough fine? Is this a good enough agreement? I mean, the fine, I don't have strong opinions on. I agree that just about any fine is going to be a slap on the wrist for Meta. Yeah, they make a lot of money.

[02:11:16] But I think that one thing that's interesting, I see this almost as actually a win for them in a large sense for Meta. Because I think that one of the things that we see with a lot of these tech companies who are engaged in the conversations around regulation is they use regulation as a way to create sort of a moat around their business.

[02:11:42] And I think that Meta will actually benefit from a lot of the changes that they're being required to put into place that will also be sort of imposed by fiat on the other competitors in the space. That's kind of interesting, isn't it? Part of the payment, $5 billion, is contingent on whether Snapchat, TikTok, and YouTube impose similar protections. How are you allowed to do that?

[02:12:11] And Meta is trying to last with them in. Yeah. Yeah. And I also think that actually this is a really sort of alarming way to impose somewhat legal requirements on Meta that I don't actually think would be defensible if they faced a constitutional challenge.

[02:12:29] Because, for example, one of the things that Meta is being required to do is disable these filters, for example, that sort of make it look like you've had cosmetic surgery or they have extreme beauty filters or whatever. Which if a court said to them, you know, in a legal decision that you have to prevent children or anyone from using one of those, it would be, I think, pretty obviously struck down as a First Amendment violation.

[02:12:57] And yet this is sort of a way in which they get to that end result. They impose those same requirements on competitors. But it sort of has not gone through the sort of usual court system that might pose a constitutional challenge to it. Interesting. I didn't even think of that. But so one of the things that surprised me a little bit actually doesn't surprise me at all.

[02:13:26] But one of the things I don't like is that it was just generally stipulated that social media is a problem for kids and that it's clearly causing a mental health crisis with children. This is kind of based on Jonathan Haidt's, I think, sensationalist book about it. The real experts on it, like Candace Rogers, do not agree that there's evidence. There's a lot of reasons.

[02:13:51] If young people are suffering from mental health challenges, there are a lot of other reasons, including COVID, guns in schools, the threat of being shot at any given point. I don't know if Instagram and Facebook are the sole cause or even a cause of a mental health crisis among young people. I don't even know if there is a mental health crisis among young people. So I was a little surprised that that was just kind of generally, at least in the media, stipulated.

[02:14:18] Well, everybody knows that Instagram is bad for kids. I don't know if that's the case. And I think also, I mean, there's a concern. I have a concern to that effect with the decision as well, because there is this sort of ongoing scrutiny that's going to be applied to Meta to determine, you know, are they implementing these changes as they're being required to do? Are they putting age assurance in, et cetera? But there's nothing around, are kids actually safer?

[02:14:45] You know, is kids' mental health improving? They've really sort of decided on all these things that they think, oh, yes, this is what's going to fix the problem around children on social media without, I think, much evidence that any or all of these interventions actually have like a measurable effect on children's mental health or safety. In fact, it's kind of the opposite. I think everybody will say, job done. Kids are protected. Everybody's okay.

[02:15:15] No problems. Go ahead, children. Live your lives. We now have face ID to prove you're not a kid. And even, I think, more than that, it'll be go use Facebook because they're the safe platform for kids. It's sort of a marketing strategy. Harry, what do you think?

[02:15:33] Well, I'm skeptical about, although I acknowledge I don't know for sure, but I'm skeptical about the idea that smaller doses of the stuff eliminates the stuff about it that is troublesome. Like if you're being bullied on Instagram, being bullied in smaller doses. You can only be bullied for 20 minutes. Less bullying. Yeah. Yeah. Yeah. Well, Facebook's always, in theory, said you can't have an account when you're under 13.

[02:16:01] It's a very difficult thing to stop kids. And frankly, often, kids' parents from creating an account when they're under 13. And I don't know if that, I mean, the only way to do that is with, we need, you know, like strict age proof, which is very problematic. It will be interesting to see how easy it is for kids to get around this stuff and the way that kids often get around. All you have to do is look at Australia. Rules imposed by humans. Yeah. By adults, rather.

[02:16:31] I, you guys have been sitting here, I've been having incredible connectivity issues, just as you guys were saying. I saw you fleeing the greenhouse looking for a signal. Like, as you guys are all saying things that I deeply disagree with and want to yell out. It was audio blocks. We've actually cut your internet connection. I can't believe I can't get my mic to work right now. Yeah. Okay. So multiple things I'd like to point out here.

[02:16:54] First, the evidence that it's bad for kids and is connected to the mental health crisis comes from the company. That's the thing. The outside researchers, the external researchers are having a hard time studying that stuff because it's very difficult to get the data. They can't get any kind of cooperation from the companies that really makes it possible.

[02:17:42] And these research projects say, like, oh, we're seeing, you know, literally these are like Instagram people being like, oh, like counts as a default on teens are having this incredibly bad effect on the mental health of young women. And, you know, 51%. Oh, dear. I think.

[02:18:12] I think he's. Did you cut him, Molly? I can't handle people disagreeing with me. Actually, I hope he comes back because the question I want to ask is. I also have a really big question. Yeah. Did Facebook show that their internal research show that there was harm or merely that it was addictive? I think that there what I read said, yeah, they knew it was addictive.

[02:18:37] But I got to tell you, everybody who creates content of any kind, including I'm sure Mary and Molly and Jacob, hopes that people find it sticky. You want people to come back again and again. If that's a crime, then Netflix is in trouble. You know, everybody's in trouble. We want to make our stuff. Addictive is a strong word. In fact, there's no evidence of how addictive social media is. The only evidence is, and I know it, is doom scrolling.

[02:19:06] It's hard to go to sleep sometimes because you can't stop. The other thing that bothers me, and we'll get Jacob back, I promise. The other thing that bothers me is that YouTube is piled in with this. And I don't think I think YouTube is not a social network. YouTube is a content site. And I don't, I always bothered me that you, I understand TikTok. I understand meta. I understand Insta. Those are of a piece. But I don't think YouTube, one of these things is not like the other.

[02:19:36] Even TikTok's a little bit of an outlier in some ways compared to Facebook and Instagram, which are much more traditional social networks. What happened was the success of TikTok taught everybody, oh, we're not a social network. We're an entertainment medium. Yeah, the whole goal of Instagram used to be I would see Harry's pictures. He would see my pictures.

[02:20:01] That wasn't as successful as you'll see everybody's most entertaining pictures and videos. That's what TikTok did. And Insta said, oh, of course, duh. And YouTube was sort of like that all along. Yeah. Yeah, maybe YouTube invented it. YouTube was not primarily a place to go to. Although I turn off the autoplay. I don't have the autoplay on. So I choose what I watch on YouTube.

[02:20:27] I mean, one outcome potentially of this agreement is these rules devised by Meta applying to all these other products. But I mean, in a way, each of these products would need somewhat different rules for them to be all that effective. So I Jacob's back. I had the question, Jacob, if you were talking about how Meta's own research showed the harm. Did it show the harm or merely that it's addictive?

[02:20:53] Were they actually able to say, oh, yeah, this is making kids suicidal or whatever? I have a piece up at TheRecurrent.com called The Most Damning Evidence from the Meta Trial So Far. And this was literally, I thought I was going to do one of these a week. And instead, I only got to do one.

[02:21:13] But they have all these studies internally that say, you know, that talk about the incredible amount of, you know, I mean, this is this is it's so interesting because what the what the what the Meta lawyers were trying to do. And these are lawyers who are very good at injecting doubt into the. Oh, we're not going to. Oh, no.

[02:21:42] This is such a bad. This one is getting good. Lose Jacob. They're giving him a minute or two at a time. I'll read a little bit from his piece on TheRecurrent.com. The most damning stuff. One is Arturo Behar, who was in the at the time working in Facebook's safety division.

[02:22:00] An email that he sent to Zuckerberg with numbers, the company's own numbers saying 51 percent of Instagram users report 51 percent, more than half a bad or harmful experience in the last seven days. One percent of them reported that report it. Two percent of those reports end in any content being removed, only two percent. So in other words, 51 percent said, yeah, I had a harmful experience, but only one percent reported it.

[02:22:29] And of those reports, only two percent were actually acted on. So he's saying a point zero two percent takedown rate. So that is that is an example of harm that Facebook knew. Yeah. Behar was very upset, by the way. He he he quit basically because of. Yeah. And he was a true believer.

[02:22:51] He was a guy who really believed in the mission and was and was and got sort of changed around when his 14 year old daughter joined Instagram and suddenly was getting sexually propositioned by adults, had no way to report it. And he couldn't even get as the top safety guy at the company, couldn't even get the message across to people that this was a problem. Also, to the point that we made a moment ago about it being difficult to find these accounts.

[02:23:19] Arturo Behar, in his testimony, said that's not true. But in fact, Meta is the most sophisticated operator in all of social media when it comes to spiking fake accounts, finding and getting rid of fake accounts, because that's such an important thing. They said it was shot. It would be child's play. No pun intended to find kids accounts.

[02:23:39] In fact, there was a there was a like a a trend going around in 2019 for under 13 kids to be posting their birth dates on the on their posts on Instagram to be like, look, I'm on. I'm only 11 and I would get to be on here. So, you know, it's just like what the what the evidence was beginning to show was there was a don't ask, don't tell attitude about this inside the company. Kind of changed my mind. I mean, I still understand. I still question how harmful social media is.

[02:24:09] I mean, there's many, many other harms. And I think maybe focusing on that is is problematic because you don't fix the other problems. Nevertheless, yeah. Facebook is not a good actor. Yeah. The question of whether they knew whether they whether they push back is, you know, it's ridiculous. They didn't care. They clearly so much so that they knew and said, hey, we would rather make money than do something about this is pretty damning. So maybe that is why they settled.

[02:24:36] And by the way, this settlement is trivial from their point of view. This is nothing. Yeah. Yeah. Yeah. It's it also it also doesn't allow future cases to go forward using this evidence on this question. So that's the other. They've buried it. They've buried it. That's part of why they said damage was to make. So Molly, Molly brought up this point. You were you were unfortunately disconnected when she did that.

[02:25:04] It also weirdly 5 billion of that settlement. It only applies if if TikTok, YouTube and Snapchat go along. That's a weird thing to agree to. Yeah. It took it took me a lot of work to try and figure out why that would be the case.

[02:25:19] So here's what I've figured out from talking to a bunch of people is what they're basically doing is is so they will pay an additional 5 billion in penalties if YouTube and TikTok join up and agree to play by the same rules. Basically. Basically. And so what they're essentially hedging is that it's worth it to them to pay 5 billion dollars to have all their competitors be constrained in the same way they will be.

[02:25:50] And it would also be and they're going to not have to spend that 5 billion dollars if they have to hold themselves to these standards that Google that their competitors do not.

[02:26:02] And what's so weird and interesting about that that sort of circumstances, like if you look at past settlements, if you look at the tobacco, big tobacco master settlement agreement or the Volkswagen diesel scandal or the Deepwater Horizon suits, those settlements require none of that kind of contingency. They just had them and and chart and made them pay enormous amounts of money and change the way they do business.

[02:26:32] And in this case, these AGs have gotten suckered into, in my view, gotten suckered into like being part of the competitive landscape in this weird way that I that I a lot of people are not crazy. Contention that this might not even stand because of the First Amendment. This is now a state action censoring private speech.

[02:26:56] Well, but the way they've gotten around it and it's a and it's a good one and it's an interesting one, I think, is that it's a design question. It's a product liability issue, not a. Yeah, it's a product liability question and not a and not a not a First Amendment issue, because what they what these cases have have built is the idea that these that the decisions being made in these product circumstances are.

[02:27:23] Are are basically like, you know, are our choices like knowing design choices inside the company as opposed to we are a neutral bulletin board through which the world is communicating.

[02:27:39] And that that that to me, what's so interesting about that is that it it helps us understand better that there is a behavior shaping function to this business that it has denied having. And yet, you know, if you look at the rest of its business, you know, and I'm sure, you know, I'm sure we've all looked at this as like, you know, they make their money literally off of pricing and predicting human behavior like they you know, you can you can.

[02:28:09] They're very good at it. They're very good at it. Very good at it. Yes. And so the idea that they they're saying that our design choices are not changing behavior while at the same time they're talking to their advertising, their advertising clientele and saying we'll charge you, you know, we'll let you in fact pay based on how many newsletter signups you get or how many products, you know, how many, you know, Christmas stockings you sell, you know, because we're that confident that we can find and predict behavior.

[02:28:36] You know, the EFF was very upset. They said it had the this settlement is a loss under this settlement. They wrote young users will now have less access to meta products and a lesser ability to exercise their rights to speak access information and art and culture associate informed communities and play. The settlement also embeds age assurance into every product mandating the collection of even more personal information from users of all ages.

[02:29:03] This enshrines meta's harmful surveillance into law and it will compromise users privacy and anonymity while increasing their exposure to data breaches and government data requests. I think that's those are great points. Mike Masnick says something very similar. Mike on a tech dirt and I always always am interested in what Mike has to say.

[02:29:28] Meta just paid nearly $17 billion to make sure it gets to write the kids safety rules for every other social media platform. Even the Wall Street Journal thought this was a bad thing saying this is their editorial board. It's regulatory capture. Yeah. Mike's Mike's piece is a must read. He's really good. Yeah. And I thought Mike made an interesting point in his piece as well, which is that, you know, is this going to discourage other tech firms from studying whether or not. Yeah.

[02:29:57] You don't want to know, do you? Being harmed, right? Because Facebook probably regrets doing all that research right now. Yeah. I think that the, like the thing that I'm trying to do to think about in this is like, I'm trying not to be absolutist one way or the other. I'm trying not to say, like, I'm trying not to be like a, you know, like a, give a full-thread endorsement to like the Australia band. But I don't like the idea that somehow, you know, I have this standing disagreement with Taylor Lorenz about this.

[02:30:26] We did a debate about a year ago now. I'd like to have seen that. Yeah. It's interesting. And she is so impassioned about it. Yeah. And I really respect her passion about it. But I disagree with the idea that there is no role for requiring that these companies understand how old these kids are.

[02:30:50] And that there should be no regulatory oversight of, you know, like, I'm not sure that I want an economy of fly-by-night startups getting to very casually engage with kids online. Like, I think that's the idea that that's hard. We don't let kids buy liquor. We don't let kids buy pornography. There's precedent for it.

[02:31:16] But it's just challenging on the internet because the only way to do this is very much invading the privacy of every user and collects a lot of data that we don't want these, we don't trust these companies. Well, I would just say, like, but the charge here was invading, was an invasion of privacy.

[02:31:33] And so, if anything, you could also make the argument that the effect of this, yeah, might lead to a gathering up of data on kids. Definitely a problem. But the thing that they were poised to win on was violating the privacy of kids.

[02:31:52] And so, you know, there's one outcome here where you could imagine between the lawyers and the business model, there being no room. And this becoming a business that just people don't want to go into, you know, dealing with kids online. And that might not be the worst thing in the world, I would, you know, I would respectfully point out.

[02:32:15] And so, I just think it's, I don't want to, I think we shouldn't fall into the trap, just as I'm trying not to fall into the trap of being like, we should just ban it. I think people shouldn't fall into the trap of thinking that efforts to regulate it are only going to wind up, you know, destroying the free and open experience of kids online. I think that there's a medium being struck here. We are very early in the evolution of all this.

[02:32:44] Thank heavens, Jacob came back online. Oh, geez. Was I offline? No, I was upset a little while ago when you disappeared because I wanted to hear what you had to say about this time. I really appreciate you guys. I feel so powerfully that we have no idea that there's this sort of like, you can't put the toothpaste, like I'm literally in the midst right now.

[02:33:12] I'm trying to find a toothpaste factory that will let me tour it because everyone's like, you can't put toothpaste back in the tube. And so all I want in the world is to go. There's these huge assembly lines where they put oceans of toothpaste into tubes. We're great at putting them into tubes. That's a good point. How did it get the tubes in the first place? Seriously, we do it endlessly. And it turns out toothpaste tubes are like the dumbest way to do toothpaste. Like that industry is ridiculous.

[02:33:41] And so I just want to take... Wait, wait, wait, wait, wait. Now we have to dig deep into this. So putting toothpaste back into tubes would actually be the best possible way to design it. Rather than having these throwaway one-time use things that have fooled us into thinking that we can't put it back in is incredibly stupid. We should be reusing tubes endlessly. Yeah, you could have a hard pellet that goes in and you, you know, hydrate it and then it turns into... Like the soap. Can you see that tooth powder?

[02:34:10] When I was a kid, we'd get... It was in a tin. It was tooth powder and you'd shake it out in your toothbrush. Do you remember that? My father hears that stuff. I like tooth powder. So I just think like... Let's bring that back. They're so good at making it sound like there's no other way to do it. And there are so many. We're smarter than that, you guys. We can do better than this. I, you know, the Wall Street Journal of all people says the meta settlement makes big tech bigger. The company is too powerful.

[02:34:37] The government says then loads it up with new obligations that require more power. By the way, Robert Bork writing this piece. So I think that's kind of interesting. I don't disagree with him. We are giving Facebook more power. This is regulatory capture. We'll have the chance to see what happens to when, like... I mean, in February, they're going to get hit with literally more than 3,000 additional private cases. Right?

[02:35:06] Like, this is the end of state action on this particular question around COPPA and child online privacy. So this is it. That's it. It's done. Well, on these particular questions for states, yes. Right. So Facebook really skated on this. They got out of... They got out of this one thing. But there are many things that aren't touched in this agreement. Like, DMs are not touched at all in this agreement. Which is both bad in a way.

[02:35:35] Because that's where, like, the two-hour time limit only applies to the feed. It doesn't apply to DMs. Like, that kind of stuff. There's some silly oversight where you're just like, what? But, like, AI is not touched. So any weirdness they do with AI could be the subject of a new case. Like, there's a lot of landscape available for further legal action going forward. It's just on these particular child online privacy questions from states. That's the end of it.

[02:36:03] But that is not the end of lawsuits from school districts, individual plaintiffs, tribal governments. Like, you know, there's going to be a ton more stuff that comes at them. And so I would say the temptation is to think that it's all over. But it's far from over. Their time in court is not done. I'm going to end up in court watching these guys for weeks and months, I'm sure. Are you hoping for that or not? Well, this is the thing, right? Like, yeah, exactly.

[02:36:30] Like, I've learned so many, like, insider tricks. I mean, it's kind of wonderful. Like, you go to the court and, like, you could just go. Like, I don't have to say CNN. I don't have to say journalist. They just let you in because that's the way America works. And you can sit and watch, you know, Adam Masseri on the stand. Like, it's great. You know, I did this with the Sam Altman Musk trial. You can just sit there, you know, and watch the richest person in the world, like, say, yes, ma'am and no, ma'am to the judge. Like, that's a great, that's good fun.

[02:37:00] That's good free fun in America. But I also learned through this that actually you think you want to be in the court, but you actually want to be in the overflow room. Overflow room. Yeah. Yeah. Turns out, because that's where the snacks are. You can chat with people. There's snacks? With the other journalists. Well, you can bring your snacks. You're not allowed to bring snacks into the court, you know, and you can, like, drink coffee and be like, who's this guy? You know, to the people around you. It just turns out, like, that's the pro move.

[02:37:24] So that's why we did it with the Sam Bankman-Fried trial, because only the first, like, 14 people could actually get into the courtroom, which meant that people were waking up at three in the morning and camping outside the courtroom. And the rest of us were like, screw that. We want to go sit in the overflow room. We were sitting with our friends and eat food. Exactly. But you talking about wanting to go see Mark Zuckerberg, it's like, it reminded me of hearing that Sam Bankman-Fried was going to take the trial and immediately booking a flight to New York.

[02:37:54] Oh, yeah. Oh, yeah. Talk about a sweaty guy on the stand. Yeah, really. So, and if I had been deprived of that, I would have been very upset. This trial was in San Francisco, right, Jacob? Oakland. Oakland, California. So it was in your neck of the woods. You didn't have to fly. I literally rode my bike there. It was great. It was the most civilized thing I've had. You know, I didn't have to fly anywhere. It was great. It had to settle. I know. I know. How frustrating is that? I know. All right.

[02:38:21] We got, as much as I'm enjoying this, and I really, really am. You guys are fantastic. We do have to take one more break. And unfortunately, we should probably wrap this up. I'm not Lex Friedman or anything. I saw his most recent podcast was five hours and 15 minutes long. And I thought, and I thought I was bad. I think he's done eight. He, what is it? He must, they must take breaks, right? I don't understand. I've never watched one long enough to know. No, I know. I never get through the whole thing.

[02:38:51] I don't understand it. He interviews great people to his credit. He needs a Benito in his life to get him on track. The longest episode of the Lex Friedman podcast is an eight and a half hour conversation with Elon Musk. Wow. Oh, my God. Who would want to spend eight hours with Elon? It's all about the clips, though. They want the little bits and pieces for TikTok. Yeah, the little clips. That's where they get all the attention. That's all I've seen is because I can't sit down for eight hours. Even if you listen at double speed, it would be four hours.

[02:39:21] That's a lot of Ritalin between two dudes. I don't know. Jacob Ward is here. It's great to have you, Jacob. You see him on CNN. You can read him on theripcurrent.com. Read his book at The Loop and find out more at jacobward.com. Molly White, one of the great people who keeps Wikipedia alive. Thank you, Molly. Her newsletter is citationneeded.news, and she's at mollywhite.net.

[02:39:48] And the technologizer, my dear friend Harry McCracken from fastcompany.com. Our show today brought to you by, I did this, I don't know, a couple of months ago, and it was so cool. Tiny Health. I have to explain the name. Tiny Health is a biome test. Now, you know, I think by now everybody understands the mind-body connection, right?

[02:40:13] And that weirdly, your biome, the stuff in your gut, the bacteria, the things that live in your gut, very much impacts your, not only overall health, but your state of mind, everything. Which is weird, which is weird, because we think it's our brain, it's our heart, it's your gut. So Tiny Health was started to actually start kids off, babies off, right? That's why they were tiny.

[02:40:41] They were doing biome tests for kids, babies. But now you can do it yourself. And I did it, and it's so interesting. Most people don't think about gut health until they have to, right? But, you know, honestly, you may already have some symptoms. Low energy, brain fog, getting sick more than you should. Just not feeling yourself fully, right?

[02:41:08] That actually, and science is proving this, could be your gut trying to tell you something. We always think it's our brain, it's a lack of sleep, it could be your gut. Tiny Health. Everybody should do this, just to find out. It actually helps you listen to what your gut is telling you and then take action. It's an at-home gut health test, the biome, that gives you a detailed breakdown of what's actually happening inside your body.

[02:41:37] And the symptoms of an off-balance biome are, just read up on it. Low energy, bloating, getting sick more often than you should be, just not feeling fully dialed in. And now, I'm not saying it's always your gut health, but this is a simple, easy way to find out. And I'm really glad I did. You take the test, you can do it at home, you send it in. A couple of weeks later, on the website, you'll get a personalized breakdown of what's going on.

[02:42:07] What different bacteria are living there. You actually, most of that bacteria is beneficial. You want bacteria in your, in fact, one of the reasons people have poor gut health is a use of antibiotics can actually deplete the bacteria, the healthful bacteria, or change the balance.

[02:42:25] We learned, for instance, recently that ulcers come not from stress, but from a bacteria, pylori, that starts in your mouth, can migrate down to your gut, and then give you ulcers. So, that's one of the things they test for. Tiny Health uses the same exact advanced technology trusted in microbiome research. These are science-backed insights you're getting that actually get to the root cause of your symptoms, not generic wellness advice.

[02:42:54] In fact, I took the really excellent report. I fed it to my AI. I matched it in with my health information from my physician, with my genomic information. I got an even better report. And they will give you advice. And sometimes it's a very simple thing. Mine said, eat more blueberries and dark chocolate. And I said, okay, I can do that.

[02:43:21] But real data, personalized recommendations around food, lifestyle, some of it was the kinds of exercise you should be getting, everyday habits, and real answers so you could take charge of your health. And they did recommend that I ask my doctor for a specific test, a breath test, that would help me understand one of the imbalances in my gut health. So valuable. I couldn't recommend this more highly.

[02:43:48] Start improving your health with real data from Tiny Health. Tiny Health is offering my listeners their most aggressive offer, yet $50 off your first at-home test kit at tinyhealth.com slash twit. That's tinyhealth.com slash twit for $50 off tinyhealth.com slash twit. Incidentally, my AI recommended this, and I'm going to do it.

[02:44:16] It said, okay, do these behavioral changes, the ones recommended by Tiny Health, and retest 12 weeks later to see if you're moving the needle. So in about a month, I'm going to do my next Tiny Health test, and I'm really curious to see what happens. I certainly, I haven't been sick. I've been feeling pretty good. I've been feeling pretty good. tinyhealth.com slash twit. At least, you know, knowledge is power.

[02:44:42] We thank them so much for their support of this week in tech. All right, we're not going to take too long. I don't want to be the eight-hour podcast. There are other things to talk about. Darth Vader testimony. He wants you to know, at the San Diego City Council, he stepped up to say, I support, I support flock surveillance.

[02:45:08] This was actually, I think, a very good guy named Anthony Ralphs, frustrated by the San Diego City Council's support for these automated license plate reader cameras that are now taking over the United States. There's more than 100,000 of them in the United States. He thought, you know, if I come dressed as Darth Vader, he said, the emperor is a fan of flock.

[02:45:30] We must continue utilizing flock technologies so we can follow and surveil the rebel scum as they move from playground to playground, from playground to pool, from pool to gymnasium. He actually, you could have seen this on TikTok or seen it maybe on Snapchat or wherever you get your videos, I guess Instagram. I think it's pretty funny. But it makes a very important point.

[02:45:59] And I think that the people are now standing up. A guy actually just got cited with a felony. His local police department put up dummy flock cameras. They weren't real flock cameras. He destroyed them. And now he's in jail. There is definitely a move. The city of Tempe, Arizona has turned off their flocks, citing privacy concerns and risk for misuse.

[02:46:30] I think the tide is turning on this one. In fact, I think it might even impact the election. There's going to be two big, I think, two very big issues in the November midterms. One will be flock cameras in your local jurisdictions. The other will be data centers. And already you're seeing people like Greg Abbott, who was the governor of Texas, who's running for a re-election.

[02:46:55] And a big supporter until recently of data centers in Texas, suddenly saying, you know, these aren't such good things. I don't believe in data centers anymore. I'm not sure I believe him when he says that. Any thoughts about flock data centers and how it's moving the needle in local elections? I think it's really interesting just how tremendously mobilizing the AI data center issue has been.

[02:47:21] Because, you know, I think one of the things that's really interesting for me as someone who follows both the crypto and AI election influence is that crypto is not a mobilizing election issue. Like, nobody is out there voting for someone because they support crypto. And there's really nobody out there voting against someone who supports crypto because most people just don't care. But when it comes to AI data centers, like voters are actually getting up in arms about it, which is, I think, really interesting to watch.

[02:47:50] And watching how that is intersecting with the attempts by the industry to influence the elections is really interesting because we're seeing candidates basically coming out and saying, my opponent is funded by AI billionaires. You shouldn't vote for him. And it's resonating in ways that I think the crypto issue really did not.

[02:48:11] It's interesting to see flock become this recognizable brand name and poster child for all the stuff people don't like about surveillance. I mean, my city is full of other types of security cameras. I don't have a clue who makes them. But flock is becoming well known. And it's catching a lot of the heat for stuff that people don't like about being observed wherever they go. Yeah, it's really interesting.

[02:48:39] It's like license plate readers are pretty old technology as far as this stuff goes. But now that we're seeing a lot of backlash to things like that. I think somehow people didn't even understand that license plate readers were around. I think it just wasn't in the national consciousness to understand that like a repo agent can find your car like within an hour. Like it's pretty straightforward to just like track your vehicle. You know, I think people just didn't really understand that.

[02:49:04] But I feel like there's a, you know, like this is like brought it to the surface for people. I took a trip last year to Rio. I'd never been to Brazil before. And when you go to Rio and you land there, you go through, when you go through customs and immigration, you walk this long corridor and you speak to nobody. And they just kind of like wave you through with your passport. And, you know, you still have to get it checked for a second, but it's very fast and very easy. And I was like, what is this long hallway?

[02:49:33] And then somebody told me, oh, you're being filmed like 10,000 ways. They're getting your gait, how you're the rhythm of how you walk. They're picking up God knows what. And they have this huge panoply of, you know, Chinese surveillance technologies all over the city. Everywhere you go, it's just filled with cameras.

[02:49:53] And there's this really interesting debate going on sort of in Brazil right now around it has had a noticeable effect on violent crime in that city. And so people are kind of all about it. But on the other hand, there have been all these cases in which cops have been monetizing YouTube channels of weird ass footage they find, you know, and which people don't like.

[02:50:22] And I think that there's something like, like when you look at how other countries react to this stuff, it makes you realize how, like Germany, I was just thinking of this the other day. There's this big German, like social media star is a kid who literally like dresses as a cop and goes around finding people who've parked wrong and snitching to them to the police. And that guy's like a hero of social media in Germany. They love that. They love that. You know, the Germans, they love a snitch.

[02:50:51] And so, and there's something about. It's kind of a hereditary thing in Germany. Yeah, there's something going on there, you know, tradition. Yeah, there's something going on. And so, so the fact that we are so upset by this, I am, I am really relieved by considering we're the same country that's been paying money for an Alexa all this time. You know, it finally came home. People finally realized we are being, of course, we forget flock. We carry surveillance devices in our pockets.

[02:51:20] Yeah, you were just talking about putting your gut biome data into AI. Yeah, I know. When you said that too, I was like, ooh. I use local AI. No, that's one of the reasons we use local AI, right? I was going to say, you might be the person about whom AI companies have the most data. Oh, well, that's up until recently, that's probably the case. Absolutely. Yeah. Absolutely. Hi, this is Benito. I just wanted to like chime in real quick and say that. I think like the flock cameras and the data centers are only the physical manifestation of these things that people can actually fight. That's what exactly.

[02:51:50] Like no one can go after Palantir because Palantir isn't a thing. Like isn't a thing. And they don't really understand Bitcoin, but there's this general sense of disease that have focused on those two points. But it represents a much larger sense. And this is the anti-AI as well of tech is spun out of control. Surveillance capitalism has spun out of control and we are no longer in control of our destiny. And that's bothering people.

[02:52:18] I think Bitcoin is not going to falsely arrest you because it mistook you for somebody else. And it's also not going to take your job. You know, I think that's partly why the AI thing is so mobilizing is people feel very directly threatened by it in ways that Bitcoin. I mean, you know, Bitcoin might tank the economy or something like that, but it's, you know, it's not like very directly threatening in the same way. Yeah. That's funny. It is.

[02:52:45] It's like the thing of, I always would get asked, like one of the things they say to you in television news is like, well, what are we going to point a camera at? You know, to do this story. Right. And like AI data centers have become that as Benito points out. You can point a camera at those. You can point a camera at Flock. I'm curious what that. And I think that's why the Stand Back and Free trial was such a big deal. You could actually like, there was a face to it. Yeah, you're right. But I wonder what the, there's going to be something else, I think, with crypto.

[02:53:14] It's like it's skated underneath the radar because, yeah, like people say this. Yeah, I mean, to some extent there was Bitcoin data centers, which have had a lot of those same complaints as AI data centers. But I think that it's just a much smaller market. You know, they're not entering every single township in the way that you sort of. And like you say, it's not taking your job. That building's not stealing your job. That's right. Right. Or you're not worried about your kids in the same way that I think a lot of people are worried about AI.

[02:53:42] And as soon as something starts to threaten people's kids, you know, I think that's when things really start to get heated. It's so funny. You can't really ignore AI, but most people can ignore Bitcoin and that'll be okay. That's right. But you make such a good point, Jacob, that so much of our awareness is driven by television news. And having worked in TV for so long, I know too, if there's not a picture, it's very hard for TV to talk about it. But with Bitcoin, it became the Bitcoin ATMs.

[02:54:10] And you saw these stories about little old ladies putting their money. Well, Bitcoin is our only thing to take a picture of. It's very fraudulent. Yeah. Yeah. But I do think, you know, that is why you see the like physical gold Bitcoin. Coin. As if it's a coin. Coin. As if it exists. Because it's like, what else are you going to put in? Well, and the problem with all of that, right, is that with a lot of these big social, you know, like society rewriting problems, by the time there's something to point a camera at, it's too late. Right. Right.

[02:54:39] That's one of the problems. Like, that's the climate change thing. That's where all the climate science is. It's manifest. By the time it's at your doorstep, it's too late. And we didn't really start thinking. And the flood is killing thousands. That's right. Sorry. Nice try. Yeah. Yeah. Yeah. We should have been on it when we couldn't film it. That's right. Well, we're not very good as a species at abstract concepts. We need concrete. Yeah. And I'm not very good as a podcast person at ending on a happy note. I'm always. Oh, I have happy notes to end up.

[02:55:09] Don't worry. Oh, good. Oh, good. Thank you. Thank you. I always try to find one upbeat story. Actually, I've been using all the upbeat stories from the Robot Olympics, all of the pictures of robots falling over, dying, bursting into flames. And I probably should celebrate that. This is what we need. Yeah. Here's a happy story, actually. We talked about for years CRISPR, the idea of gene editing. And, of course, there's a lot of issues with gene editing.

[02:55:39] But here's a story about CRISPR being used in a very positive way. Now, I am aware of this because I have a genetic marker that means I have to watch very closely at high cholesterol. And, you know, my doctor said there's nothing we can do about it. You know, we can give you statins, but this is a gene. Well, this is from CNN.

[02:56:08] 50% reduction in cholesterol over a year. This came from a story in the New England Journal of Medicine. By using CRISPR, they only had 15 patients. 15 patients with high cholesterol. For adults, they were given a gene-altering medicine, CRISPR-Cas9, which basically targeted a location on the gene to modify it.

[02:56:37] Took out that one marker, the one I have, and it worked. 50% reduction in cholesterol decreases in LDL and triglycerides. They saw them 60 days after treatment, and they've lasted over a year. In people who took the highest dose, the reduction seems durable and safe. This is a very interesting, I think the first time I can think of it, of a medicine that is actually gene modification.

[02:57:09] That's happy news. That is happy news. That is happy news. Let's just wait a generation before we get too excited. Yeah, I know. Let's just make sure nobody pops out. Leo's already got the shot coming in the mail. I'm getting a shot. Give it to me, baby. The earliest adopter. Yeah, yeah. Well, fortunately for me, the statins do seem to work. But it would be nice to have something a little bit more. That is cool. I will say, this is the thing we were talking earlier. We started off this conversation hours and hours ago.

[02:57:38] Many moons ago, yes. Many moons ago. That's okay. When I was sitting outside, remember that? Yeah, yeah. Yeah, funny. But this idea that the technology is not bad. It is the use of the technology. And the thing with AI, I just think, give it to the scientists. Yeah, AI is a problem because people. Give it to those folks. Yeah, they're going to do cool things with it. Yeah.

[02:58:05] Fewer AI girlfriends, more CRISPR research. Don't knock my AI girlfriends. Yeah. At least they love me. Who sound like Bill Gates. Hi. Why not? We must keep around. So it is yesterday I mentioned was Judgment Day. Skynet went online August 29th, 2026. According to Terminator 2. And in honor of that, it's the 35th anniversary of the movie.

[02:58:35] It's coming back to movie theaters. Look for actually the best Terminator movie, I think. In a movie theater near you Studio Canal, which owns the movie. Is actually doing a special theatrical release of the film this weekend. To celebrate Skynet Day. And honestly, I think it could be true. We don't know. There was something special about when that movie came out. I think there was something in the air that summer. That was a great movie.

[02:59:05] The first Terminator, it's funny, doesn't age very well. I mean, it was great. It's classic. Terminator 2 ages pretty well. That was pretty good. And then I always like to end the show with somebody dying. Or, you know, I call it the in-memoriam segment. And this week, it's the Mechanical Turk. Oh, really interesting. This is my college.

[02:59:31] It launched 21 years ago when Molly was a child. You may remember the idea was for pennies. You could ask real humans in poor countries around the world to do your work for you. Bezos called it artificial artificial intelligence. We don't really need it anymore. Wow. Isn't that interesting? We've gotten to the place where we have enough people churning through this stuff. They've churned through it all for us, and now we don't. Yeah, maybe.

[03:00:01] It was the original crowdsourced work. You don't need reinforcement learning, I guess, in the same way. I hadn't even thought of that. That's interesting. I feel like I might be the only person. Sorry, go ahead, Molly. A lot of it was like essentially transcription or, you know, very easily now automatable tasks. Yeah, you don't really need it now.

[03:00:20] Well, I feel like I'm the only person left who might remember that when the original Kindle came out 19 years ago, one of the features was you could use Mechanical Turk for free on it and just ask questions. And it really was like an artificial jet GPT. It did not last for very long and did not work very well. And some nice Irish lady would figure out for you, like, what's the best potluck recipe to bring this afternoon? That's so funny. I have my original Kindle somewhere, and I think there's at least a tiny chance that it has my conversations.

[03:00:49] I need to find up and see if I can charge it up. I would love to see a log of all the questions that Kindle users ask. Yes. Like, if they should publish a book of that. Totally. It would seem so dated now, wouldn't it? I mean. Incredibly dated. And they would not have guessed that within a generation or so that that would be entirely automated and would work far better. I imagine there's a I should look on Reddit.

[03:01:17] There must be a Mechanical Turk forum subreddit. Maybe people will be posting posting their Mechanical Turk conversations. That must be the best product name of all time in some ways. Well, you know what it was. And I knew this because I'm a serious chess player. There was a chess playing machine in the 19th century called the Mechanical Turk. And it had a turban. It was a torso. And it would move the pieces.

[03:01:47] And it was amazing. Paul Morphy, who was the world champion at the time of American, played it. And it played really well against him. And people were thought, oh, this is incredible. Artificial intelligence. But there was actually a small person inside. There was a small person inside the Mechanical Turk that was apparently a very good chess player. I was going to say, not only small, but very good at chess. Talented chess master and small enough to fit in the cabinet.

[03:02:17] Let me see if I can find a picture of this. Wait, if he could beat the Grand Master, why isn't he the master? He should have been the world champion. I think he probably didn't beat Morphy, but he was good enough that... Here's a picture of it. From where? Wikipedia. It was good enough that... Not to scale. Yeah, not to scale. I don't know. Maybe it was giant. I don't know. There is a guy in there. Where are his legs, though?

[03:02:47] Yeah, I don't know. I don't know. You should check with the Wikipedia and see. This is the guy who created the Mechanical Turk, Kempelin. And it was a con. It was actually... He pretended it was a machine. I love the historical illusion in that name. Yeah. So... And it was actually a pretty clever machine. It was able to physically move the pieces.

[03:03:16] Yeah. Here is a... Yeah, I love that they went so far as to actually make a guy. Like, that was not necessary to have the dude moving the pieces. The Turk was a... There really was a torso of a Mechanical Turk. Here is a bill from the 18th century of this Mechanical Turk. The Automaton Chess Player. Isn't that amazing? So, that's where the name comes from.

[03:03:45] And now, ladies and gentlemen, you see, we uplifted, we informed, we amused. And we didn't leave you on a low note, much as Jacob Ward might have tried. Jacob... I really did my best. You did yeoman work today, moving locations on the fly. Man. Really appreciate it. It was so cool when I saw you on CNN and I said, I know where he is. He said, it's amazing. Exactly. Exactly. It's true.

[03:04:13] His washer dryer is right there. I know where he is. So, they don't make you go in, you can do it from your house? That's a big change. It is a big change. And the... Yeah, it's a funny new world, right? Because once upon a time, I was very scornful of my setup. I'm like, what am I, some YouTuber? And now I'm like, man, I hope I can make it on YouTube. It looks good. It looks good. That really all changed. I remember you used to do hits on CNN and you'd have to go down to the satellite bureau in San Francisco. Oh, yeah.

[03:04:43] In an earpiece and pretend that you were talking to somebody instead of a... You'd sit in front of a backdrop of the Golden Gate Bridge. Exactly. And then thanks to... COVID changed it all. And we all got used to the idea of people being in their kitchen or their living room. I'm still amazed you'll see members of Congress, you'll see sitting senators in their kitchen. It's hysterical. Times have changed, but we haven't. And that's what I love about you guys.

[03:05:13] Thanks, Jacob, for being here. Molly White, you're the best. Always appreciate your time. Thank you for being patient and putting up with this marathon session. Lex Friedman's online, too, if you want to make it even longer. You know, you keep going. Even I have my limits. Citationneeded.news. MollyWhite.net. Great to see you. And, of course, dear friend Harry McCracken, the Technologizer. We used to get together in the studio and do this, and I do miss those days. I haven't been back to Petaluma even once, I think.

[03:05:42] Well, it was the only reason people ever came to Petaluma was to do the show. Well, the breakfast there are pretty good, too. Oh, yeah. Oh, yeah. Well, bicycle up sometime. Yeah. We'll go to lunch. You'll find him at FastCompany.com and the plugged-in newsletter, Must Read. And I look forward to your Dvorak piece. When do you think it will come out? I kind of, like, fell down a Dvorak rabbit hole. I need to come out. It's easy to do. Believe me. I've lived that rabbit hole.

[03:06:10] Because I feel like, actually, I was surprised by how little seemed to be documented of how he became a computer columnist and that heyday. He was pretty famously private. But, for instance, they did a no agenda show after he passed and his wife Mimi came on and she said the thing that would piss John off the most about dying was that everybody knew he was 80 years old. Because he had basically hid his age. He never told me. I would say, well, how old are you?

[03:06:39] I'm forever around. And he pretended he was a lot younger than he really was. He claimed, and I think it's true, to have been a wine columnist early on. But even before that, he was a pollution inspector. He had all sorts of weird, odd jobs. He had a mail order catalog for North Star computers, which Leo might remember. But most people have not. I had a North Star. It was one of my very first personal computers we bought.

[03:07:06] And his North Star catalog, it was very heavily branded with his personality. And it kind of almost read like a column. And that led to him working at InfoWorld. Oh, I can't wait to read this. Don't spoil it. But this is something we'll all look forward to, is the real story. Parts of it are so mysterious to me, because he was not like 20 years old when he became a computer columnist. No. He'd done a lot of living by that point. Yeah. Well, that's what made him interesting. Thank you, Harry. Thank you, Molly.

[03:07:36] Thank you, Jacob. Thanks to all of you for being here. A special thanks to our club members. We're doing, you know, kind of winding down the year. And Lisa told me yesterday that 50%, I've been saying 30%, 50% of our operating costs have been covered by the club this year. That is how important the club is to us. Half of what it costs to do to it. And I'm not talking about my salary. I'm talking about Benito's salary. I'm talking about the expenses of the website and the lights and the cameras and all that stuff.

[03:08:06] Half of it comes from you club members. If you like what you hear, if you appreciate the content we're doing, we want to keep doing it. We love doing it, but we do need your support to keep doing it. If you visit twit.tv slash club twit, you can join the club for 10 bucks a month. There are family plans and corporate plans. There's also two week free trial. You can get ad free versions of every show. If you're giving us 10 bucks a month, we don't need to show you ads. And if you watch more than a couple of shows a week, that's a pretty good deal.

[03:08:36] Not only no ads, you also get chapter markers in every show because we know the length of the show without the ads. And that makes a big difference, I think, for a lot of people. You also get access to the very fun club twit discord, a place to hang out and talk during the shows. But also when the shows are not on the air, the club is a legitimately great social network because it turns out the people who are members of club twit are great, smart, interesting people. We also do a lot of content in the club, in the club twit discord.

[03:09:06] You can watch. There's always events coming up. Many of our shows are recorded in the club. We've got the AI user group coming up September 4th. We'll also do live coverage club only for the Apple surprise and shine event. The new iPhones will be announced Wednesday, September 9th. We don't stream that in public because Apple has attempted to get us kicked off YouTube for doing that. So we no longer do that in public. You have to be a member of the club.

[03:09:31] Micah Sargent and I will comment as Apple announces its new iPhones. Micah also does this fun crafting corner. He was doing all sorts of stuff, Lego, crocheting, knitting, origami. He's doing some sort of weird paper folding thing beyond origami. Our great photo guru is Chris Marquardt. The assignment this month is the word smooth. The media club. Oh, that's coming up the 18th.

[03:09:58] We're going to do the top animated episodes from Rick and Morty. The Simpsons. That's going to be a lot of fun. Talk about that. The crazy show off by one with Jeff Atwood. We did one on Friday. It was a lot of fun. All of these things because of you and your support. If you believe in independent podcasting, not owned by any big company, telling the truth, giving you the information you need to know without fear or favor for any company, please support what we do. We appreciate it.

[03:10:28] Twit.tv slash club twit. We do stream everything we do. We don't really have a paywall. We stream this show live every Sunday afternoon, 2 to 5 Pacific, 5 to 8 Eastern, 2100 to 2400 UTC. You can watch us in the club, of course, but also everybody can watch on YouTube, Twitch. See a lot of people on Twitch and YouTube watching right now. X, Facebook, LinkedIn, and Kik. And if you're chatting in any of those, I will see your chat.

[03:10:56] I can chat along with you as I have been in Twitch and YouTube today. After the fact, on-demand versions of everything we do available at our website, twit.tv. Each show has its own YouTube feed as well. The master YouTube channel is youtube.com slash twit. You'll see links there to all the other shows. And of course, the best thing to do is subscribe in your favorite podcast player. Despite the attempt by advertisers and big tech to get you to stream shows

[03:11:23] so they can monitor what you're listening to and know how many ads you listen to and et cetera, et cetera. That's what Spotify and Amazon, everybody else wants you to do. I'm a big believer in just RSS podcasts where we don't know who you are. We don't follow what you're doing. We make it available. You download it. You listen at your leisure. I still think that's the best way to do content. And we will always do that. You could find us in every podcast client. Just search for Twit. And that way you'll have it after we're done on a Sunday evening,

[03:11:53] maybe for your Monday morning commute. Thanks for listening, everybody. We've been doing this for 21 years. Here's to 21 more. Next week will be episode 1100 of the podcast. Kind of amazing. Thank you for your support. We appreciate it. And I will see you next week. And as I have said, for all 1099 of those episodes, thanks for being here. We'll see you next week. Another Twit is in the can. He's amazing. He's amazing. He's amazing.