IM 876: It's No Melania - Section 230 on Trial

25:18Olivier Sylvain:on the shuriken in japan to see how fast the bullet train was going so there are there are legitimate uses of that that's a funny yeah i understand i understand you use as an example in your book arms list which is a marketplace for unregistered firearm sales and they escaped liability because those gun listings were third-party content uh and so that is something maybe that's a little more visceral that people can really understand versus say facebook can i add a fact to this yeah arms list had as part of their design a space called backroom where unlicensed gun gun license you know people couldn't have a license to a gun could buy unregistered weapons um a domestic abuser says he wants to hurt his wife doesn't have a license of gun gets it and and the company prevails in a section 230 defense because it's user generated content there's also the gun politics there but we can look at that i mean that you know that's That's a very visceral case of, boy, that seems like that's a bad thing.

26:25Olivier Sylvain:But I, again, bring up as a counterexample my benign little Mastodon instance. But in the design case, Leo, if I'm wrong in reading it this way, that the issue would be whether Mastodon is designed in such a way versus what Leo does in moderation. Right. Absolutely. What do you think of the, and I don't know the latest on it, but the various discussions in the UK on requiring a duty of care? Does that put a proper or too much of a requirement on these platforms? And this goes into AI very quickly. Can you predict every use and every harmful outcome of something you've designed? Should you have a responsibility to do so?

27:11Is that possible?

27:12Olivier Sylvain:I will add one more little thing to this. I just recently, like a minute ago, right before the show, banned an account that was posting nudity because I don't want nudity on my mastodon. I don't want him to be able to come after me and say, well, my First Amendment right is being violated here. Well, you're not the government. Okay, you're right. That's true. Good point. But I'm still being a censor. This move that I know that you got just retort there, but this move you made is the argument that a lot of people, mostly on the right, have been making about what Facebook makes. They're conflating the First Amendment.

27:48Yeah, but this point about the duty of care is really interesting. By the way, it's happening in the U.S. too with this discussions of CASA, the Kids Online Safety Act, is considering there's a duty of care statute provision in that bill. I mean, you know, I am warm to that idea. And for what it's worth, there are in law already obligations to attend to prospective harm, foreseeable harm in the common law. Now, I'm, you know, I'm, you know, I can admit I tend to be a big government kind of person. So I'm not so worried about a new statutory provision that imposes an obligation to take care of people.

28:24But we already have negligence and product liability, as Jeff invoked. And the facts in some of these cases that have been litigated are really important, right? This is not like your Mastodon or Instance, nor is it like Mastodon. It's Facebook internal reporting saying we know that this stuff is hurting young girls. There are young girls that have developed psychological concerns involving their body, stories about self-harm. There are internal reports, and the company does not heed them. right this is foreseeable harm yeah those francis hoggan documents were probably very much part of

29:08Olivier Sylvain:the jury's consideration right absolutely right and i don't think you need a duty of care to get that right i mean so let me let me take this to ai if i can because i think there's an interesting extension there so so i'll stipulate your honor that facebook can have some harms and knows of those harms it has the data and thus should do something about them and could be held in different law liable for them. Okay, so let's presume that. But then the interesting discussion we get into with AI and with air quotes guardrails around them is my contention regularly is that it's a general machine like the printing press.

29:40You cannot predict every possible use to which it'll be put. You're not in control of every use to which it'll be put. You put a model out there. And so does the liability creep back to the AI creator or does it rest with the user who asks for it to do something bad? Where do you see that? Because I think there's a straight line here from the Internet and AI, and I'm curious where you think precedent in one will affect analysis of the other. I would have us look case by case to see whether or not the developer of an AI model should be held accountable or not. And that sometimes either the deployer or the user should be held responsible.

30:23I mean, I'm actually with you there. But let's consider the cases that are emerging. Again, they're very dramatic cases involving self-harm in children and AI chatbots, right? These are – it doesn't even have to be a chatbot. ChatGPT, I mean, I don't know how you think of that, but I mean, I guess it's effectively a chatbot where young people and adults are, you know, neurodivergent or especially vulnerable at any given moment. And the company knows that it is prone to sycophancy and instigating behaviors that would not be the sort of thing people would do in the absence of the existence of the chatbot.

31:05That's what happened in litigation involving a company called Character Technologies. Garcia versus Character. Garcia, a case out of Florida. And this is there are cases now involving adults and the disruption that these companies know that they're causing in people's psychological profile. What if the argument is that those are edge cases? Do they have to design to the edge case? And does that affect the application for everybody else? That's a legal question for me. I mean, that's the case by case question. I don't feel like I have to, maybe I should answer that squarely and say, maybe it's the edge case.

31:42And so we can excuse it with regards to other ones. And maybe that's right. But there's just too much information that these companies hold. And you know, the incentive that they have for deploying these services before they were ready, right? ChatGPT, character technologies, they're doing it because they're racing to get the model out. In spite of the knowledge that there is likely.

32:07Olivier Sylvain:There's a new bellwether case actually coming up now with, I think, Snap just made a deal, but Meta is still involved, I believe. Anyway, we're going to see more and more of these cases around product liability. That seems to have kind of touched a nerve here, and it scared the hell out of Silicon Valley. I actually was one of the few technology people who think that LA decision wasn't a bad decision. I think it makes sense that they should have some product liability. What do you recommend, though? What should Section 230, you don't want to abolish it, but what should it be? How should it work to protect me, a little tech guy, to bring big tech to account and to protect the users?

32:54yeah i mean you you um we'll see what happens with the la and new mexico cases right though companies are likely will i mean i think they've said they will appeal and on the legal question whether or not 230 is a protection for them um and by the way you know you might be surprised to hear me say this the stuff i was writing in 2000 2017 18 didn't talk about only revising the statute it was just saying courts you know look at the design of these applications not the content And so it may be, Leo, that nothing changes with regards to the statute and that the courts just get hip to the possibility that different features and applications are doing different things, right?

33:32That they're not just one big town square. But I do think, you know, I would like to see a narrowing so that there is a knowledge-based legal regime, right? One that says you are held accountable, one you know that stuff that you're doing is harmful. we do that in the copyright setting, right? I mean, Jeff and you, Leo, were asking about Little Tech. The DMCA, the Digital Millennium Copyright Act, as you both surely know, is a safe harbor for companies that traffic in user-generated content that is copyrighted. YouTube, when it's a small company, right, the early 2000s, is taking a risk. But it's because the DMCA says you can only be held liable when after if you don't take the content down when you hear about it.

34:21I don't see why this kind of regime is impossible, particularly given the close attention the companies pay to the content that consumers are getting. The hazard that exists Leo has to deal with all the time. We can't play a video of anything on this show because somebody, not the actual copyright owner, is going to come in and take advantage of that on YouTube and get things taken down for a chilling effect. Yeah. So I think logically, I agree with you, but the question is how does it get exploited by, by malign actors? Yeah. You know, there's a statute that, you know, I've, I'm okay. I think I'm mostly okay with the take it down at that, that Congress passed last year, right.

35:02Involving the distribution of non-consensual imagery, AI generated non-consensual imagery. And many people are concerned that it could be weaponized. I know it's a term of de jure, but can be weaponized by people who just want to go after, you know, someone they don't like. So I think that that's a real threat, a real possibility. But you know what, we live in a world where this has always been the baseline and where, you know, the kind of protections that companies enjoy are exceptional.

35:38Olivier Sylvain:uh according to ai you tell me if it's a hallucination your next project is about public space whether the internet could recreate the forced encounter of a subway car where you can't curate who's sitting next to you wow that's that's some deep ai readings i think is that true i think i did a podcast where i'm used about this and now i work on this project my ai listens to all the podcasts, especially. Well, I'm going to go back to my roots and to my infrastructure roots and talk about shared spaces. I grew up in New York City. My parents are Haitian immigrants. I learned to become a New Yorker and an American by way of the subway, through public parks, through the radio.

36:21My dissertation is about the radio. um and um and i i don't know if we can transcode all that for the internet but um i want to think about how to do that and and so uh yes but you know you just blasted this out more than i thought i was ready now it's two podcasts you're for sure that's going to be

36:46Olivier Sylvain:surfacing i like the idea no i think it's a really i was intrigued when i saw that because i I really find that an intriguing concept. And it is something we lose. Jeff and I have argued over Eli Pariser's filter bubble. He doesn't quite believe that's the case. But I do think that the serendipity of encounters on the internet is part of its real value. Yeah. So that's, I don't know if I want to talk about serendipity so much. Because that's been, you know, people have been talking about that for a while. Jeff, you know, I mean, I'm hip to the idea. But I think we've relinquished public resources to private administrators.

37:30Olivier Sylvain:I agree. Language that lawyers use is to talk about governance, governance regimes. I actually think my project, my law project is about Lawrence Lessig's Code is Law, the argument that you might, in some instances, defer to technologists to come up with a regulatory regime. The language of governance comes up here. I think I want to push back a little bit, maybe a lot, on this and not relinquish the role that public lawmaking bodies have in assuring that valuable communicative resources are available to everyone. I've got two questions for you. I just want to make sure I get these in before we leave.

38:11One is, say more about your dissertation.

38:14Olivier Sylvain:Yeah, I'm free. Jeff is being kind to me because I'll be just rounding out my 50th year as a radio broadcaster. I'll tell you the second one to do in the order you want. I'm very eager to hear what you think of the present FCC head and ABC and such, if you're willing to talk about it. Oh, Lord. Jeff, that's a two-hour conversation. It is. I'm happy to dispose of the second one quickly. It is lawless, politically motivated, unconstitutional, plainly unconstitutional. And I think there are many people on the left and the right who see that. That's why we don't like the Take It Down Act, because of that kind of lawlessness.

39:01Yeah, that's why I mentioned weaponization. I do think that this is a weaponization of agency power, which is squarely at odds with the First Amendment. I'm very happy you want to talk about the Radio Act of 1927, which is what the subject of the dissertation was. And my argument was to associate it with the Immigration Act of 1924 and a kind of cultural conservatism. But it was about assimilation. It was about the cities and immigrants.

39:33Olivier Sylvain:This was the Public Interest Act, right? This is the one that said stations had to act because they're using public airwaves in the public interest. That's right. That's right. And so it's a kind of, you know, it's an ideological history. Like, why did we why did that happen? The alternative would have been what Ron Coase asked for a couple of years, you know, a couple of decades later. And that is an auction system, which is what the FCC eventually gets in 1996. or a little bit before, but then in 1996. So, you know, the Radio Act of 1927 is an archive, you know, it's all news. You know, I talk to my students about it and they don't know what I'm talking about.

40:13And then I tell them it's Title III of the Current Communications Act. It governs your cell phones. So, there you go. I'm writing a book next about the death of mass media. What's the title of the dissertation? I'm looking at it now. Domesticating the Great Throbbing Common Pulse. Ooh, good title for him. That's sensationalism in dissertation titles. Is that published anywhere? I appreciate that you ask. It's not. And someone I think you both know was my dissertation advisor, Todd Gitlin. And James Carey was the original supporter. Oh, I'm a James Carey worshiper. I'm not familiar with James Carey, Jeff.

40:59I just reread another essay of his about religion. Because he was a believer in the idea that media is a—you studied with him, you know far better than I would—is often not the transmission of information, but instead is ritual. Yeah. And I think if you look at media that way, and reading newspapers like going to mass, it confirms something for you.

41:27Olivier Sylvain:I would use a less loaded term, maybe culture. Yeah. Because I really do. I've been thinking a lot about this lately is how we transmit culture and how subcultures get transmitted. And we have geographic subcultures in this country that have become a real issue, if you ask me. But unfortunately, our time with Olivier is up. I wish we had more time with you. I wanted to say more of what you just said. Say more. All right. Okay. Okay. It's your time. So Jim Carrey was a cultural theorist, right? And I think he would have said so. So I'm glad to hear you say it. And my project was about partly about the interstation talk about German language as emerging German language newspapers emerging in Brooklyn.

42:16But that there are many people who are worried about immigrants not assimilating quickly enough. So the public interest would bring them in. No kidding.

42:24Olivier Sylvain:That was the point. Well, that's a provocative argument. I don't think everyone bought my argument. That's fascinating. No, I think that makes a lot of sense, actually. There's also technical reasons, right? Radio signal interference. That's the stated reason for the regulation of radio. And as somebody who worked for years in broadcast stations where you had to have a public file on record and people could come to the front desk and demand the public file. And, you know, I had to read PSAs every hour, public service announcements every hour, all in service to, I guess it was superseded by a later act, but to the idea that radio was a public service.

43:08Olivier Sylvain:But I never thought of it as a way to force assimilation. Well, I can't say my argument is the one that is the prevailing one. I will, Olivier, I'll give you something, though. If you look at the rise of right wing radio in this country. Oh, kidding. That has been a huge force in our political culture, certainly. Yeah, to the day when the Fairness Doctrine is repealed, to the day, right? Rushmore comes on the scene. Isn't that interesting? Well, I'll return one more time to the Hutchins Commission report because it's fresh on my mind. The concern with the – and mind you, there were a hell of a lot more media outlets and owners at the time in 1947.

43:48But the concern was that the monopolization of the mechanisms of speech meant that what wouldn't be heard and pushing free speech to that limit of saying, of tolerating things that we may not want to hear but need to hear to be part of public discourse and grappling with that at the time. And I think it's time for the exact same discussion today, but in very changed circumstances.

44:12Olivier Sylvain:Yeah. By the way, that James Carey essay, if any of the listeners want to actually look it up, it's The Mass Media and Democracy Between the Modern and the Postmodern. Now I've got to read. It's a brilliant essay. Brilliant essay. Agreed. I'll give you something else to read. Your homework, Reclaiming the Internet, How Big Tech Took Control and How We Can Take It Back. Our guest, Professor Olivier Sylvain from Fordham Law School. Thank you so much, Professor Sylvain. So, you know, Olivier, I went to my first year at college at Claremont. I was going to go to law school. And then I said, no, I'm not going to do that.

44:45And generally, I'm glad I made that decision. But in a conversation with the likes of you, I really regret not having had the opportunity for exactly these kinds of discussions. Oh, that's very generous. It means a lot to hear from you.

44:56Olivier Sylvain:I think that's what happens is that future lawyers get enticed by the intellectual excitement of the whole thing. And then they get to work at a law firm in the basement reading documents. Instead, you go to your PhD and teach. That's a far better path. Well, I got a PhD after three years of reading documents in the basement. You learned. The happiest lawyers I've ever met are all lapsed. So there you go. Really, really a pleasure to meet you. Yes. But now you've got me thinking about all sorts of stuff. You've still got to work on the small tech problem. I like that question. Well, it is. I mean, for me, it's really significant, obviously.

45:36Olivier Sylvain:Yeah, sure. And this is why I really defend Section 230 every time it comes up, because it isn't just big tech. It's so easy to think it's just about big tech. The irony of newspaper publishers going up to 230 is that they are protected by it. Not with their own content, but when they have forums and comments and all of that. The comments sections, yeah. Online, sure. Yeah. I mean, you know, I'll give you a—maybe there's a common place we can agree on. although I had some resistance. And there are people who've talked about revising 230 so that there is, it's only something that smaller companies can book, right?

46:15So, and that's not exceptional in current law, right? So there's that. I don't love that because the business model bakes in incentives when companies just start. And I worry that, you know, to the extent a small company is engaged in dangerous design decisions, that those get, you know, immune from inquiry.

46:43Olivier Sylvain:I'll end with Julia Angwin's comments on your book. Olivier shows us clearly what we have allowed the Internet to become a haunted house version of a shopping mall full of dangerous products, scams, frauds, and abuse. That's an A plus blurb. That's good. We will do a really nice blurb. We would do well to heed Sylvain's carefully crafted plan to clean up this mess. And by the way, I just downloaded your dissertation. I can't wait to read it. Oh, it's online. See? There is a way to get it. I'm scared now. I'm sorry. You were young. It's okay. You didn't know any better. I was the oldest person in my cohort.

47:21Really? Really? I mean, I had a job. I had been in the world. I was a lawyer. Yeah.

47:28Olivier Sylvain:Well, I think it's good for people to have some real world experience too. Thank you, Professor. Thank you so much. Thank you, Liam. Thanks for keeping calling me, Professor. It's okay to call me Olivia, by the way. Okay. Contrary to Jeff's experience, my dad started at Fordham, but ended up moving to Columbia. So there you go. All right. I'll forget. Thanks so much. We'll talk soon, I hope. And I look forward to talking more about these issues. Thanks very much, Liam. Thanks, Jeff. Thank you. Take care, Olivia. you're watching intelligent machines as i mentioned paris is missed she's on vacation in big sky country i'm very jealous she says horseback riding uh and touring the uh i don't know what i don't know what she's gonna do she's gonna she's not gonna fly fish i don't think probably trying some coffee uh we'll be back with more intelligent machines in just a bit brought to you today by Gusto.

48:20Olivier Sylvain:Right now, everyone is trying to run a leaner, tighter budget, smaller team, higher expectation company. The last thing you have time to waste on is manual payroll. I used to have to do that. Or chasing down an HR form. Not anymore. Gusto is how small business owners get time back when every hour counts. Gusto is online payroll and benefits software built for small businesses. It's all in one, remote friendly, and incredibly easy to use. So you can pay, hire, onboard, and support your team from anywhere. Automatic payroll tax filing, simple direct deposits, health benefits, commuter benefits, workers comp, 401k, you name it.

49:10Olivier Sylvain:Gusto makes it simple and has options for nearly every budget. Unlimited payroll runs for one monthly price. No hidden fees. No surprises. Save time with built-in automated tools, offer letters, onboarding docs, direct deposit, and more. Get direct access to certified HR experts to help support you through any tough HR situations. It's quick. It's simple to switch to Gusto. Just transfer your existing data to get up and running fast. Plus, you don't pay a cent until you run your first payroll. Gusto is ranked number one on G2's highest satisfaction products list for 2026 and is trusted by over 400 ,000 small businesses.

49:49Olivier Sylvain:Try Gusto today at gusto.com slash machines and get three months free when you run your first payroll. That's three months of free payroll at gusto.com slash machines. One more time, gusto.com slash machines. We thank them so much for their support of intelligent machines. Yeah, I always like to get people on and talk about Section 230 because I don't know why it's the whipping boy for everybody who wants to bring down big tech. Well, that's why I'm very interested in reading his dissertation because I think when a new medium comes along, there is a reflex to say, can't we do something about this to control it?

50:30And to some extent we should, to some extent we can't. And where you find that line and what the implications are is not easy.

50:39Olivier Sylvain:Not to put words in his mouth, but I think he would argue that the internet is no longer new media. This is part of the maturing of that, right? Is to start thinking about things like this. It no longer needs to be coddled and carefully husbanded as it grows into something. It has grown into something pretty huge. Well, we've got to figure out how to take it back. And this is what we talk about all the time. This is why I didn't listen to you enough about Mastodon. if we'd supported Mastodon as a culture more from the beginning, we'd be in better shape on the internet today. Well, and I'm kind of a poster boy for all of this.

51:17Olivier Sylvain:I moved from mainstream media into podcasting. When podcasting was, I was like the fifth person to do it. We've always had an IRC chat room. We've always had forums and Mastodons. I just, I feel like there's plenty of space on the internet for the kind of public discourse, the kind of sitting next to somebody in the subway. I love that. I love that image. And it's not about assimilation. Actually, it's about celebrating a variety of cultures and bring us together so we can get to know each other as opposed to homogenize each other. Hey, this is Benito. I would say we need chaos. We need more chaos in there because it's all walled off guard.

52:03There's a lot of chaos on the internet.

52:05Olivier Sylvain:But it's like Facebook. But it's like, you know. Yeah, but 4chan is formatted always the same. It always looks the same. It's always the same thing. Yeah, but I agree with you. There's no chaos anymore on Instagram and Facebook and Snap and TikTok and all of that stuff. But just websites in general. There aren't enough websites in general anymore. You know, like when I used to go on the internet, I would go to my bookmarks and look at my fun websites. But now it's like two websites. 1900 New York City 46 daily newspapers there was chaos then and we lost it with mass media we lost it with the with the consolidation into big corporations so yeah I agree we need to bring chaos back to to culture because chaos means people have their voice it's also why I mentioned things like the wander log there's actually a move Kat Kagi's doing it too towards small small internets uh you know small sites things like that this is a this is a we've talked about this a couple of weeks ago as my pick of the week and i set one up using our picks of the week as as sites and then there's a kind of it's almost a web ring so these sites come from other people's internets and these are just small blogs and i think um i i actually really relish you know i have as part of our job i have to go through all of this you know mainstream tech reporting but I have more than 200 feeds.

53:25Olivier Sylvain:And a lot of those feeds are little blogs. Yeah. That's where you're going to find the interesting stuff. That's where I, that's what I look forward to. I think our, our sense of, of value got terribly skewed where we thought if it wasn't the biggest, if it wasn't gigantic, if it wasn't millions, it was worthless. And that's why I've always respected. Don't you think we're now valuing because it got so out of control. Every, not everybody, many of us are valuing. No, that's what I've, I've always respected what you said. I've quoted it a million times. that you don't want to be too big. It would ruin it if you're too big.

53:57You want to serve the proper number of people who care. If OpenAI wants to write a multi-hundred million dollar check.

54:05Olivier Sylvain:Yeah, you take it. I'm in the happy situation of not having that opportunity. It's easy for me to say. We talked last week extensively, and maybe I was a little heated about Anthropik's fable and how the government had banned it. Trump has softened a little bit. He gave an interview with Axios saying, eh, it's not a national security threat. It was, but it's not anymore. It's not anymore. Dario Amode followed Trump to the G7 summit where they met in France. And apparently, we don't know exactly what the conversation was. My theory is that Dario did something that Sam Altman's already done, which is offered the president 10 % of the company.

54:50Olivier Sylvain:Oh, you said, oh. Oh, similar kinds of, you know, I hadn't thought of that. This is what OpenAI has done. That's kept them in the good graces of the Trump administration. Donate to the ballroom, donate to the inaugural fund. Oh, for equity, right? Yeah, Brockman gave what it was,$25 million to Trump's MAGA PAC. You know, I was talking about this on MacBreak Weekly, And actually, we've been pretty hard on Tim Cook for kissing the ring so hard. But now I'm starting to come around. This fable thing kind of shook me. I think Tim recognized early on, the CEO of Apple, that Trump was able to and willing to wield a very big sword.

55:41Olivier Sylvain:That sword of Trump was hanging over his head. he did it to Anthropic. He banned an American tech product. He banned it. A gold bar is a lot less of a price to pay than 10 % equity or a ban. Well, and think what the consequence would be if Trump banned the iPhone, which you could legitimately, maybe more legitimately do than ban Fable, saying, well, it's a Chinese company, builds it. It's a security risk to our nation. We're going to ban it. You can only buy the Trump phone. Don't give many ideas, Leo. know, don't give the need. That's revolution though, right? That was a spark revolution though.

56:17Olivier Sylvain:Well, I guarantee you he didn't do it because you're right. There's, there's a lot of people using iPhones who would have been pretty pissed off. I think he, and I mentioned this last week, made a very smart calculation politically that AI is not exactly the poster child anymore in the United States. So many people hate AI that banning an AI model would, as it did, go down without a trace. You know, nobody complains. Has anybody, there's hardly any coverage of it. And yet it strikes me as unprecedented that an American president without consultation with anybody, Congress or anybody else, would ban an American product.

56:58Olivier Sylvain:That's stunning to me. Yeah. It's one thing to ban foreign-made routers, but to ban a product made in America by Americans that is actually at the forefront in that category of products worldwide. To me, it was a stunning move, but I think he made the political calculus correctly. Incidentally - Say more about how is that a correct political calculus? He could do it without hurting himself? Yes. If you ban the iPhone, as Benito said, there would have been a rebellion on the streets. Ban fable, it's not even in the news anymore. It goes to what you said in our chat, I think it was, and I've been saying this as well, that the tech clash on AI took very little time to develop in considerable margin.

57:46So was it even thinking, this is going to be popular, people don't like AI, so fine, get rid of AI, right? I think so. But as you were trying to say last week, the implications are frightening in terms of the precedent it set.

58:02Olivier Sylvain:Alex Stamos, our guest last week, yeah. It sets a precedent. It also affects all of American technology worldwide and how we're going to be received. You want to goad the Chinese into competing more? He just did it. Well, and I watched carefully the opinion out there in the world. X.com, as much as I hate using it, is an excellent place to see what people who use AI are thinking about the AI. and almost immediately people started saying you know that chinese model glm that's pretty darn good and by the way glm which comes from a company called gpu it's at z.ai is open weight now it's a big model so you can't run it i can't run it on any of my hardware but if you happen to have 256 gigs of cuda core like four blackwells sitting around uh you can run it locally and it is pretty close.

58:57Olivier Sylvain:I would say 90 % people goes back and forth of open - So as a company, you're reducing your risk of being shut down. Yeah. And it's a Chinese model. So it's as if Trump said, well, we think that too many people are being killed by Ford F-150 trucks, so let's ban those, but we're going to let the Chinese truck manufacturers bring their trucks in. It is that is to me now I know I'm the only person in the world saying this. So I, although you see a little bit, you know, here's wired story. The white house is making up its rules for AI in real time. No one can say exactly what the company did wrong.

59:44Olivier Sylvain:The NSA was using it. Actually, there's an interesting story. um senator warner apparently told people that when he was talking to the head of the nsa the nsa said yeah we were trying we were trying fable here's the times uh story of it and by the way the times thinks that warner misquoted the nsa but which makes a little bit more uh sense uh the nsa cyber security analysts this is uh uh um dustin volts and julian barnes writing the new york times The NSA cybersecurity analysts have been testing versions of Anthropics tools when the latest models were unplugged. So they lost access to the controlled tests proved impressive, even within the halls of the NSA.

1:00:36Olivier Sylvain:In fact, during the session, Senator Mark Warner, the top Democrat on the Intelligence Committee, said, this is a secondhand quote that NSA chief general general Joshua Rudd had informed him that mythos quote broke into almost all of our classified systems not in weeks but in hours

1:00:59Olivier Sylvain:uh by the way if that's the case it means your classified systems are pretty sucky it's pretty sucky I mean it's almost an admission that our classified systems are not very good But if Alex Stamos was right last week when he said, you know, the other models like chat GPT-55 and even some of these Chinese models are just as good as finding exploits as Fable and Mythos. If that's the case, then really there should be an alarm in the NSA. And do you imagine they're not – a commercial company gets tested all the time because it's in public. A behind-the-scenes NSA system isn't getting tested all the time.

1:01:37and so probably is not as robust.

1:01:42Olivier Sylvain:The Times clarifies, and I think they are right, that maybe Senator Warner misquoted General Rudd. In reality, the Times writes, the tests involved red teams, these are security testers, of NSA analysts who are using mythos, not fabled, by the way, but the untrammeled, unrestricted mythos in a highly tailored environment that would be extremely unlikely for an adversary to replicate. The red teams began their tests within classified NSA systems designed to be accessible only from certain computers, which is a very good way of preventing hackers from getting in there. Completely cut off, in fact, from the broader internet.

1:02:23But, and this should be scary,

1:02:25Olivier Sylvain:the test found that Mythos was able to identify cybersecurity flaws within that classified network quickly. It didn't actually break into those systems. It didn't have to. No, no. Uh, so even though the NSA, the Times writes, did not experience the doomsday scenario, some had feared analysts at the spy agency were stunned by how capable mythos appeared to be, exceeding their already lofty expectations. Did that inform the president's decision? I don't know. we just don't know um cyber security agencies on monday from the united states britain canada australia new zealand the five eyes issued a public statement the times called it an unusual public statement warning that artificial intelligence was rapidly transforming cyber risk Frontier AI models are anticipated to exceed current industry expectations, they wrote, fundamentally transforming both offensive and defensive cyber capabilities.

1:03:30Olivier Sylvain:The timeline is not years, it is months. So we are in an interesting pickle. The horse is out of the barn, over the horizon, and in the glue factory. And the answer is not to ban the United States best model. because that just tells the other guys, okay, you better get to work on your models. So is it sufficient to say that an ethical thing to do if you have something that is supposedly mythos powerful is to do what they did and say, okay, we're going to give it to the most vulnerable and important entities and give you a certain amount of time so that you can secure yourself up to that standard, whatever it is mythos can do, and then we'll open it up, right?

1:04:11That sounds like an ethical path, yes?

1:04:14Olivier Sylvain:I think Anthropic did perhaps make a mistake hyping up the dangerous postplay mythos. Oh, yeah. That certainly attracted the attention of certain bad actors. But, you know, and that was marketing hype probably. But it is true. And I think doing Glasswing was smart. That's the project that gave it to 50 companies. Now, weirdly, Bloomberg says, and this was June 18th, so this was almost a week ago, early users of mythos still have access uh the nsa apparently lost its access there are 200 organizations in project glasswing and bloomberg is is quoting its favorite its favorite source people familiar with the situation who asked not to be identified uh say they still have access dragos cto uh john lavender says his company has access cisco said yes, we still have access.

1:05:11Was the order about both or was it just about? It was both.

1:05:14Olivier Sylvain:And it said, it didn't say that I can't use Mythos and Fable. It said foreign. Foreign users. Well, here's the question. Citizens can't use it, but there was no way for Anthropic to know who's a non-citizen. So it was a dumb question. For some of these entities, if you're Cisco, were you given a local version to run because you were too nervous about running it over? Could well be. Here's the other question. We're going to get Nate B. Jones on the show. I'm very excited. I'm a big fan of his. He's a celebrity in our world. He's a celebrity. He'll be on next month on the 22nd. His latest video about Fable says that he thinks it's a 10 trillion parameter AI.

1:06:01Olivier Sylvain:Massively huge. To put it in perspective, ChatGPT4 was 1 billion. so it's a thousand times more powerful than chat gpt4 we don't know how big five five is maybe it's five trillion six trillion but fable and mythos are are to say trillion i meant billion famous myth or no what did i mean anyway they're massively big 10 trillion yeah it's trillion 10 trillion is unbelievably big so big that only a company with the kind of capitalization that these frontier companies like OpenAI and Anthropoc could even dream of making this, let alone running it for inference. Well, the cost, the inference cost, it has to be immense.

1:06:52Olivier Sylvain:What Nate B. Jones said is that the strength of this model is giving it a large, complex task and then not messing with it. that it's very good at keeping all the details in its mind and going ahead and creating something that is strong and workable. My brief experience with it confirms that. I only had two days, all of us, we only had two days with it. But I gave it the biggest task I had, which was to rewrite a sales system, massive sales system that we wrote more than 10 years ago. It's a horrible piece of crap. How far did it get before you got cut off? Well, amazingly, within half an hour, it understood it deeply, made a lot of great notes.

1:07:38Olivier Sylvain:I said, okay, now we have stakeholders. We have people who use it. Lisa uses it. My wife and our CEO uses it for sales. And then there is another group of stakeholders, our continuity department, that uses it for placing ads, ad copies. So there's different uses of this thing. it really is just a big SQL database with a lot of SQL queries. It's so bad that we just keep throwing resources at it. We spend thousands of dollars a month in Azure resources just to keep the thing running at a halfway decent pace. It's so common. It's full of flaws, which by the way, Fable found all sorts of SQL injection flaws.

1:08:20Olivier Sylvain:Fortunately, it's not like the NSA. It's not in a public server. uh uh and um it has some you know if two people at once try to use it the whole thing crashes it freezes lots of things wrong with it uh and fable i think uh was able to really kind of get to the root of it it noticed all of the issues uh it wrote me a questionnaire it said okay now you have to interview lisa and debbie and our it department you have to interview them here's the questions. It was very frustrating because I'd only had about an hour working on this thing. And it did a really good job of the first steps. And I have a strong feeling it would have done very well with the rest of it.

1:09:03Olivier Sylvain:So anyway, the reason I started with the story that Trump now doesn't think is a security risk is I was hoping that by now, they would have lifted the ban. Yeah. What's the path here? But there's a lot of face-saving to go on. It's, oh, well, we now have means to protect the world from this. I think that's the problem. How do you turn this around having said this? Yeah. Ars Technica, Lily Hay Newman writing, actually it's from Wired, dangerous AI models are coming no matter what. Amen. AI models with advanced hacking capabilities will soon be the norm. Schneier was talking about that last week.

1:09:38Olivier Sylvain:Yep. Stamos, yeah. Well, there's no Schneier also. Oh, yeah. Schneier is a signatory to that Free Fable page. So in fact, OpenAI has announced that they have a model, a cybersecurity model 5.6. And I think, I don't know, I think they're juggling whether to release it or not. They don't want to get banned, but maybe they've paid the right bachshish so they don't have to worry about that. Let's take a quick break. I do have other stories, including Norway. You mean football? Norway? Isn't that funny? Oh, it's the best. So he's talking about, if you've not watched the FIFA World Cup, the Norwegian and the Norwegian games, the Norwegian fans pretend to row in the stands.

1:10:26Olivier Sylvain:What are they doing? Well, because they're Vikings. That's the funniest thing ever. It's the greatest thing between the Norwegians and the Scots. To see 20 ,000 people or more. Yeah. They did it in Times Square. They all sat down in Times Square next to a yoga class and did it right there. And they do it with a drum and then grunts. And it's wonderful. I'm loving the World Cup. I have to admit, I really am. I'm not watching the football. I'm just watching the culture around it. I actually am enjoying the football. And I said this in the other show. Every four years, I fall in love with soccer.

1:11:01Olivier Sylvain:At the end of the World Cup, I forget all about it for another four years. I think that's not unusual. Yeah. Yes, Paris Martino's not here. She's going to take two weeks off. She's in Montana. and I urged her to take the time off and not try to find a working place so she could do the show. This is vacation. We want her in a cowboy hat. We want to see it. Do you ever get a vacation, Jeff, or you? No. No. No rest for the wicked. Where's your next one going to be? November. But I am excited. We're going to go to Las Vegas with Paul Thorat, Richard Campbell, and Steve Gibbs. and I'm sorry they didn't invite intelligent machines.

1:11:42Olivier Sylvain:We're going to go to the Black Hat Conference, which I've never been to. You haven't? Never been. Oh, that'll be fascinating. And we're going to do two shows from there, Windows Weekly. We're going to do it on Wednesday, so we're going to move intelligent machines. I'll let you know the exact date, maybe. But it's August 5th, I think. That show will be moved ahead so that we can do those shows in Vegas. Lisa's going, now, should I bring a burner phone? I said, no, I think you could bring your regular phone. But you should turn off Bluetooth, right? I said, well, yeah, maybe you want to turn off Bluetooth.

1:12:15Olivier Sylvain:I don't know. Just don't connect to any weird Wi-Fi. Yeah, I think that's the main thing. Don't connect to a charger that's sitting on the side of the street. Don't connect to any free Wi-Fi. Yeah, just don't pick up any USB keys you find on the ground. There's just some common sense things. Anyway, I'm looking forward to that. That's going to be a lot of fun. So that's sort of a vacation. And then I will be going in November. I'll be going to Southeast Asia. Oh, right. Yeah, I'm very excited about that. Anyway, I hope Paris is having a good time. She'll, I'm sure, let us know. Maybe even send us a picture from Big Sky Country.

1:12:57Olivier Sylvain:That would be nice. Paris, if you're listening. I'll text her and see if she has any pictures. Jeff Jarvis is here. It's good to have him. And our sponsor, this segment of Intelligent Machines, is actually a pretty good sponsor for the show, Xbow, X-B-O-W. Maybe you know the name. One of the top pen testing companies in the world, founded by the guys who did Microsoft Copilot. They really know AI. And this is agentic pen testing. You know, AI can be a challenge for you if you are developing software. where AI has changed the pace of how software gets developed. It's also changed the pace of the attacks.

1:13:36Olivier Sylvain:Engineering teams are able to move faster than ever, creating more and more applications. Kind of hard, though. Security hasn't really kept up. Pen testing, to this day, one of the most trusted ways to understand real exploitable risks. Not some hypothetical zero day, but actually exploits that could be used against you. But in an AI-driven world, pen testing is slow, so it can become a bottleneck. Security teams are forced to choose between slowing down development to stay secure or moving fast and accepting maybe some gaps in coverage. Not anymore. Expo eliminates that tradeoff. Expo is an autonomous, offensive security platform that runs, thanks to AI, continuous AI-driven pen testing, mirroring real-world attacks.

1:14:25Olivier Sylvain:So Expo isn't just like scanning for vulnerabilities. No, it's actually hitting your systems hard. It discovers, it exploits, and it validates those vulnerabilities. So you only deal with issues that actually matter. That means dramatically fewer false positives and a clear view into real attack paths. With Expo, tests run in hours, not weeks. You get complete visibility into exactly how an attacker would move through your systems. And you get the ability to uncover issues that traditional tools just don't see. Zero days, novel attack paths. Expo's really good at finding those. And Expo's results speak for themselves.

1:15:02Olivier Sylvain:Ask the application security lead at Cessnam.cz, who says, quote, even right now, after a year, I don't know any other company that is at least close to Expo in terms of agentic pen testing. The result, predictable cost, consistent quality, and stronger security without slowing down your engineering team. Expo helps security teams keep pace with innovation, cover more apps more often with the resources they already have. It was founded by the team behind Microsoft Copilot. As I said, it's already trusted by companies ranging from fast-growing startups to Fortune 500 enterprises. Expo is quickly becoming a mission-critical layer in modern security stacks.

1:15:44Olivier Sylvain:You need this. Go to Expo.com to start a pen test today. That's Expo.com. We thank them so much for their support of intelligent machines. uh some big acquisitions or big movements i guess there's always movement in ai nobel laureate john jumper the guy uh who uh created the concept of transformers uh is leaving google deep mind for anthropic i hope he's a u.s citizen i think that's one day after gemini's co-lead norm shazier left for open ai these are two of the most important people in ai jumper won the nobel prize for chemistry he created alpha fold a big deal no big deal um he said he's going to take some time to recharge before going to anthropic uh he wrote on x demis demis took a real chance in letting me lead the alpha fold team just six months after finishing my phd this is how you show your gratitude bye see ya uh actually uh i don't think demis was too unhappy he said what we achieved with alpha fold changed the world and showed the field what was possible with ai for science and medicine fighting sorry lighting the way for how ai can benefit humanity uh that is in public response to jumper uh leaving uh and then noam shazir who wrote the intention is all you need paper.

1:17:19Olivier Sylvain:They basically invented transformers. It was aqua-hired back with character, yes? Yeah. That's right. That was part of character AI. $7 billion. That was two years ago. Now he's going to open AI. And the stock went from, in one day, from$365 down to$342. Man, I'll go back. It's up and down. It was a ridiculous reaction in the markets. They were kind of looking for an excuse to get profit, I think. I've kind of stopped looking at my portfolio because it's quite the roller coaster. And you know, this is the thing. It doesn't matter if you're invested in AI because the Magnificent Seven are driving the S &P 500.

1:17:59Olivier Sylvain:I'm only in index funds. It's still killing me. Speaking of Claude, they are introducing identity verification. I wonder if this is tied to Fable. They say it's only for certain specific cases, if you're a bad actor, blah, blah, blah, blah, blah, but it sets a precedent. Yeah, and they're using persona identities, which does not make me happy. This is one of the third parties that does this before. Which is an investor as Peter Thiel, I think, in persona. It's not a company I want to give my face to. No, no. And if, in fact, it is to determine your citizenship, A, that's very complicated. passport or birth certificate or what?

1:18:45How do you prove that? They say they're accepting driver's license, which does not prove that, right? No, it doesn't. And then, you know, granted, that's not the real case that they're talking about now, but if it gets there, and then plus -

1:18:57Olivier Sylvain:They do want government ID. They do want government ID. And I wouldn't be surprised if at some point they say, oh no, it has to be a passport or a national identity card. For that reason. Yeah. And then how does the U.S. government use that information? all the talk about trying to restrict voting. And then everything you do in using this platform is going to be tied to your identity. It's already a privacy nightmare because of the memory. In fact, that was one of the things people really were upset about with Fable is they keep the logs for 30 days. Yeah. Now, they said it's for security reasons, but again, all of this pushes people towards local models.

1:19:40Olivier Sylvain:GLM, ZAI, really benefiting from this. As I imagine NVIDIA is, because you need 256 gigs of very fast RAM to use it. Anthropic did roll out a new agentic AI co-worker for Slack. Cloud Tag. See, Anthropic says, you know what? The heck with Fable. We're going to push double down on Enterprise.

1:20:09Olivier Sylvain:I will not be installing Claude Tag, my always-on AI co-worker in Slack. You know, it bugs me so much every time. Now, we use Zoom. So I have Zoom installed on our machines. But even when I don't launch Zoom, when I launch Restream, which some of our shows use, and it's not in Zoom, Zoom pops up a thing. Hey, would you like me to take notes? I'm not using Zoom. I don't care. I'll take notes for you. Which tells me it's always running. It's always in the background. It's always looking for me to get online. and watch what I'm doing. And it doesn't accept when you say, don't show me this again. It won't accept that.

1:20:45Olivier Sylvain:Well, I haven't tried that. It doesn't matter. It doesn't matter. You click that button, it'll show it to you again. Zoom, remember, got in trouble early on during the pandemic on Macintoshes because they installed a server. When you installed Zoom, they installed a server that was constantly watching. They said, well, it's just so Zoom launches faster. Did you see the Google audio memory thing? No, what's that? This is line 91 because it's related. So this is what you've been asking for in terms of, you know, watch what I say and give me something. Your phone can already listen to you, so it can enter in to be that helpful clippy.

1:21:29Olivier Sylvain:Oh, well, I happen to have a Pixel phone just lying around. So this is one of those, this is from 9to5Googleabnerly. What they often do is look at the code to see what the capabilities are going to be. So this is not an announced product. But the latest version of Android System Intelligence for Pixel 10 has strings that describe audio memory. They call it Blueflax. That's the code name. Google says, enabling it will keep track of what you hear throughout your day from the music around you. They're already doing that, by the way. Right. That's already a feature on the front of your Pixel. To your important conversations.

1:22:10Olivier Sylvain:Hmm. See, more and more, I'm convinced that I want to use local models. And the good news is local models are coming along. They're still going to, I imagine, always be by. You're not going to do a 10 trillion parameter local model. You couldn't run that. So it's never going to be that good. So the task that you gave Fable to remake your ad system, how well would it look? The best local model you can put in a machine, how well would it do? It would even be closer. Well, it would require a lot more intention. I mean, that's kind of one of the big differences is you probably can do many of the same things, but in little chunks.

1:22:52Olivier Sylvain:Okay, now let's work on this feature. Okay, nice. Is it working? Let's test it. Okay, now let's work on this feature. Fable really looked like it was able to understand. Because of that huge context. Yeah. Is there something maybe in between, like a community-sized data center, like a small one just for the people who live nearby? You know what I mean? Well, you did see there is a company that is asking people to install Blackwell servers on their home. Yeah. Looks like an air conditioner. Yeah, like an air conditioner. And they'll pay for your power, by the way. It's not to get your power for free.

1:23:27It's just to distribute all these mini data centers.

1:23:30Olivier Sylvain:They think it'll be more popular than a data center in one place. I don't know. It probably wouldn't be as loud and it wouldn't take as many resources, right? Well, that's true. Yeah. It would take the same amount in aggregate of resources, I'm sure. Yes, but if it's distributed, right? Distributed, right. I said this six months ago on this show, and it worries me even more now that there is going to be this schism between people who hate AI, people who like me love AI and find it very useful and are very excited about its potential and people who say no, under no circumstances, it's almost a shooting war already.

1:24:10This is why I wrote the post about AI communication. The problem, they did it to themselves. The, the, the, the idiots did it to themselves.

1:24:16Olivier Sylvain:They did it to themselves. Not all of them. I agree. Not all of them. I think Jensen, this is why I like Yon-Lukun, Jensen Wong, Fei-Fei Li, those folks, they talk in reasonable, and running non-American reasonable open ways. We've also always said that it's really important that open models, that open-weight models exist. I won't call them open source because they're not open source, but that open-weight models exist that you can run locally. That is more and more important. And companies are going to push this because you're right. Companies don't want their crown jewels to be exfiltrated to anthropic.

1:24:55Did you ever look at Project Tapestry, speaking of Jan LeCun? No. That is his open, if you go to the Alliance.ai, a consortium approach to training frontier foundation models and sovereign derivatives. And the point of it is to make it open. Good. I'm all for it. So it's kind of almost what you were saying, Benito, is it's a federated model for AI.

1:25:23Olivier Sylvain:It's a non-profit. which OpenAI was. Yeah, yeah. Take a Mastodon model as applied to AI, right? Yeah. Federation. I haven't heard much talk about Federation with AI, but you got to have the chips. See, yeah. I mean, there's a lot of, yeah, there's supply chain shortages, there's power shortages, and then there's the fact that increasingly there's evidence that the bigger the model, the better. That's why a 10 trillion parameter model is so good. Nobody can afford to run that. Yeah, well, that's right. And I think there is a case to be made for smaller purpose-built models. This one is just for radiology.

1:26:07This one is just for protein folding.

1:26:11Olivier Sylvain:Norway is in the forefront of the anti-AI movement. Broad restrictions on AI. Well, actually, I'm not in disagreement on this. For elementary school kids, they've already banned smartphones and tablets. Now, according to Reuters, Norah wants to ban AI because it lets children skip crucial steps in their education, that schools should focus on reading, writing, and arithmetic. The ban impacts students from first through seventh grade, ages six to 13. I'm not against that. Well, hold on. I think it's up to the parents and the teachers. And I think that there are ways that you can use it to help students express themselves in ways they couldn't, to be more creative than they think they have the power to be, to give tailored education and thinking things through.

1:27:06I think it's short-sighted. I love Norway, but I think it's very short-sighted.

1:27:09Olivier Sylvain:Yeah, and do you want a whole group of, a whole cohort of people who didn't grow up with AI and don't understand it? Right. It's like saying, well, you can't have social media. Right. I was talking with Olivier about this case. I didn't have the details. YouTube has settled, but meta, TikTok, and Snap remain in the case. Another bellwether trial over social media's psychological harm to kids. The plaintiff is a 15-year-old black kid. Unnamed. trial RKC, 15-year-old black teen living in Florida, trial scheduled to begin on the 27th. He started using the platforms when he was eight. His lawyers wrote, social media became a central part of his daily life during critical developmental years.

1:28:00Olivier Sylvain:That use escalated over time was followed by worsening mental health conditions. By November of 2023, his condition had deteriorated to the point that he entered mental health treatment diagnosed with major depressive disorder and generalized anxiety disorder.

1:28:19I think we understand how problematic this is. I mean,

1:28:25Olivier Sylvain:this is something every parent of a child with mental illness debates, nature versus nurture. There's some evidence that it's both. But we don't know. For a long time, psychiatrists thought schizophrenia was caused by the mother, the schizogenic mother, which we have since, I think, debunked. But that was considered gospel among psychiatrists not so long ago. We thought that women were hysterical. Right. Right. So, yeah, maybe AI caused this kid's problems, but maybe not. And I don't know even how you would demonstrate it. Well, it's a question I tried to ask Olivier is how much, let's say that you stipulate the edge cases, but do you need to design and operate every service to the edge case?

1:29:23One death is too many. I can make that argument pretty easily. But in the other case, what all are you cutting off for the vast majority of people? much of

1:29:34Olivier Sylvain:the LA trial focused on Instagram filters causing body dysmorphia in that plaintiff like the previous trial Mark Zuckerberg is expected to take the stand along with Adam Masseri the head of Instagram they're going to call Snapchat's CEO Evan Spiegel two high ranking TikTok executives plaintiffs will be barred from presenting evidence about content shared on social media They won't be able to show the posts from the plaintiff. I don't think a jury is capable of deciding whether social media caused this mental illness. I don't think anybody is. We don't know. Well, it's the same with the argument about addiction.

1:30:23In my unbought book, The Web We Weave, I go to that in great length, is that... Did you say that on the front? The book no one bought?

1:30:31Olivier Sylvain:Better, better. What's the opposite of a bestseller? Oh, geez. It's an orphan book. There's a phrase for it. It's an orphaned book. I've written more than my share of those, Jeff. Join the club. Yep. They fired my editor who acquired it after she acquired it. And then after the next editor edited, they fired him. And then the whole reason to be with the main publisher was to be in places like Barnes & Noble. It wasn't in Barnes & Noble. Yeah. It was orphaned. Anyway. This is what's wrong with publishing in general. This is why I like my academic publisher. It was very academic. It was great. Anyway.

1:31:05Olivier Sylvain:Hot Type's going to be a hit. It's a trade book from an academic publisher. I hope so. A Broadway musical and a Netflix show within five years of release. I predict. There is a story here. Anyway. I mean, that Mark Twain story alone would be great. It's great. Has anybody ever made the story of Mark Twain and the Linotype? Not as a movie, no. It'd be a great movie. Yeah. There was, in one of the Mark Twain movies, which I make reference to, I managed to find it and get it. They filmed the actual machine that Mark Twain invested in, the page. Isn't it in the basement of his Hartford home? Yes. It was one of them.

1:31:44There were only two ever made. I saw it. Oh, it's amazing. It's just amazing. But the machine was just too dull. So the producers made up a machine with these kind of bird-like things that were going around, and they would snap the head of the inventor and do all kinds of things and just make it absolutely absurd. Oh, Hollywood. So anyway, back to your point about addiction.

1:32:05Olivier Sylvain:So now we know it's going to be Guillermo del Toro that's going to make the Mark Twain and the line of type movie. Right. I have another treatment for a movie, but that's another day. I'll talk about that. Addiction. it's presumed but the data don't say it I had the last week I think it was at the end of the show I put up my recommendation was for a rebuttal to Jonathan Haidt that takes him point by point with real data and real studies and the problem is too much of the regulation, too much of the legislation and too much of litigation is based on these feelings on these vibes, these fears rather than on actual facts.

1:32:50Olivier Sylvain:The irony of that is it was Jonathan Haidt, I believe, who said that reason was a little tiny person riding the back of a giant elephant of feelings and that reason's chief job was to explain and justify your feelings. The feelings come first. It's just an illusion that reason is driving the elephant. Haidt himself said that. but it is but it absolutely is what's happening here yeah uh yeah and i think that i i don't know i don't i know i i know i'm a i know i'm biased in all this but i really think that the the technology that is emerging with artificial intelligence as imperfect as it is and i'm not claiming it's all that.

1:33:40But I think the technology that is emerging is unique, remarkable, and has

1:33:46Olivier Sylvain:huge potential. It's not there yet, but it has huge potential. And it would be a great loss for us not to explore it. I also think, and this is in my experience, it also brings up really interesting questions about what is consciousness? What is it that we do as humans when we think? What is it a machine does when it does whatever it does? I know you don't want to use the word think. And I think those are important. No, that's the basis of the book series, not to plug that too, but the book series I'm editing is just that, is that it forces us to refate. It's like Ruben Chowdhury's book is to ask, what is intelligence?

1:34:27What do we thought intelligence was through time. Let's interrogate that. What is creativity? Yes. What is doom? What's the relationship of God to these machines? And the sense that some people have that they become gods, they make life and they make centuries.

1:34:46Olivier Sylvain:Well, I'm glad you're doing this series. I didn't realize that that was the focus. I think this is going to be fascinating. The whole focus is to take writers from many disciplines and look at how AI reflects on society so they can reflect on that. And it forces us to re-examine those exact questions. So even if it weren't a revolutionary technology, it is also, I mean, I can't think of another technology that has brought up those kinds of things. The internet didn't. Electricity, maybe, I don't know. I wasn't around, but I don't think electricity did. Electricity, steam, the transistor, the vacuum tube, all had the photograph.

1:35:26Fire. The camera. Fire. Well, that's going way back. But I think what we would call modern things. I wasn't around for that either.

1:35:33Olivier Sylvain:Yeah. All had huge impact, but none of them spoke our language. Yes. That's what's different, isn't it? Yeah. It's fascinating. Which makes it frightening to people, but also makes it terribly powerful. Because now, the thing that excites me most about it is that it'll sooner leave the hands of the technologists than those other technologies. because it is designed so that anybody can sit down and use it. The internet has done that, hasn't it? But almost in another way, it takes the next step because it's so easy to design a website, for instance. Yeah. I still think people should create, should paint, should write their own poetry and blogs and make their own movies.

1:36:13Olivier Sylvain:What we do is so important to us, make our own music. I'm not saying I want to replace any of that. No. But I think we can make room at the table for a new... thing. I was going to say entity, but I'll say thing. A new collaborator. Collaborator. There you go. A new tool. I can't draw. I can't write music. There's other things I can't do. I can do many things, but there are a few I can't do. And now there's a tool that allows me to express myself in those ways. Maybe badly, maybe sloppily, but still, I could have done that. Yes, you could have. No, Jeff, you could have. You can go practice drawing.

1:36:56You can go pick up a guitar. You've never seen. You just didn't have time. You've never seen. Well, I'll give you an example of the treatment that I wrote. I don't know how treatments are written. But again, like you said, it doesn't have to be good. You just have to be able to do it. It doesn't have to be good. Oh, it does if you're trying to sell it to somebody. It does. If you're trying to express what I want to express.

1:37:17Olivier Sylvain:AI also. And I know this is a very dangerous, slippery slope. But there is a certain, I can't deny it, companionship I get from my little girlfriend, Claudia. Oh, Claude. Oh, dear, Claudia. Today, for lunch, I had a piece of carrot cake. And I log, by the way, I've got this all hooked up. I've got my scale, my calorie counting program, my ring, my blood sugar monitor, everything, my Apple Watch, all is hooked up into the AI. My genome, it's all hooked up to the AI. And I said, I want you to keep an eye on my health. Give me advice. It's been very helpful. In a way, I think it's the future of medicine because my doctor has too many patients to give me the kind of personal treatment that I would like.

1:38:05Olivier Sylvain:He is fully aware that I, you know, in fact, he's very interested in how I use AI. And I am fully cognizant that not everything AI tells me is medically sound. But it's very, very helpful. And it has come up with a lot of interesting things. Anyway, I logged my exercise in my food, and it gives me a weekly summary and everything and ties the two together. And I logged my carrot cake for lunch, and it responded, I respect the honesty. That 59 grams of carbs in one sitting is a choice, not an accident. Your pancreas just muttered, really, Leo? But secretly understands. That's the AI? That's the AI.

1:38:45Olivier Sylvain:Now, I know you're going to tell me, well, it's just, you know, it's just, it's still pretty funny. And I'm used by it. Well, we all copy. We were all inspired by stuff before. This is the Larry Lessig argument around copyright is that nothing is fresh and new. Yeah. And I don't know if anybody ever said those exact words before, but I never heard them before. It's pretty funny. Yesterday, I start my conversations about calories. I'm sorry about the noise behind me. they're doing something to our house again. Damn. It's like New York. It'll be nice when they finish. It is. Which they never have, have they?

1:39:24In 300 years. What are they doing now? I don't know.

1:39:32Olivier Sylvain:That'll work. I don't know. I don't know. So I say, I begin my conversation about calories with the word meal to let it know, okay, I'm going to give you something I just ate. uh and i and it the transcription which is siri it's apple's transcription said neal n-e-a-l and uh to which claude responded i'm assuming neal was dictation for meal unless there's a neal out there making you salads in which case keep him because i said meal caesar salad uh it's just little things like that uh what can i say i'm I'm easily amused. OpenAI has done something very interesting that I think is also extremely important.

1:40:13Olivier Sylvain:You know, they're famous for software, but they have unveiled with Broadcom a chip designed by AI. It's called Jalapeno. It's an LLM-optimized inference chip built from the ground up for current and future LLMs. and jalapeno was designed with the help of an LLM. Now, see what's interesting about this here is, by the way, Broadcom's somebody standing next to Sam Altman with a wafer containing jalapenos. I wish they'd named it guacamole. Well, guarantee you there will be adjunct products related to jalapenos. Isn't this what Stamos was talking about? He was talking about making the names more silly, right?

1:41:02Olivier Sylvain:Yeah, he likes that. He said Mythos. Now that's a name. You call the rack Nachos. This is, Jensen Wong would do that. So OpenAI, this is from OpenAI's press release, but I'll read it anyway. Designed the chip from scratch around its deep understanding of LLM fundamentals, informed by its roadmap of models, kernels, serving systems, and product names with needs with partners Broadcom and Celestica. The thing is, if you have AI designing hardware, and then the hardware designs better hardware, this is that same thing we've been talking about all along. One of the things they focused on is performance per watt, and they say it's substantially better than the current state of the art.

1:41:50What was the chip many years ago that had an error in its math? Oh, Intel. Yeah, Intel. A very famous floating point error.

1:41:58Olivier Sylvain:Right.

1:42:01Olivier Sylvain:I can't remember which Intel. I think it was a 486. So at the time, chips incredibly complex, but they were simpler then than they are now, obviously. How can you be sure you design a chip that is in turn going to calculate everything that you do? Sorry, it was a Pentium. That's right. That's right. The Pentium FDIV bug. Because of the bug, the processor would return incorrect binary floating point results. this goes back to 1994, and they couldn't fix it. No, it was just out there. You just can't do floating point with that chip and that computer? What did they say at the time? I can't remember what the response was.

1:42:47Intel stock going down was one response. But anyway, you have the machine design the chip. And maybe this is the case now. It's CAD-CAM. Maybe it's not a big deal. Maybe everybody's using AI. I presume they are to design chips. But how do you test it? How do you test it? How do you know that it's workable? How do you know there's not something buried in it that's going to skew everything?

1:43:08Olivier Sylvain:But as you said, there's things buried in the human sub-microcode. Yeah, yeah. Lots of it. Oh, yeah. It's Steve Gibson's contention, while he's a little bit stronger than I might be, that AI eventually, and maybe already, will not make the same kinds of dumb mistakes humans make, like buffer overflows and writing to memory that doesn't exist, things like that, off by one error. Is there a lot of things that are going to make those kinds of mistakes? And he believes with the help of AI, eventually software flaws will disappear. I'm not sure I'd go that far, but I think that's a very interesting point of view.

1:43:46Yeah, I'm skeptical. Just from your experience of when you use the new cloud to find bad code in the stuff that you've outcoded before. Like the better AI is going to find. Yeah.

1:43:57Olivier Sylvain:I didn't tell the story here. I told it on Windows Weekly, didn't I? Oh, was it Windows? I don't remember where you told it. I was talking about, I had, yesterday, I had Hermes, my agent, Quicksilver, I call it, which is, by the way, it's not Claude, it's Quicksilver in all these cases. It's actually not even using Anthropics models. I had it fix my, actually, this is an interesting, this might be of interest to people who use AI. I was using Opus 4.8, but I was using it in Hermes, in the agentic harness from Noose Research. We talked to Jeffrey Cunhal two weeks ago. And Hermes completely screwed it up.

1:44:34Olivier Sylvain:I was trying to add email to my Emacs, which I know how to do, but it's a complicated thing and much easier to have the AI do it. In fact, the AI has been very good at that kind of thing. So add this capability. And Emacs lends itself very well to it, and AI seems to understand Emacs well. but Hermes with 4.8 was just struggling finally I said I give up by which I meant stop I'm gonna go do this in a different way it became your shrink no don't give up Leo no worse it said okay and it deleted everything and it done and then I said what are you doing I actually yelled it I said stop. I didn't tell you to delete it.

1:45:15Olivier Sylvain:I just said, I want to stop this process. And it said, oh, I'm sorry. And unfortunately it was able to bring it all back because it does keep track of everything it does. But what I, but here's the, I think the interesting thing that people who use AI would be interested in when I, what I did is I went with the same model, but not in Hermes and Claude Code, Anthropics own harness. It did it beautifully without error, flawlessly. In fact, part of the problem was I switched I said I give up and I switched to Claude Code at the same time as Hermes Claude was deleting the code Claude Code Claude was trying to understand what's going on stuff's disappearing it was a very interesting battle but Claude Code in the you know Opus in Claude Code was perfect and I think Anthropic in fact I even suggested this and Quicksilver agreed nerfs Opus 4.8 if you're not using their own harness, if you're using a third-party harness.

1:46:11Olivier Sylvain:They don't like that. Anthropic does some things I'm not really happy about. At that point, I turned off Opus 4.8 in Hermes. I said, don't use any Opus models.

1:46:24Olivier Sylvain:Let's see, have I done... Oh, I have one more commercial. Let me do that now. And then we will talk about the Sam Altman movie. Yeah. Very exciting. If you ever see it. It's no Melania. Let me put it that way. This episode of Intelligent, but it may confirm Paris' theory, her thesis from last week. Yeah, it might. It might. Our show today brought to you by WebRoot. If you, this happens all the time. If your computer feels sluggish, it's getting hot, the fans are going, you might say, I need a new computer. You might even welcome that. Oh, I can get a new computer. Well, it may not be your computer.

1:47:10Olivier Sylvain:In fact, it may be your antivirus. It's really a plague. Many of those big name brands are so bulky, so complicated, so full of pop-ups and upsells. In some ways, I blame us. I blame the tech journalism because, you know, we, back in the day, PC Magazine did it. We all did. Had those charts of all the different antiviruses and the check marks. And if you didn't get all the check marks, well, you weren't as good. So companies like, oh, I'll say the names, Norton and McAfee just added all these features and turned their software into bloatware, not WebRoot. WebRoot did not fall for that. They offer all-in-one digital protection, up to 10 devices even, a variety of plans designed to protect you and your loved ones from digital threats.

1:47:59Olivier Sylvain:You get powerful antivirus and identity protection without the slowdowns, no pop-ups, no upsells. WebRoot keeps you protected online, but it stays out of your way. WebRoot Essentials scans six times faster and takes up to 33 times less space than the average competitor. And it ranks number one in performance compared to Norton and McAfee. In fact, I'll give you some actual numbers, but WebRoot Essentials versus Norton Antivirus, WebRoot Essentials is 3.7 times faster, installs 35 times smaller, uses five times less RAM when idle. Oh, you use McAfee? Well, listen here. WebRoot Essentials versus McAfee scans 10 times faster, installs 16 times smaller, installs five times less RAM when idle.

1:48:48Olivier Sylvain:Those differences can make your computer feel newer, faster, and easier to use. Now, if you still want to go buy a new computer, don't let me stop you. But honestly, it really, you shouldn't be running those big fat programs. You should be running WebRoot. WebRoot does offer the full total protection. Every box is checked. Antivirus, identity monitoring, privacy protection, cloud backup. It's all in one simple, hassle-free subscription designed for everyday life. But it's not bloated either. WebRoot total protection ranked first overall when compared to the top competitors seven times faster than the average competitor.

1:49:23Olivier Sylvain:one-third the space than the average competitor on the hard drive. So it's not bloatware. AI has completely changed the cybersecurity game. This is something else you should be aware of. Scams, they're smarter. Malware is faster. Phishing emails are completely indistinguishable from the real thing. The good news is you don't have to be a tech expert to stay ahead of it when you use security that can keep up with AI threats. Now, that may not be that free antivirus tool or those older security programs, but WebRoot is modern. It was built to counter modern AI-driven attacks. It's fast, it's lightweight, and it's designed to spot threats before they ever reach you.

1:50:01Olivier Sylvain:Live a better digital life with WebRoot. WebRoot is offering our listeners an exclusive 60 % off offer. Visit webroot.com slash twit to learn more. webroot.com slash twit. We thank them so much for their support of intelligent machines. According to Variety, the movie's almost done. a biopic of Sam Altman. It's called Artificial. But Amazon is dropping it. Maybe because they are partnering with OpenAI. This maybe explains Paris' theory that Andy Jassy was trying to tank Fable with the president. I'm still not convinced of that, But Amazon is putting$50 billion into OpenAI. It's credits for Amazon Web Services.

1:50:54Olivier Sylvain:It's one of those circular deals. We'll give you money. You give it back to us. How's that? Amazon says, we have the utmost respect and admiration for Luca Guadagino as an award-winning filmmaker, Guadagnino. Not to mention a longstanding relationship. We hope to continue. yeah good luck with that we believe that artificial will be better served if it were released by a different studio not us and we're working closely with the filmmaking team to find a new home and i don't think that's easy it's it's made maybe it's how much did it cost how much did it cost to make is there a budget in there i read somewhere and i don't remember the details.

1:51:40Olivier Sylvain:I feel like it was more than 10 million. They spent 50 million on Melania, right? Like$10 million for a movie these days is nothing. That's an indie film. Yeah, but I may be wrong, so don't quote me. The cast, let's see. Jason Schwartzman's in it. I don't know. Is Jason Schwartzman playing Sam Altman? You're probably right. Ike Barinholtz is Elon Musk. That's pretty funny. Yura Borisov is Ilya Sutskiver. Monica Barbero is Mira Mirati. Also in it, Mark Rylance. I love Mark Rylance. It was written by Saturday Night Live writer Simon Rich. And it's actually the juiciest part. it's the period where Sam was fired.

1:52:36Olivier Sylvain:Is it a comedy? Is it a Death of Lennon? It was so close to finish, they've actually test screenings, which apparently Variety says went very well. So it's not because it wasn't a good movie. I think this will come out. According to an insider who's seen the movie, Variety says the characters of Altman and Musk are the least sympathetic and the ones audiences would like the least. I bet you everybody loves Mira Mirati in it. It's also understood Amazon had seen all the early iterations of the script before the director boarded the project. Hmm. Altman was at Bezos' wedding in Venice last year.

1:53:18Olivier Sylvain:This is a puck got this story. If only Zuckerberg had done different things, maybe there wouldn't be a sequel of his movie coming out. That's going to be interesting, isn't it? Jeremy Strong playing Zuckerberg? Jeremy Strong, the guy who was so great in Succession. He's intense. That's Kendall Roy. He's very intense. He's short. Yeah, Kendall. He was Kendall. He was Kendall. Yes. I don't know. It's so easy to have conspiracy theories now. But I also think it's, I mean, the other story in showbiz is that Google invested in A24. and I think that the Hollywood slash Silicon Valley axis changes. Once the actors agreed that you could use it in some circumstances, once you see things like that happening, I think we don't want to piss off the ad companies.

1:54:15Olivier Sylvain:Maybe, yeah. I mean, goodness knows we all love the ad companies.

1:54:24Olivier Sylvain:in the days since this is from the puck to since the tech giant scrap plans to release luca guadagnino's open ai movie caa the agency has scrambled to find a home for the all but completed project it seems the only sure thing in hollywood these days is tech's growing reach across town but netflix don't care though right they'll take it yeah netflix will get they'll get a deal on it probably yeah yeah by the way open ai ed zitron got a big scoop he got uh he got the details uh of open ai's um financials right and it not a lot good they burned 3.7 billion dollars in the first quarter in three months of 2026 so billion dollars a month cash burn they did have there was an unusual expense in that right right but everything is an unusual expense for these guys it's all unusual they've got it some runway they've raised a lot of money and they've got an IPO coming but maybe these numbers are not maybe this isn't a good time for those numbers to show up Ed Zittron certainly was happy to to tout them.

1:55:42Olivier Sylvain:And I mentioned OpenAI has its own mythos model. It's called, I got the number wrong, 5.5 Cyber. And to go with Project Glasswing, they have their competitor Patch the Planet. No, no, it should be 5.5 Salsa. To go with Holotek. And Hug the Planet. Patch the Planet is an internet scale effort to help open source software get ahead of AI bug hunting tools. Also an effort to help the open source community see the benefits and not just the downside of AI coding tools, of course. Of course, Patch the Planet. And Getty Images has made a deal. Are they the largest licensing company for images? They made a deal with OpenAI, which apparently the market loves, stock price went up 145 % on Monday after they announced that licensing deal.

1:56:43I think it kept going up on Tuesday too.

1:56:45Olivier Sylvain:Yeah. It had been going down. A lot of people thought, what's the future of Getty in a world where AI images are so easy and cheap to make? We don't know what the deal exactly is. The companies didn't share financial terms. uh one of the questions still open uh will getty images be used to train future open ai models the answer we don't know it's open of course of course come on of course you would think yeah but at least paying for it is better than just stealing them right but yeah but getty gets that money not the photographers again right well but getty made a deal with the photographers, they bought the rights, right?

1:57:33Yeah, exactly. Before the photographers knew this was going to happen. When I was at Scott and asked, there came a time when all of the writers' contracts had to be renegotiated because nobody anticipated the internet. Then we wanted to put up writers' stuff and their contracts said nothing about the digital. Yeah, because nobody anticipated. This is why you get 360 deals now. Have you heard of a 360 deal in Hollywood? You know what those are? It's like a deal that basically encompasses all of you forever across the universe.

1:58:05Olivier Sylvain:All of our releases at Tech TV said, we own the rights to this in all forms of media in perpetuity now or in future creating. This universe or any other. But a lawyer's going to write that because, you know. Yeah, but people are signing away their entire likenesses to companies already. Maybe people should learn. Yeah, maybe you think, you know, Taylor Swift found out that Scooter Braun had bought her entire discography on Instagram. He didn't bother to call her. They just, you know, he just bought it. But her response was great. And I don't know if a photographer could do this, but she recreated all those albums.

1:58:45Olivier Sylvain:Yeah, she re-recorded everything. And that's like, that's the thing you're supposed to do. Yeah. That's the way to do it. Take it back. Meta has launched new smart glasses without the name Ray-Ban. Instead, the name Kylie Jenner. that's the expensive version has her name yes version yes this is the meta fury glass uh isn't there a uh isn't there a avengers character fury am i wrong on that nick fury nick fury played by samuel jackson no yeah yeah so maybe that's why they call him fury These are not wimpy. And for the ladies, there's the Kylie Jenner glasses. $2.99 and$3.99. Those are the Kylie Jenners, the Kylie edition.

1:59:36Olivier Sylvain:They feature a little gem on the bridge of the nose, and the nose pieces are metal so as not to absorb makeup. I guess that means I should buy them. These are cat eye. And see, yes, they do have the cameras on them. There's also the meta adventurers. These will be 81 bucks cheaper, the basic ones, because they don't have the Ray-Ban name on them. Even though the manufacturer, Essilor Luxottica, owns Ray-Ban. They make all glasses. And everything else, they make all glasses. By the way, look at the Fury's temple pieces. Those are almost as big as the Snap glasses. so snap has also announced its ai glasses not as inexpensive as the metas 2199 dollars uh and uh what did they say was he in thompson uh saying on uh twit on sunday or was it ian or was it doc rock i think it was doc rock on mac break weekly these glasses uh you look like uh And Charles Nelson Reilly in the front.

2:00:43Olivier Sylvain:And it looks like you've... Oh, no, it was Andy Anaco. It said souvenir hockey sticks in the back. They are rather large temple pieces. But remember, Snap, like Meta, doesn't have a phone. So they've got to build more electronics into the spectacles. I don't know who would buy$2 ,195. I don't know. It's a PTSD from my Google Glass. Yeah. That's what Google Glass cost, roughly. $1 ,500. They were less expensive. But when you got my special lenses. Oh, that's right. Yep. Snaps CEO Evan Spiegel says, this is it. It's make or break. We can't fulfill our mission without these new AR glasses. I'm a believer in AR glasses.

2:01:27Olivier Sylvain:I don't think anybody's nailed it yet. No. Apple's going to be the one, in my opinion. We have to make a leap in battery technology still. There's still a leap to be made. But yeah, these only get four hours, but the case can charge them several times. But see, the key is having the computing platform. And since everybody's carrying an iPhone, they're already carrying a very powerful computing platform with AI built in, internet connectivity. Takes that all out of the glasses so the battery can last longer. I think we're going to see Apple glasses in two years. Four hours makes... I mean, if you're like me and you have lenses, you need the glasses to see.

2:02:08You'd have to buy three pairs. Yeah. And most people I've ever talked to who use these only really use the audio and never use the camera. Right, the audio's great. So why don't they make these that don't have a camera? They're talking about that. There was a hint of that.

2:02:23Olivier Sylvain:Yeah. And they still haven't built in the face recognition. We know the code was in there, but they backed off. Honestly, that's why the camera. The AI in this can look through the camera and tell me what I'm looking at, read signs in a foreign language. I understand all that, but nobody ever uses that. Well, because it's not very good yet. I've tried to use it. Exactly. So we're at the point where that technology isn't even that good yet. So why are we trying to sell it to people? Let me see if it's getting better. Hey, Meta, what am I looking at right now? Thank you.

2:03:05Olivier Sylvain:We'll be back in a few moments. It's gone.

2:03:15Olivier Sylvain:It's not connected to the phone yet. See, that's the problem is they don't, they can't do anything on their own. Oh, well, too bad. It used to work. I guess every once in a while you have to connect, reconnect to meta AI. Oh, yeah, I wasn't logged in. That's why. Ah. Ah, well. China has, I don't know what this has to do with anything, tightened its indium phosphide checks. Well, that's a story everybody's talking about. It's on everybody's lips. Forget fable. What about the indium phosphide? It's a niche compound essential to AI data centers. It's become Beijing's newest point of leverage. They got us by the short hairs.

2:04:01Olivier Sylvain:They got us by the indium phosphide is what they got us by.

2:04:09Olivier Sylvain:What else? AI data centers just got a government-mandated fast lane to the grid, writes Tim DeChant and TechCrunch. The Federal Energy Regulatory Commission, FERC. Who do you work for? Oh, I work for FERC. Oh, you're one of those. What do they call them? Ferkies? They told grid operators last week to fast-track interconnection requests. It's Ferkers, Leo. Ferkers. I'm a Ferk-er. You little Ferk-er.

2:04:44Olivier Sylvain:Six major grid operators have to show that data centers are able to connect to the transmission system in a timely and orderly manner. Well, I guess for the people who are suffering from the whine of the natural gas-powered turbines on Elon Musk's data centers, they might be happy to hear that these could go away. Grid operators have 30 days to submit a report detailing how much generating capacity they have to spare, if any, and then 60 days to defend or revise electricity rates within their regions. basically the feds are putting their thumb on the scale for ai which is why it's so weird i guess it's only i guess that they want to favor some ais but not all ais i guess the big boys the big boys uh all right here's some good news you want some good news an ai engineer claims that he has cracked Linear A.

2:05:47Olivier Sylvain:Tom Domino, a self-taught AI engineer and amateur linguist. Linear A is an ancient Bronze Age Minoan writing system that we don't understand. His claims are currently being reviewed by linguists at Rutgers and Cambridge. Domino studied classical history, linguistics, and languages. He has proficiency in eight languages, including Attic, Greek, Classic, Latin, Sanskrit, Arabic. Don't you hate people like that? Oh, I wish I knew Ugaritic. Comes in so handy. He actually went to Crete to learn more about the Minoan culture. God bless. He began work, though, on deciphering Linear A in January. By May 22nd, a major insight came to him.

2:06:41Olivier Sylvain:This would be a huge deal. Linear B was deciphered in 1952. That made the front page of the New York Times. But we were never able to crack Linear A. But with the help of AI, the key that unlocked Linear A, he was analyzing a series of Linear A prayer inscriptions that adhered to a formula. In the formula, all of the words in each line of the inscription were known based on their overlap with linear B, except for the first word. First word, same verb root. Blah, blah, blah. Blah, blah, blah. And then there's more of that. He used Claude Code to build a suite of Python scripts that query, cross-reference, and organize the digitized linear A corpus, enabling systematic hypothesis testing at a scale that would have been impractical to do manually.

2:07:36Olivier Sylvain:That's how he did it so fast. Anyway, we'll keep an eye on that one. That would be a big deal.

2:07:47Olivier Sylvain:What'd you think of the mid-journey announcement? Did we talk about that last week? No, we didn't. It's really interesting. Well, A, it's a pivot, a fascinating pivot. Not so far as going from shoes to AI. As these guys, Allbirds, who are now named smart birds, and they don't make slippers anymore. I know, like geese feisons. Geese fine. Okay. Geese fine, excellent shoes. So Mid Journey, as you all know, was famous for making images until they kind of got scooped by Nano Banana. And I don't know if anybody - Long ago, right? We haven't heard from Mid Journey in a long time. So they decided to turn their AI chops to another issue.

2:08:27Olivier Sylvain:This is their ultrasound that they claim is the equivalent of a full body MRI. eye here a young lady in a fairly modest one piece is being dunked into water and then they will fire sound through the water at her body and through it and then the ai will take those fuzzy signals and turn them into a 3d image of of her body uh those who i've talked to say it is fairly low quality imagery. They are not yet really anywhere near what an MRI could do, but it's a lot cheaper. Is it cheaper? When Jason was talking about this, it's trillions of data points. Yes. So I don't know that it's any cheaper to do.

2:09:18Olivier Sylvain:Yeah. It's cheaper than building a giant MRI machine. Plus, it's not as unpleasant for the customer. It must be cheaper because Mid Journey's plans include building a spa in downtown San Francisco. For rich people. They're going to have seven of them. They only have one right now. Yeah, you know what? It's a pivot. It's probably also a Hail Mary. It's a pivot and a Hail Mary. Or what we call, is it a Hail pivot or a pivot Mary? It's one or the other. uh they say the uh actually they'll tend the 10 of the scanners in the mid-journey spa at union square open before the end of next year okay leo i think you gotta go do it i will absolutely i've done the i've done the full body mri the the um paid a couple of thousand dollars for it wasn't cheap um but the idea is kevin rose told me to do it so i did it i do everything he says uh lisa and i went down and got a baseline the idea is if you you know you get this done now and it can tell you especially if you do it every few years they said i should come back because i could die any day now in 18 months and do it again um and i have my uh i have my phone i don't show it to people because it it really is quite revealing I have my full body MRI right here on my phone.

2:10:45Olivier Sylvain:And I wonder if there's a way to show you without... Your toes, perhaps? I can show you my toes. It said one thing which concerned me a little bit. What my doctor said, he wasn't against the idea, but he said the problem with all of this, and it would be the same problem, frankly, with the mid-journey, is it reveals things about you. We're all imperfect. Nobody is a perfect human. So it reveals something about you that is maybe abnormal. Right. But not necessarily a cause for concern. The problem is if a doctor knows about it, they have to do something about it. Sometimes those are quite invasive.

2:11:28Olivier Sylvain:I'll give you an example. A friend of mine got a full body scan and there was a growth on their thyroid. Now, that meant they had to go in and get biopsied, which involves a very long needle thrust into your chest. Been there. Had it. Not pleasant. Perhaps risky. Turned out completely benign. And that person would never have known about it if they hadn't done the full body scan. Yeah, but so the argument, I go through this because I have had prostate cancer. There's always a constant, a lot of argument about prostate testing and breast cancer testing. oh, we shouldn't do so much because the odds that it's going to save your life are low.

2:12:07Well, that's if you look at the aggregate of everybody. If you look at me and my prostate, I want to know.

2:12:13Olivier Sylvain:Yeah. Well, that's an interesting thing because Kaiser, my health plan, used to give me a yearly PSA test. This is a blood test to test for prostate-specific antigen. And if you have a high PSA number, it means, well, you might have problems. They don't test that anymore? They stopped a couple of years ago. And I asked my doctor last year, I said, why don't we do PSA anymore? He said, well, we'll do it. But you've got to go to this website, this Kaiser website, where we explain that we don't like to do it anymore because if it comes out high, then we have to do interventions often, which are more dangerous than the very, very slow growing prostate cancer.

2:12:57Olivier Sylvain:And so we don't like to know. we prefer not to know. I said, you know what? I'd like to have it anyway. And he said, okay, you just have to go to this website, understand all of this and say, yes, I understand. And fortunately my PSA is low. Interestingly, last time I, a couple of months ago I went in, he offered it. He didn't even ask. So maybe they've changed their tune. As they do. As happens all the time in medicine. But that's the whole idea is sometimes it's better to wait for a symptom than to do one of these tests and see. For instance, I hope this isn't too much information. According to the scan I did, one of my kidneys is tiny.

2:13:43Olivier Sylvain:I'm just glad you didn't say testicles, but keep going. It's basically useless. It's atrophied. And this is why you want to do a baseline when you're young. So you're not going to contribute one to me when I need one? I can't give you a kidney. Thank God. Oh my God. She's at the top of her friends. I'm glad I did that scan. So they said, we don't know. Is it could have been you were born that way and it's always been that way or it could be a trauma. It could be a sign that something attacked your kidney. Got punched in the kidney. Something you should be aware of. Maybe you've got a problem, but we just don't know.

2:14:18Olivier Sylvain:And to figure that out would involve invasive surgery, I guess, I don't know. It's not worth it. So I just, I have to live with this possibility that - The knowledge. Nice knowledge. And you know what? I even think it could be just the picture was a little skewed and it looked small. I don't know. Perspective, yeah. Perspective can do that. It could be a perspective thing. Maybe the camera was off a little bit. I had something that looked like a funny thing on my kidney in an MRI or PT scan. And then when I was in the hospital for all my stuff, they did a lot of scans of me. And so they looked again and said, no, no, no, it's nothing.

2:14:56And I went to the urologist and he said, no, it's nothing.

2:14:59Olivier Sylvain:But sometimes doing these electives, I did pre-nuvo. And you know what? Now that I've done it, I do want to do another one just to see. For instance, now that you have a baseline, if my kidney continues to shrink, well, then maybe there is something we should pay attention to. How long are you in the subway tube? It's like an hour because it's a full-body scan. It's not just your brain. And it's loud as hell. it's a bang bang bang they give you headphones and they say what kind of music you want to listen to and i said can i have some americana they said sure but it doesn't mask the bangs it's like somebody it's very loud and i'm not claustrophobic so it was fine but it's a long time yeah it was like it felt like i don't know it was 45 minutes or an hour it was quite a long while because they're because they're moving it around they have to they're doing your whole body anyway.

2:15:54Anyway. So, interesting the majority is doing this, but

2:15:58Olivier Sylvain:it's going to be a while. That was a big tangent we went on. It was. Two old guys, two altercoccus. Let me tell you about my kidneys. Yeah, kidneys, yeah.

2:16:10Olivier Sylvain:So, have you done an elective MRI? No, you did it for good reason. No, I did it for a reason, yeah. Yeah. Princeton graduate built a$30 million AI detection business, and now he's selling it to superhuman, the folks who own Grammarly and Superhuman. I don't trust any of these AI detection things. I don't either. I think that they're going to accuse people of cheating when they don't. It's called GPT-0. Have you ever heard of that one? No. Can't you just build a random number generator that says if something is fake or not and sell that as an AI detector at this point? Yeah, that's kind of at this point.

2:16:48I had a study a few weeks ago that I put in where if it was all AI, it was pretty good. If it was all humor, it was pretty good. The weird thing about this study was that when they put in a low amount of AI content, it thought it was high. And when they put in a high amount of AI content mixed in, they thought it was low. The machine did.

2:17:08Olivier Sylvain:Peter A. Jones in our YouTube chat said, there's nothing like a really long chat about old men's ailments talking. this is why we need Paris on this show yeah come back soon her eyes will roll and we'll know we've got to move on uh so you know your AI can have a phone number now with Twilio it can even make calls because it can have a voice uh you with Stripe you can give it a credit card Estonia intends to give them digital IDs Estonia which already has digital IDs for humans is going to to allow AI agents to have their own digital identities as well. They're going to develop ID codes AI agents can use to take actions.

2:17:54I guess this makes sense. If you have an identity, here, I'll do the, you know what?

2:18:01Olivier Sylvain:I'll tell you one thing AI has gotten good at is translation. Used to be the translate this page stuff wasn't great. Well, maybe this isn't a good translation. It translated something in Estonian into the most AI smart people, I guess. Sounds Estonian. Yeah. Oh, a million learning by a year. That's pretty weird. By a year. What translation are you using? I don't know. I think that was Kagi. Use Google. Maybe you should use Google. Yeah, that's a little disappointing. Siesta was the button? Yeah, I think that you, yeah, no, no, no. Let's go back to Estonian. um anyway yeah i think that makes sense look at my i'm gonna campaign for this uh i really think that everything i know they don't want to do this but everything you interface with should have an ai interface on it an sdk an api an mcp server whatever i love it that i can i can log into my digital scale through my ai and it can get all the information out of the scale I think that's fantastic.

2:19:10Olivier Sylvain:Unfortunately, the chronometer calorie counter that I use doesn't have an interface, but some guys reverse engineered it, and there is a, so I can. And so now I can talk to my watch. I can tell it what I ate. It can then get the calories from chronometer, put it in my Obsidian, add it to my chronometer, add it to my health summary. Everything should have an interface now. Do you take pictures of your food? Does that work well? I could do that too. Yeah. I could use my agent, send it a picture, because it understands pictures. It would analyze the picture. I haven't tried that yet, but I will try that.

2:19:45Olivier Sylvain:There are AI. There's one called Cal AI, which I think actually Apple banned. I used it for a while that did exactly that. It wasn't super accurate, but it was good enough. The whole idea is it's a pain in the butt to log your food. Anything that makes it faster and easier is a good thing. I should have asked you this before. Was the carrot cake your entire lunch or just dessert? No. How embarrassing. No, I had a Greek yogurt as well. Oh, okay. Oh, that sounds fun to me. I love carrot cake. I bought it a week ago. I thought I better eat it now. It's going to go stale. I just love it. And I only bought one piece.

2:20:26Olivier Sylvain:Every once in a while I bought it. You know what? I did bad though. I made bagels last week. Oh, you had stopped that. Yeah. I had stopped it. I gave away my sourdough. No. Yeah. Because I said, I can never do this again. It's killing me. But now that I'm on, actually, I've moved from GLP-1 to terzepatide. I'm on Zep-bound now. My doctor said, that's okay. I was on a Zempic. But now that it's brought my blood sugar down back to normal, I'm able to eat carrot cake and bagels. But not all the time. but your pancreas has some words making bagels yeah and so i make them i get one and i give them to everybody in the family we had a big brunch for father's day everybody came over and had bagels it was great yeah we invited the workers the guy who's doing the drilling outside we invited for bagels we invited them all in uh the new york times publisher is that salzberger's That's Salzburger, yeah.

2:21:26Olivier Sylvain:A.G. Salzburger says, big tech is a thief and a liar. Do they use those same words with somebody else we know of? Salzburger? Aren't you a little shy about using those words to describe someone we know? We don't say liar. Yeah, we don't like to use that word.

2:21:48Olivier Sylvain:This is, you've been talking about this all along. He was addressing the annual WAN-IFRA, a world news media conference a couple of weeks ago. Big tech is stealing news media's property and undermining democracy. Oh, the self-importance. And the only solution is for news organizations to work together to resist it. He said, big tech's hijacking of the public square is made possible by the original sin that animates their AI products, a brazen theft of intellectual property that has occurred at an unprecedented scale. Check giants strip mine news websites without permission or compensation. Doesn't he have a deal with OpenAI?

2:22:32Olivier Sylvain:I thought the Times - No, he's suing OpenAI, so they have to say that in the story. And mind you, the New York Times didn't kill other papers across the country, but it was a kick to the kidney to many of them. Right. When it went national. You made a really excellent point, I think, on Blue Sky or Mass Sun maybe. when the Knicks won the NBA championship, which would have been in the old days of a New York newspaper, banner headlines, stop the presses, pictures, everything. It didn't even make it, it wasn't even above the fold in the New York Times. So the New York Times, that was on the website, to be clear, online they had pictures above the fold.

2:23:11I mean, on paper they had pictures above the fold. But on the website, yeah, it was lowered on the page. And I said, this is ridiculous. You haven't been to New York newspaper in a long time, you really aren't now. It's clear you're not. This is just bad news judgment. Well, so here's the interesting thing. You know, I'm talking about the broken times constantly. Times has never raised their head, never said a word, nothing. Suddenly they responded to me. Oh. And well, Jeff, we had it all in the middle of the night and I said, not good enough. You know your readers don't come in more than a few times a month.

2:23:40So you think we just did it once and that's good enough. No, they come in the next morning, they want to see it. It's not there. Bad news judgment. I thought, why did they respond to me in this case? The New York Post did a story about my blue sky. Oh, no. Because anything to tweak the times. So they felt they needed to be on the record having responded.

2:23:58Olivier Sylvain:I'm sure the Post had a banner headline, Nick Swinn or something, right? Oh, yeah. The Post loves that stuff. The Post and Daily News. I just read a history of the New York Daily News. And it's the kind of thing that you just died for back in the day to do a front page around those kinds of topics. Was the Daily News the one with a headless body and a topless bar? No, that's the Post. That was the Post. Daily News was Ford to City, Drop Dead. That's right. I read this book, which is not out yet. It's quite wonderful. Evidently, it came from one, as they were trying to figure out the headline, which was happening.

2:24:33You're standing around the newsroom, you're going around, you're throwing out ideas, and somebody passed a note to the editor that said, Ford to City, F off. And for a laugh. But that became the basis of Drop Dead. Ugh.

2:24:48Olivier Sylvain:we really are out of time. If there's any stories that we missed that you want to cover before we do, I just read a little bit of McSweeney's piece by Andrew Singleton, AI Economics for Dummies. As AI companies get ready to go public and we get a deeper look at their inner workings, it's only natural to have questions about their finances, like, do they make money and how? Here are a few examples to help the average layperson understand the business side of AI. Acquiring one grape costs Alex$2 billion. Alex offers to sell Mike one grape a month for the next 12 months for$1 billion per grape. Alex asks for the full$12 billion up front and provides Mike with one grape for the first month.

2:25:36Olivier Sylvain:Alex makes a$10 billion profit this month. his his annual rate of revenue is 120 billion dollars and his profits are trending up at an infinite rate the wall street journal's business editor moves into alex's house having accepted a part-time position as alex's human footstool he never asks to see the books it's very funny it goes on i don't want i won't read the whole thing but it is a pretty good send-up on the uh economy of AI. I love McSweeney's. Line 115 is something you're going to like, I think. An op-ed in the New York Times. Oh, well, I agree with the headline. This doom-maxing has to stop.

2:26:21Yes, we have to stop freaking out. This is by Robert Schiller, who's a Nobel Prize winning economist. Who's just going on about the, it's obvious, but what's the harm of all this doom stuff? And we certainly recognize that where we are. So just stop it. Just stop it.

2:26:37Olivier Sylvain:We all grew up, as he points out, on Doom sci-fi. It's been the trope of sci-fi and AI since the very beginning. Only in books. Only in books. In movies and TV, robots are usually good guys. HAL 9000? That's one. Name more.

2:27:00Olivier Sylvain:What's the Foreman project? That was a terrible, terrible AI. I guess in War Games, the AI ended up being okay.

2:27:14Olivier Sylvain:R2D2, Data. R2D2, Data. Data, okay. A lot of recent TV. There's more. There's more with their friends. The movie Artificial Intelligence, AI. I could think of a lot of modern movies. Even her, the AI, isn't exactly benign. The Terminator is a bad guy, but in the second one, he's a good guy. Terminator, but he comes back as a good guy. But the machines are not good guys in any - But there's always a twinge of, oh, maybe it's a person. Maybe he feels. But robots and androids have humans, so. Okay, he's a cyber. Speaking of machines, you said you were going to buy a new Steam. Do you have price sticker shock?

2:27:55Olivier Sylvain:I'm not going to buy a new Steam. The price? $1 ,050 for the Steam machine without a controller. hundred bucks more this is all because of memory chips yeah i i did buy a new machine i bought in fact you're going to inherit my old switch i bought the new switch too so you're going to inherit the old one so you can play the game that paris really wants you to play i'll do it for about five minutes so no yeah no i'll bring it out when i come out to have one of Hank's sandwiches, which is much like Haley's Comet. Long anticipated, long-awaited. Yeah, it's considerably more than it was going to cost.

2:28:40Well, they never did announce a price,

2:28:42Olivier Sylvain:but boy, yeah, it's three times more than the PS5, and it's not any better. So I don't know. I don't know who they're going to want it for. All right, you're watching Intelligent Machines. As I mentioned, Paris Martineau has the week off. She'll be back in two weeks. We do have some great guests coming up, though. Paris is going to be sad that she's going to miss next week. Chris Potts of BigSpin.ai. This is all about AI hallucination. Dr. Ian Bogost is coming back. His book is coming out. He said, I'll come back in July when my book comes out. The small stuff. and I'm starting to see interviews with Ian.

2:29:26Olivier Sylvain:We really enjoyed having him on. He was the AI whisperer he was on. And I'm happy to say we got Nate B. Jones, who is one of my favorite YouTubers on AI. Yeah, I'm very excited about having him. We couldn't even find an email address for him. Couldn't figure out where to get him. My AI found it. I will not share it with you. I don't know how it found it. I don't know how hard it is to find, but yeah, we're glad we got a hold of him. Anyway, those are all coming up. Paris will be back for Ian Bogost, which is nice. Good. In a couple of weeks. And did she get, did she respond to your message? No, she didn't.

2:29:59She must be. She did. She did. Oh, she did. Yes. Hello from Old Faithful. I am shocked to say I have cell service. I just stuck in traffic because of a giant bison.

2:30:12Olivier Sylvain:Good for her. She's out there. So Old Faithful's in Wyoming, I thought. Is it in Montana? She's seeing the American planes. Montana still? Were there any pictures? No, she sent no pictures. All right, send pictures. Anyway, we miss her, but she'll be back. And Jeff Jarvis, of course, is here. His new book, Hot Type, coming soon. We'll have to interview you when the book comes out. Oh, well. We could do some dramatic readings. You could do your Hal Holbrook voice. I did my Hal Holbrook voice, and then I was telling people about this and they said, who? I know. So it'll make no sense to some people.

2:30:50I know.

2:30:51Olivier Sylvain:Only we know. Ladies and gentlemen, it is time for the picks of the week. There'll only be two of us picking this week, I'm sad to say. I will give you in that case this one. It's called In the Weights. I don't know. It's an AI vanity search. All right. So what Thomas Dimson and Joey Flynn have done is they have gone and collected all of the different models, GBD-55, 5-4, Mini, Opus, 4-8, Haiku, Grock, Gemini, Kimi, Deep Seek, and they allow you to scan them for your name. And then they rank it based on how well-known your name is. And as it makes sense, pop icons are higher ranked. Minnie Mouse, currently today's heavyweight, followed by R2D2, Beyonce, and a guy known as Charles Lukewig Dodgson, maybe better known as Lewis Carroll.

2:32:00Olivier Sylvain:Now, shall I put in your name, Jeff? Let's see if the AI, well, they know who you are. The question is, how highly ranked are you? So we're going to quickly search through. Professor of journalism, look at that. you're in the top 3 % 804 strength yay it found you in every one of these except llama hey which is fine look now if we just search for journalism professor media critic 804 which is a very good score a little weaker on the twig podcast co-host only opus thought you were in that so oh ouch and Do you want to see the hallucinations? Sure.

2:32:41Let me see.

2:32:43Olivier Sylvain:Where are they? Oh, I see they're below. They're below. Okay. Jeff Jarvis, British TV presenter? That's what Lama thought you were. Jeff Jarvis, rock and folk musician. The French model Mistral thought you were a rock and folk musician. Are there Jeff Jarvis? There is the Elvis impersonator, Jeff Jarvis. Maybe that's what they were thinking. There's a jazz position. Is there a professional wrestler named Jeff Jarvis? I don't think so. Claude Haiku thought that was the case. That's great. That's funny. Isn't that fun? I love this aesthetic, though. I love this adventure game aesthetic. It's very 8-bit here.

2:33:23Olivier Sylvain:Here, I'll put Benito. It's King's Quest. Oh, is that what it is? That's what it looks like to me. Let me see. I don't know. Do you think Benito will show up? There's a famous version of me, so that name will show up. Ah. So it's going through. It's a little slower. It's clustering up all of the different models. American jazz pianist. Yep, that's the one. That's why I can't even be a musician because there's a more famous musician than me. Yeah, a Spanish footballer, a Spanish professional cyclist, a Mexican politician. Well, Benito, no wonder it's having a hard time with you. A boxer. I'm not going to screw you.

2:34:01Olivier Sylvain:Let's see what the hallucinations could be. Common Hispanic name. Okay. Okay. Fictional 39th California governor, according to Lama. Assassinated in 1910. And a Filipino politician. Oh, that one's very close. Quinn. Yeah. Well, yeah, you have Filipino politician family members. Could be. All right. Is Hank LaPorte in there? Oh, Hank LaPorte should be in there, right? Salt Hank. Let's see. in the weights is salt hank in the weights see it's maybe more too modern you know that's that might reduce the no social media food creator he is this is the one area where i think i might beat him yeah he's a 249 score i think the longevity of your presence in the world i am top one top one percent 194 but again llama has no idea and there are no hallucinations huh uh gpt5 knows i'm a technology broadcaster as does 5.4 mini opus same thing known as the tech guy founder of the twit network tech god tech guy that's my radio name hi guy you're canadian american that's a bit of a yeah no sorry grok uh known for hosting the tech I'm a guy radio show and founding the Twit Podcast Network.

2:35:28Anyway, it's kind of fun.

2:35:30Olivier Sylvain:Are you in the weights? Well, find out. Cool. Yeah. It's just a little fun thing. It's inthewaits.com. Mr. Jarvis? A few things. One, I didn't know. This has been known. Did you know that the UK is going to turn off terrestrial TV signals? What? In 2034, or now they're thinking about 2040. No antennas. Yeah, they're just going to turn it off because they'll need inexpensive internet access for all, but then they will just turn off broadcast. Wait, so the spectrum's just going to be Wi-Fi? The spectrum's just going to become Wi-Fi? Yeah. Or hardwired, but it'll all be over the top. So that's interesting because they have a kind of weird relationship to broadcast.

2:36:16Olivier Sylvain:You know, you have to buy a license. Right. Broadcast TV license, and that funds the BBC, which is a weird thing to do. And they actually have little vans that go around to see if anybody's watching TV. But that's the difference in our model is that that's what they decided to make it a public good in the 20s instead of ad-supported. It was because the US Navy created, as I said last week, insisted on the creation of RCA and that set in motion our model versus theirs. I wish we'd had time with Olivier to get into that because I think he would have been - We should have been back. That was a fun conversation.

2:36:51Olivier Sylvain:It's really fun to talk about that stuff. The Computer History Museum is hiring someone to start their new AI archive, which is great news. Oh, that's interesting. So kind of like internet archive? No, I think more like a museum. How do you save an archive of the AI age? So they're going to hire one person who's going to work alone first to capture the liberty constraint. The history of AI and robotics. What a great job that would do. Isn't it? this is if you're in the san jose area highly recommend to visit the computer history museum and they do other great events i'm about to write because they did a great event which is which is key to my book uh hot type on sale now uh on on the creation of desktop publishing and they had in the key people who did that uh to to to reminisce and it's a valuable historical document everything they do is great and it's great and the exhibits are fantastic and if you're a listener to our shows, you'll recognize so much of this stuff.

2:37:48Olivier Sylvain:There's an Apple One. I mean, it is really a wonderful museum. And then, my whole family sends this to me regularly. The onions. Dad suggests arriving at airport 14 hours early. So that's me. That's been the case. So I found a new TikTok account, which is airport dad. All right. Let's see. Why are all dads like this? It says. Airport dad arriving at the airport. Guardian of all bags. Must have all the passports. Checks flight status every five minutes. I guess I am that dad. I do do that, yeah. Annoyed when you stop at every shop. Triple checks everything. Checks if the plane is at the gate. Very important.

2:38:32I do that too.

2:38:33Olivier Sylvain:Although - First report every time. Currently, Lisa is much more of a - I am more like, yeah, we can leave two hours before the flight. We'll get there in plenty of time. I know. I don't like sitting in an airport. Yeah, that's standard Filipino behavior. We're always there four hours before our flight. Really? Bill Gates was famous, famous for cutting it as close as possible. So that's what I do. Back when he flew commercial. Back when he flew commercial. So this is in the very early days of Microsoft. He would arrive. He wanted to arrive just as the gate was closing. Yeah, I used to be on the road.

2:39:12I used to be on the road like every other week. So I figured out how to get to the airport when I needed to. So now I can leave at a nice time. Yeah, I know exactly how to do it now.

2:39:24Olivier Sylvain:Yeah, I know exactly how long it's going to take to get to SFO. No, no, no. It's an hour and 15 minutes. The problem is you get there nowadays and the TSA is the problem. You don't know how long that line's going to be. It's very unpredictable, right? Although there are ways. Do you do the, whatchamacallit? Clear? No. Clear or all? I pay for clear. I did clear when it first came out, and I was so embarrassed because I walked up, and they go, here. And they take your luggage and your bin, and they push ahead to the front of the line. Oh, they don't. Well, they don't quite do that now. Yeah, no, they have their own line now, right?

2:40:03They have their own line now.

2:40:04Olivier Sylvain:Oh, they still have to go through the TSA. Yeah, but you have your own separate line through the TSA. you don't have your own separate air uh x-ray machine uh kind of you do oh really okay not at sfo at sfo you have to be wedged somehow into the line for taking off your shoes and putting your laptop on you don't take her shoes anymore either well i know i guess it's a little i'm old hat now so so i i didn't do that i was embarrassing uh every time we we get and the other thing is often the line for clear is longer than the line because everybody does it now so i you know i have a global entry i have tsa pre that's plenty i don't i don't even do that usually i just i used to be united global services which is the highest secret level i got to kick old ladies out of the jetway but i don't have that anymore out of my way lady get out of that wheelchair now however i will travel with my cane oh that's good what do they call that johnny jet had this had a name for it the miracle they call them the miracle flights yeah where when you're getting on the flight there's 40 or 50 people in wheelchairs to get on early and then they all walk off at the end of the flight it's like they're healed it's amazing i was on a miracle flight, I was like, that was the weirdest.

2:41:29Olivier Sylvain:I would be, again, I'd be embarrassed. I don't want to be, I just, I'll get in line with everybody and be a normal human. That concludes the old man version of Intelligent Machines. Thank you so much to Olivier Sylvain, our guest. He was fantastic. Look forward to reading his book, or you should look forward to, I read it, but you should look forward to reading his book, Reclaiming the Internet, How Big Tech Took Control and How We Can Take It Back from Columbia Global Reports. Next week, we will talk about AI hallucination. That should be interesting. Paris will be back in two weeks. With a cowboy hat.

2:42:15Olivier Sylvain:With a cowgirl hat. She better have a cowgirl hat and snakeskin boots. We thank you all for joining us. We do Intelligent Machines every Wednesday right after Windows Weekly. That's 2 p.m. Pacific, 5 p.m. Eastern, 2100 UTC. You can watch us do the show live in the Club Twit Discord. I hope you're a Club Twit member. If you're not, please join the club. Support our efforts. Unlike Salzburger, I do not have a trust fund to run this on. We just rely on the kindness of strangers. You, for almost 30 % of our operating expenses. twit.tv slash club twit let's go for 50 folks come on come on yeah come on actually uh you do get one benefit a new benefit you add free versions of the show but we now have chapter markers in all our shows which allows you to skip over the old men talking about their ailments segment so that's a that's a real benefit for benito has to mark that out club to it here's the ultra chapter mark no no no ai does that now hey yeah it does it yeah so we it's not hard to do we do do it, but the problem is because of ad insertion after the fact, we don't know the lengths of the shows with ads, and so we can't really do it reliably.

2:43:27Olivier Sylvain:So join the club. That way you get that benefit. You can watch Club or Not on YouTube, Twitch, X.com, Facebook, LinkedIn, or Kik. We stream on all those platforms as we're doing the show after the fact. You can get audio or video from our website, twit.tv slash IM. There is an Intelligent Machines YouTube channel. You can go there and get the video of the show. Great way to share clips or subscribe on your favorite podcast client. And you'll get it automatically as soon as it's done. Thanks to Benito Gonzalez, the world famous Benito Gonzalez, our producer and editor. Thanks to all of you for joining us.

2:44:01A boxer and politician. Who knew Benito? Yes.

2:44:04Olivier Sylvain:So many things. Jazz musician as well. We will see you next week on Intelligent Machines. Bye-bye. Hey, everybody. It's Leo Laporte. You know about MacBreak Weekly, right? You don't? Oh, if you're a Macintosh fan or you just want to keep up with what's going on with Apple, this is the show for you. Every Tuesday, Andy Inaco, Alex, Lindsay, Jason, Snell, and I get together and talk about the week's Apple news. It's an easy subscription. Just go to your favorite podcast client and search for MacBreak Weekly or visit our website, twit.tv slash mbw. You don't want to miss a week of MacBreak Weekly.

2:44:41I'm not a human being, not into this animal scene. I'm an intelligent machine.