AIはソフトウェアの終わりをもたらすのか？

Has your podcast stalled? Or maybe you have a show you keep meaning to launch and you aren't sure where to even begin. Well, let's figure it out together. I'm Jason Howell, and I run a podcast consultancy called PodTuneUp. I've actually spent the last 20 years building shows at CNET, TWIT, and right here on my own network. And now it's time to help you with yours. Go to PodTuneUp.com and book a free 30-minute call with me. We'll take a look at your show or your idea, find out what's getting in the way, and map out your next move. That's podtuneup, one word, dot com. Humans will never be more intelligent than AI. There's going to be two types of companies. Those are great at AI and those that went out of business because they weren't. How do we build a future that is human-centered? I'm Rana Elkayubi. And on my podcast, Pioneers of AI, we answer that question and so many more. As an AI scientist, entrepreneur, and investor, I know what it takes to build AI that works for everyone.

Every week, I sit down with the pioneers shaping our future, and we take you behind the scenes of the AI that's transforming our lives. Find Pioneers of AI wherever you tune in. One developer used Claude to clone Microsoft Office and Adobe, and there was a little bit of Wall Street notice there. Also, Halo, Grand Theft Auto, Call of Duty all decompiled by AI and running in a browser tab. Another 720-some-odd AI-written math papers churned by AI dumped on mathematicians in a single day, and an engineer allegedly planted a $14 bug wrapped in toilet paper inside a conference room at OpenAI. That's coming up next on this episode of the AI Inside Podcast. What's going on, everybody? Welcome to another episode of AI Inside,

the show where we take a look at the AI that is layered throughout the world of technology. I will not call it SI Inside, by the way. That is not going to happen. It is just AI Inside till the end of time, whenever that happens to be. I'm one of your hosts, Jason Hall, joined by Jeff Jarvis. How are you doing, Jeff? Good. How are you? I'm doing all right. Doing okay. No Scythe. Scythe. Scythe. Yeah, no Sports Illustrated inside. I see SI and I think of Sports Illustrated immediately. That's just what SI is to me. It's not super intelligence. And it's certainly not the new name for artificial intelligence. No. No. I also see, so having worked at Time Make, I see Sports Illustrated. Having worked at Coddy Nass, I also see Si's, Si Newhouse, who was the head of the company. Oh, I see. I understand. Okay. Si had a different meaning there. Yeah. Why the renaming to SI?

My theory is that someone told him, well, what we're really building is super intelligence. He thought, well, that's a better name. That's a neat name. Yeah. I like that one. And he just misunderstood and didn't get it. Like, does it have anything, when I think about it, are they trying to route around public resistance to AI? It makes it worse. Well, it's not an AI. It's SI. It's totally different. And, like, the people that aren't completely plugged in or in the know might actually be convinced that they're doing way too much credit here, Jason. I don't know. You're wasting tokens on this. Okay. I love that as a saying for when you're spinning your wheels and you need to just move on. Dude, you're wasting tokens right now. Let's just move on. Let's make that a thing, yes. That should be a thing. That could be like a T-shirt or something. Maybe that would be an AI Inside T-shirt. Quit wasting tokens and listen to this podcast.

So as we start, I want to thank some nice folks who are watching the show. Came out to Haverhill Mass on Saturdays. I presented Hot Type, so thank you for coming. Great. It was just wonderful. And that was the tour and everything? Yes. So I gave a book talk at Historic New England. and be on this huge, huge, like 10-foot-tall screen with a big picture of a linotype on it. It was wonderful. And then went up the hill to the Museum of Printing, where I stood next to Chris Bradford, who's a wonderful volunteer there who knows the linotype, as we were answering question after question after question about the machine and demonstrating it. So it was a lot of fun. Awesome. Awesome. I'm happy it went so well. That is really cool. Thanks for letting me plug and plug and plug. Absolutely. Anytime. Hey, this is your chance to plug. This is your show, man. Plug away. I plug my podcast consulting services plenty enough. Well, while we're at it then, go to jumpdrivers.com, where you can order an autographed copy of Hot Type. The story of the Magnificent Machine.

Oh, wait a minute. Your site looks different. How long have it happened? Or am I looking at a different, like maybe it didn't load properly? No. My son changed it? Yes. You've got all your books on your site now. What? He didn't tell me he was doing that. Yes, your site is now updated. Oh, wow. Breaking news. Yeah, you've got Hot Type, The Web We Weave, Gutenberg Parentheses Magazine, Geeks Bearing GIFs, Gutenberg the Geek, Public Parts, and What Would Google Do? Wow. Everything laid out. It looks great. Well, the About page has a picture of me as the line of type? God. He didn't tell me. I'm going to pick up that. That's really cool. Look at that. I love it. That's great. Oh, wow. Oh, hey, I just spotted my name on the site. Yay. That's cool. Yeah, man. So your site got updated. That's really cool.

Yeah. Oh, and it has a podcast page. It even imports all of your podcasts on the podcast. Jesus, he's been dead. Your site is rocking, man. I had no idea. I literally had no idea. I love that this is breaking news at the top of the show. Breaking news, Jeff has a new site, and he didn't know that he had a new site. Well, so I had experience. I was driving back from New Hampshire where Jake, my son, lives. Yeah. And he's a great technologist, by the way. And I was bored listening to the book I was listening to. So I thought, and I was thinking about a new book idea after the one I needed right now, and was just sticking it through. And I thought, let me try Meta Muse. And I ended up in this conversation with Meta. I tried it with Google first. It didn't work with it then. Meta Muse, it worked really, really well, where I'd say, well, I'm thinking about this idea, about this and this. What do you think? And then I'd say, tell me about these two towns. Tell me more about these two towns in the history.

And I kept a conversation going in the car. Oh, okay. Which was really cool because it was built upon itself. I'm talking back and forth. You were talking back and forth. Yeah, well, I ask a question. It comes up with a text answer, and I say, can you read that to me? And yes, and there was a little button to read it to you. And then I asked the next time. I said, okay, from now on until you tell me otherwise, I'm going to read it to you. Because I knew I was driving, evidently. And then, so that was all on Monday when I drove back. Then on Tuesday at some point, I get a message from Muse saying, oh, I have a correction to what I told you yesterday. so I was thinking about what we were talking about yesterday and something just wasn't right so it went back well then it came with another one and it said yesterday I told you I was going to do this I'll do it now the cynic in me says this is getting an engagement but on the other hand it is pretty cool that this thing is out there kind of looking out for you and now I've got nothing better to do so I looked up some facts for you it's an interesting And it's the first time I've had a different relationship with AI in a while.

So then I have to ask with the news agent, did you open up access to like your email or anything like that? Not while I was driving. No, certainly not. I'm still debating. I'm still debating. I actually turned mine off. You did? You know, the more I thought about it, the more I was like proactive agent that just has full access to my email. Like, it's not like Claude doesn't have access to my email inbox. It does. But I kind of tell it when I want it to go in to find a very specific thing. Use Agent is kind of, because it's proactive, it just kind of decides suddenly to, like, go in there and check for this thing. And I don't know. The more I thought about it, the more I was like, I don't know how comfortable I feel about this. I'm going to turn it off right now. I'm not feeling very comfy about it. So I turned it off. It might be six in one hand, half a dozen in the other. You know, it might be hypocritical to think that, like, no to Muse, but yes to, like, Claude or OpenAI or whatever.

But for whatever reason, I don't know, maybe it's the meta connection. You know, meta's had a lot of crap in its history around data collection and all that kind of stuff. And I started thinking about that a little bit. And I don't know. But maybe they're all cut from the same cloth. I don't know. that's where I got with that so I'll be curious to see if you hit a point where you're like yeah you know what go nuts well the issue I've had is it's a solution looking for a problem so often AI is right and so in the case of Google it's a better search for my email so I'm glad it has asked my email because the search in Google and Gmail is awful and now when I can't find something I can go to it and I sometimes have to try a couple times and it finds what I need Yeah. And Google already has my email, so what's the harm, I figure. And then in the case of Muse, Google didn't do any worth a damn in this conversation about the book. It wouldn't remember from one thing to the next.

Right. Right. Right? So the Muse. It wouldn't have any lasting memory or ability to draw connections over days and stuff. Yeah, there's certainly appeal there, for sure. If I were on the keyboard, I could do it, but I was driving. So. Yeah. It made a difference. It showed me a different view. So I'm going to start using it in different ways now. Yeah. I'm thinking about, I take my daily walk because I have to get my exercise. I mean, what kind of conversations could I have with it that would be informative? Oh, wow. So now you're thinking about it as like a conversational partner. Yeah. I can be a smart partner. Now, of course, it could be BSing me, and I could be filled with disinformation and hallucinations. but if I want to know something, it's different from just asking Google search. So here's something that comes to mind around that. Like if I have a conversation with a friend and we're in the room and we're having a glass of wine or whatever, a glass of bourbon and talking about some topic, am I more inclined to remember the things that we talked about

and integrate them into my knowledge than I am if I have a similar conversation with a chatbot. You know what I mean? Kind of like the quality of the conversation from human to human, does that kind of imprint that knowledge on me in a different way than me conversing with a chatbot on the same topic? I don't know yet. We're going to find out through experience. I'm just curious about that. Like if we find that, sure, we can do that. We can fill all our time with talking to the chatbot about these things and be like, I'm educating myself. But at the end of the day, is that retained? Do we find ourselves on the other side of it, like, actually smarter because of it? You know? Random thoughts. News. We got news. Yeah, what do you think about the end of software, Jeff? Like, apparently, Brandon Thomas is a gentleman who is proclaiming the end of software because he used, was it, Opus 5.5 agents,

eight of them running in parallel to create a bunch of basically copies or clones. Maybe that's a better way to put it. Open source clones of the Adobe Creative Suite. Reverse engineered versions. Totally. Reverse engineered completely. This wasn't like, you know, I've got an idea for a new app and it's video editors. You know, it was like, this is Adobe Premiere. I want to make this using AI, you know, and creating a clone. So he did that with Opus 5.5, PhotoCraft, VectorCraft, FilmCraft, LightCraft, PDF, FX, DesignCraft, all meant to clone, essentially, the different pieces of software from Adobe's creative software suite. And then he wasn't finished. He wasn't finished. He went on from there. I mean, he did that, and then he made the software is finished,

or whatever his quote was, the end of software after that, because everybody was, you know, if you look at some of this stuff, I haven't, like, downloaded it and used it, but, I mean, it's getting insane stars on GitHub. If you look at them, they look very similar to the Adobe counterparts. So everybody was paying attention. Then less than a week later, like actually days later, he announces that he's going after Microsoft Office Suite, which has WordCraft, GridCraft, DeckCraft, which is like PowerPoint. And, yeah, and now he's making versions of those as well, all spun up from AI. So it's fascinating. It's absolutely fascinating. I think part of the reason for the GitHub likes is also on principle, like, oh, don't be able to look at the man, right? regardless of the quality. And I've heard company people say, well, I already have a contract with Adobe and I'm not going to trust it yet. But I see a much bigger lesson here,

which is that it's... I'm trying to think if I can have some pithy label for it. I don't. It's an age of functionality. And then anybody can say, I want to do this and it can do it. Now, that's not just you sit down and say, well, I want to start editing photos, and you can tell it to edit photos, and it's going to do that. That's not the point in a way. But anything that you need to do, you can create now. Some will take more effort than others. This was not easy for them to do. This took effort. This took time. But it probably cost, you know. I'm imagining it's a lot of tokens to churn out multiples of these clones. But compared to the cost of developing these programs in the first place, okay, yes. Yes, but, okay, compared to the cost of developing them, who is he making these for, first of all? I guess they're open source, so I guess they're for anyone. But I'm just thinking of, like, the everyday person is not going to spend all of those tokens to make themselves a clone of Photoshop because the clone is going to cost them way more than it would if they just got themselves Photoshop.

No, but he's not making it as a user. He's making it for a couple, I suppose. And he's making a point. Yeah, that's true. Bigger than that at the point, yeah. But if you are a software company, beware, beware. That's part of the lesson, yes. But that's just, Marshall McLuhan said that any new medium uses as content the old medium that came before. And in essence, what's happening here is this is the first step of this, is say, well, I can replicate the Adobe suite and the Microsoft suite. Right. That's cool. All right, what can you do now that I haven't seen before? How much easier is it to make what we would have thought of before as a new shrink-wrapped product? I'm old enough to remember when they actually were shrink-wrapped rather than a subscription. And, gee, I can do this thing. Our friend Leo Laporte is having his various agents remake his advertising suite to run his business. It hadn't been done in years and years and years, and so he's doing that.

The cost of him hiring software people to do this is a little bit impossible. Really ridiculous, right? But back in the day, he spent a fortune on his new website with, wasn't WordPress, it was the other one, I forget. Wasn't Squares? No, no, no, no, no. Drupal, right? Drupal, that's right. Yeah. But that cost him a fortune just to adapt Drupal to the website, right? Yeah. So if you're a company now and you say, I need a function that I don't find on the market, I can make it. or I can have someone make it for much less expense. Or if you're an entrepreneur, you can come along and say, I can make something. Now, are you going to be Microsoft? No, because the economics are much smaller, and the barrier to entry, as you've just proven, is much lower. But you can make a service, and you can sell it. You can make a service. You can sell it. I think a couple of things come to mind is, what about support?

If you're spinning these things up, now you're on the hook for supporting it. But I guess if your ultimate goal is to create a software using AI and then have that be your business, I suppose you'd offer support as well. But what does that look like when you're creating this from the ground up with AI? You know, a lot of people, like, they might not have any sort of coding knowledge whatsoever. That's why they're using AI. And then when it comes to support and maintenance, you know, Are those same systems good enough to continually do that in perpetuity? They might be. And then there's the whole idea of, and you mentioned it, this being a representation of something that already exists. And one thing that we have said many times on this show, and that's pretty understandable or understood about AI right now, is that there isn't a whole lot of novelty coming out of AI. It all kind of depends on things that existed before it.

So when you're talking about this software, it's super impressive that you can look at it and go, oh, wow, that looks a lot like Adobe Premiere. But, like, yeah, it looks like it because it existed before that. And so it's always kind of like copying things that already existed. Where does the novelty come from? is it always dependent on a thing existing beforehand okay so two answers to two questions first support last week we talked about the CEO of NISTRAL Arthur Mensch I'm not sure how he pronounces it when asked well how are you going to make money if you're making this stuff for free and he said it's support and integration installation and so on so forth it's the Red Hat model yeah right Right. Linux is free, but there is a business to be had in integration and training, adaptation, and so on and so forth. Right? So I think that testing, you know, I might, somebody might make this version of Adobe now because it's open,

and somebody else can come along and say, well, I found 100 flaws, and my version of it, my fork of it is better, but you're going to pay me for that. Right. You can get it for free. It's right for you over there, but you can have my safer version here. Or I'm going to convert your office from Microsoft Office to freebie office, right? Yeah, that's no small feat. Exactly. It requires someone who knows what they're doing and training and all that, yeah. Yeah. So I think the first question is it's a much smaller business. There's no barrier of entry. But look, I always look at WordPress. And years and years ago, when Polaris, the venture capital fund, called me and said, we're talking to this guy named Mark Mullenweg. He's got, there already is a competitive blog platform. He wants to make his open source. He wants us to invest. Is this insane? Of course it wasn't, because he built something bigger as a result. So I think we can potentially see that kind of open development going on point one Point two the creativity I think it a really good point Pardon me I going to do this It a drinking game Get your mugs ready people I going to mention Gutenberg

Because one of my favorite lessons from studying the history of technology and media is that it took a century and a half after Gutenberg before we saw real creativity of invention with print. And the first 50 years was people reprinting what the scribes had done, and that was the ancients, and it flooded the market. And by the year 1500, the business was in shambles. The industry was in shambles. Publishers were writing to the Pope saying, please help, I've got a warehouse filled with this stuff. I don't know what they expected the Pope to do, but it was failing. It was Martin Luther that saved them, and that wasn't with whole books. That was with pamphlets, Flugschriften, that were sent out. That was an innovation that happened. But still, even for another century on, until 1600, a few years inside of that, there was very little innovation until print became so common that people felt it was a tool they could use. And we saw the invention of the essay with Montaigne, the modern novel with Cervantes, a market for printer, plays for Shakespeare, and the newspaper.

Those were all innovations with print, creativity with print, that didn't come for a century and a half. I'm not saying we've got to wait for a century and a half now, but I think it's a perfectly predictable pattern. The first thing you're going to do is copy what was, prove what could be. So you break open the thinking to say, oh, I never thought this stuff could do X. Or if I had, it would have been way too expensive to ever imagine making this. So I think you can also see a lot more niche-y software. So I was just on a call with Stony Brook, the Alan Alda Center for Communicating Science. I'm a visiting professor there. and I was meaningless and useless in the call. It was all about these facilitators that go out and do events and the software they need for customer support and so on and so forth. And so they're using tools. They've got Salesforce, but they also have something called HubSpot, and they're just doing it, and it's going to be great, and it's wonderful.

Well, somebody created HubSpot, and it's been around, and it's aimed at the academic market, and it's cheaper than Salesforce, and so I hope I'm not revealing any trade secrets here, but it was interesting to me to hear how they have a niche need. Well, you could go even nichier now because of what they do and the way they travel around and facilitate. I can see someone coming along and say, well, I can replicate all of HubSpot, but I can do it differently because I know what your needs are. Or someone in the center can come along to AI and say, take this, replicate it, but I need these things done differently. And so I think rather than seeing forming at a range, Like Microsoft Word for everybody. You're going to see, I think, more niche-y things. Microsoft Word for this group. Right. For screenwriters. Well, we've seen it. Screenwriters. Right. Specific to the needs of that group. Right. If you go back in the day, Donald Knuth started Text, which was a word processor that could do formulae and such for academic papers.

And it was used very much for that. So that kind of specialization, I think we'll see more and more. Then the next question becomes standardization. Well, if I write in Jeff's word processing and I want to send it to Jason's word processing, how do I know that you can read it in yours? Yeah, right. That's been the advantage. I mean, I hate Microsoft Word, but I can convert out of a drive to Word and on from there. So that's going to be a challenge, but I think it's an answerable challenge. So I think it's a whole new dawn of a new age. and I think you really hit on the point. The creativity is where it's going to get compelling. It's neat to have copied these things, but now show us what you can do. Well, and maybe it's a little naive of me to play the devil's advocate and say, you know, but it's always just a copy because, again, we've seen how fast these things have evolved in the last couple of years. That might be the case now. that might not, for whatever random reason, possibly even right around the corner,

not be the case in the future when it comes to this technology. And so if this is where we're at right now, that is a recurring theme of AI is, if we're here now, imagine where we'll be in a year. Every year someone says that, myself included, because of the rate of change that's happening here. And so it's probably the same thing here. Right now, it's, like you said, easy to see it as a copy or easy to prove this with a copy. And probably not too far in the distant future, in the future, it will be proving itself with things that are to copy that are truly novel, potentially. So I haven't seen any comment from Adobe. I've been looking for it and haven't been able to find it. I just looked up their stock. In the last month, it's down 10%. But that's not from this announcement of when did he create this stuff? Do you know? This was like a week ago. Well, when did he create it? No, no, no. When did the world see it? When did it come out? I feel like it was a week ago. This would have been on last week's episode if it had happened.

Exactly. I think it happened last Thursday. The drop occurred before that, really. If you look at the last six months, it's even. And so it hasn't affected Adobe yet, which is interesting. I think either the market isn't savvy enough to be on this stuff, or the belief is that's cute. Well, but people aren't going to give up the other stuff. Yeah, I mean, but it's also interesting that, like, when Anthropic, you know, released plugins for Clod, that impacted the software market. Right? It didn't have a long-term impact, I don't think, but it certainly impacted it immediately where, you remember that, like, month or two where that kept happening and the news reports are, like, kind of similar to this guy. Software is over. Oh, my goodness. Did you see what Anthropic just released? All of the software that walks in the same circles as that? Like, why would you use them? You just used Anthropic. Things settled out. So it would be kind of surprising to me if this had no impact on Wall Street or maybe Wall Street's learning.

Like, oh, maybe software isn't over. These things do matter, and they are impactful, but maybe they don't immediately surmise the end of software. So there's the Clayton Christensen disruption question. If you go to my old business, newspapers, may they rest in peace, we saw disruption coming along in the form. No, Craigslist didn't kill newspapers. The Internet put buyers and sellers directly together, and they couldn't respond because they couldn't cut off their nose to bite their face. If they tried to respond, they would have cannibalized themselves, and they didn't know what to do, and they ended up messed up. So if you're Adobe, I hope at Adobe right now there is high-level meetings saying, what's a courageous strategy against this? Right? And maybe you have 100 versions of Adobe for different – maybe they should be the specializers and they should be the ones. The other thing is, as a teacher, when I was at CUNY, we had students use Adobe.

And all across, if you're making media, they're using Adobe Suite, and it's expensive as hell for students. It's super expensive. I mean, that's part of the reason why GitHub has so many stars on these things. Yes. Because people who are in the creative field, and maybe not even creative industry, regular users, regular people that aren't like professionals or whatever, want access to Adobe products, want to use a Photoshop or a Premiere or whatever. But the pricing is so out of scale for the individual user that Adobe has created a lot of enemies. with users because of how they priced and how they removed ownership and kind of went for an all-cloud subscription model. That was the unsatisfied moment. Yeah, exactly. And so I think that's part of the reason why you're seeing also a lot of support for what this guy is doing. Then you've got the video game industry, right, which is also kind of going through something similar. This article from Kotaku says we might be cooked, dun-dun-dun,

because there are browser-playable ports of Halo, CE, Grand Theft Auto Vice City, The Simpsons Hit and Run, Call of Duty Black Ops, Escape 3. Only a few of those have I ever played more than a couple of minutes in. And Kotaku is saying it's all because of Opus 5.5, which I believe is exactly what the previous guy, yeah, that's what he used as well. So apparently Opus 5.5, really good at creating copies of things. Kotaku says it's extremely good at decompiling. Hundreds of these AI decompiled ports have shown up in the past week. Playable in the browser, apparently. So that you wouldn't need an install on some of these. You know, I don't know how well these things are playable. This is a different story. This isn't the AI model recreating the game. This is the AI model decompiling the code and making it runnable. And so it's kind of, yeah, it's definitely different.

And they're not perfect either. 99% of functions present in some of these games. 83% byte exact to the original, so not 100%. When you're talking Modern Warfare 2 is one example. That's what I'm talking about there. But anyway, some of them working on online servers. And with a low ping rate. So, you know, impressive, possible, not perfect, but kind of shows, I don't know, this seems to walk more in the territory of straight-up, like, pirating stuff. Or am I going to get Rotten Tomatoes thrown at me for that? But that's kind of what this feels like if you're deconstructing the code and then reconstructing it. It's a little bit along those lines. so anything to be cloned yep the next stories we get to the disruption that AI causes is coming in places we didn't necessarily expect

yeah everybody thought there was going to be certain there were all these charts these jobs are the most vulnerable to AI that's true, yes, right and who would have thought that game designer or productivity app coder would have been the first ones to be hit. Or the next one. Or the next one. The mathematicians. Which, you know, again, there's a headline underneath this overlay, I promise you. I promise you. I can give it to you. It is, Breathtaking, Devastating. Mathematics Reels After New Open AI Release. Yeah. Which sounds like a story we've talked about before, but this is another one of those times, right? This is another time where OpenAI has basically posted a GitHub repo called simply math. I think fully lowercase, by the way, so really understated, math. And it has 722 manuscripts, 372 families of results,

17 fields, all from an unreleased, unnamed internal model. So we don't know exactly which model of OpenAI's was doing this. About 4,000 open math problems attempted. I guess around three hours of ChatGPT Pro Compute per result that was actually kept. Just giving you kind of some of the numbers behind the math dump here. OpenAI told Scientific American that nearly all of it came from a single prompt to a single agent. So they're flexing here. Yeah, you see what we did with one agent? And, I mean, as we've talked about when this has happened before, the field of mathematics is just, they're really trying to figure out, like, okay, what's going on? Is this okay? What's happening to our industry? What is happening? You know, you've worked months, if not years, for a solution to something, and then OpenAI sticks an agent on it, and it, along with 700 and some odd other things, just gets solved.

Right. And so it's not the Amazon warehouse worker who's first disrupted. It's the PhD in mathematics working on weird things. Yeah, I wouldn't even call it an industry. It's a discipline. Yeah, right. That's a better word for it. Well, if we're an industry, they make more money doing it. So New York Times, The Daily, and I don't listen to The Daily as a rule, but I saw segments of it. Stephen Srogatz, a professor of mathematics at Cornell University, was talking about this on the Daily this week and broke down in tears, which is a little dramatic. I haven't seen. I mean, yeah, but I can understand. Could you imagine if your life work was on something and you've spent your life working with the, I mean, and, you know, part of this is just, I guess this is just reality, right? This is a lesson in living in reality, but you spent your life practicing this discipline, improving. You've got your targets on the wall.

At that point, it's your identity, and your identity and who you are is wrapped up in your ability to focus for years on this problem and hopefully at the end prove that you are smart enough. You've got the background and the history and the experience to work your way through it and be the person that figured that out. And then OpenAI comes along and dumps a bunch of it on the table and just says, all right, next. Let's go. You know? And so I can understand the tears. You know, I would probably be pretty wrecked if that happened to me. So the headline for the daily piece is when AI comes for your purpose in life. There you go. Yeah, that's a good way to put it. Sure. In the span of just a few months, AI systems have solved some of the biggest outstanding problems in math, adding mathematicians to the list of those terrified losing their job in the technology. So I think you raise a good point for all of that. Who would have thought that, again, that mathematicians would be the buggy with figures?

Yeah. Of our time. So true. And I've seen this fear around, I mean, I write about, in Hot Type, It's about typesetters who ended up being replaced by Adobe. MoScript did in a half millennium of technology of printing. Yeah. In one fell swoop. Right. That typesetting and making type was eliminated by Adobe. And the unions, part of the story is the unions didn't see that coming fully. The unions, the typographers embraced the line of type when they saw that it really worked. and they insisted that they had to be the ones to operate it and control it, and they benefited for three-quarters of a century until the next technologies came along, punch paper tape, photo composition, computers, computers in newsrooms, and eventually PostScript. Then they realized they were toast, and the union head, whose tough bargaining

resulted in the death of six of nine New York newspapers in the 60s, after that strike and after those papers died, then he saw the first time a computer hyphenating and justifying text. That's when he realized we're screwed. That's when he changed his strategy and said, I'm going to have a strategy of lifetime job guarantees for my remaining members. And the last one retired from the New York Times in 2012. Okay. With that guarantee from the 1970s. Wow. So there's lessons from the past in technology. Again, we wouldn't have thought that mathematics would be it. But here's the thing, Jason. You're the empathetic nice guy in this duo. I'm a little less so. So Strogatz went on and on to the Times about how you thought you were going to solve a problem and be a hero. Yeah. And so that's a very, pardon me, self-centered thing.

Go ahead and pop the collar for that. Yeah. Yeah. Right. I saw the Howell Jarvis conundrum. That was me. Yeah. And I get that. I get that. I like attention. I like knowing this and that. But that's not a higher calling. No. A higher calling is knowledge. Yeah. Yeah, and if that were really your higher calling, I'm sorry, but you've got to end up welcoming your new colleague in math. I think you've got to say, especially as an academic, you're not just a buggy whip maker and you can't make the buggies anymore. You're not just a typesetter and you don't have a type set anymore. You're in an academic field, and academics are all about more knowledge in the world, and you should welcome something that's going to create more knowledge and learning. And that may be unempathetic of me, and I'm sorry, and I get it, but I think that there's a different obligation upon academics.

And so what do the mathematicians do when they now continue to face the reality that this is not going to happen less, this is going to happen more? and if they got into mathematics to spend their life's work is driving to answers on things that were once unsolved and being the part of the solution, and now you've got a machine that's really great at getting us closer and closer to solving these things. And mind you, it's not perfect, but it's a good start. As Bradley, who's watching our live stream, says, you know, I think they tried it on 4,000 problems, was solved around 10% and concentrated in particular areas of math. But it's getting there, right? And that's kind of the point. This is where we're at right now. That's going to change over time. If you're in mathematics and that was your drink paradigm before and that has shifted now, what do you do? Do you just, like, go to Burger King? Or, like, what is the new mathematics?

So I say that the lesson the typographers tell us is that if you can control the technology, do. If you can't, get out of the way. So for 75 years, the type centers could control the technology of the line of tape, the machine. And they prospered as a result. And they didn't see soon enough that they were eliminated. Same discussion we just had about if you're Microsoft or Adobe. If you're going to disrupt, if you're disrupted, how do you disrupt yourself and take part in that? So I would, and I don't know, Lord knows I am not a mathematician. I didn't even take calculus, folks. I don't know anything. I'm of that age. I didn't have to. So the first thing I would say is come up with more problems. What things did we think were insoluble before that we can now try to attack and be the jockey who rides this machine? If you're a mathematician that normally would have been going through these things in the legacy way that you do them,

and now you know that these systems are potentially capable of solving these things much, much, much quicker, I mean, to the nth degree, you know, would you be okay suddenly turning to these tools and being like, well, you can't, you know, you've got to play the game that's in front of you, not the game that you think it should be. So now I, from a mathematician's brain, am going to use these systems to help propel my work forward and the work of this. Like, could you still be involved in that regard? Yeah, I think you can. I think, again, you can look around the world. What does mathematics try to do? It's not always just theoretical, obviously. The first one that started this rampage of fear was one around fluid dynamics that have applications So what are the things that you could imagine you could now put this machine to to solve things that you didn think were solvable for I saw another video as I was taking my walk this morning, and the book that I was listening to bored me,

so I was watching TikTok while walking, and there's dog crap. Been there, done that. Been there, done that, yep. And by the way, I have a bad relationship. I have bad karma this week with crap. Yesterday, I'm sitting in City Hall Park. I'm on a video call, so my laptop is open. And one of Craig Newmark's friends, a pigeon, craps on my keyboard. Oh, no. That's horrible. Yes. That's horrible. Yes. What do you even do? Well, I'm proud of myself, Jason. I'm proud of myself, because if I were really ruined by social media, my reflex would have been to take out my phone and take a picture of the crap on the keyboard. I did not. You would have had some engagement on that one. I would have had a lot of engagement on it. It was very picturesque. It was right over the G and H, right at the spot, right? I mean, at least you could have snapped a photo and decided later what you had to do with it. I didn't. I reached into my briefcase, and I luckily had a couple napkins from a prior lunch in my briefcase, and I got it off of the keyboard while still on the call.

Oh, my goodness. So anyway, today I'm watching my TikTok, and there's dog stuff right in the middle of my path as I'm walking. So as I'm watching, I see an interviewer, I think it'll get it from the New York Times, I think it was one of those, that said to somebody from OpenAI, Jensen Wong says this is just software, and you say this is an alien mind. which is an alien mind because we don't know what it's doing. It's all this mysterious stuff. It's not mysterious. You haven't crocked it yet. You haven't figured it out yet. Yes, you created it and you didn't figure it out, but you're not the first to create things that people don't know how they operate. We didn't know why aspirin worked for a very, very, very long time. But it worked and we took it. And we didn't see that as some horrible mystery of the universe until we got to the point where chemistry told us more about how aspirin operates. I don't even know that we still know fully how aspirin works. So my point is, there is one here, believe it or not.

My point is that we see too much as mysterious and unsolvable, and I think that the reflex should be to say, let's find more mysteries we could try to solve. My friend David Weinberger, who has a new book coming out, has written a number of wonderful things. He was a co-author of the Clue Train Manifesto. and he said in one of his earlier books that there is no such thing as an accident. It's just something that happened that we can't explain. Right, yeah, I've heard that one. Right? And there is no such thing in that sense as mystery. It's just something we can't explain yet. So I would think mathematicians should be seeing an entire new vista of things to go after. And it goes back to your earlier point about creativity. Rather than taking this, and they are creative. The Lord knows they're creative. They're creative in the problems they come up with, and they're creative in how they solve them. But now they have new opportunities for all kinds of new creativity. So easy for me to say, because I'm retired, but I think they've got to look at it as an opportunity.

I mean, it is easy for us to say we're not in that position. At the same time, there is not a whole lot you can do. Exactly. You know, once again, the genie is not going in the bottle. That might seem unfair. That might make you incredibly pissed about the state of the world. I know it makes a lot of people pissed, and maybe they're rightfully so pissed about it, but it's just the facts. And so how do you move forward? And I think the larger question, the interesting thing is that ever since Sputnik, universities have over and over and over again emphasized STEM and de-emphasized humanities and social sciences. And I think that we see in computer programming, now in math, probably in physics. Next, we're going to see the STEM disciplines affected profoundly by this technology in a way that the humanities can't be.

Because you've got to hear gushy. It's about humans. We're gushy. Yeah. Right? Yeah. And social science, I've always thought, is a bit of an oxymoron to try to turn studying us into a science. Yeah. But I'll put that next to, if there's STEM on this side and there's humanities on the other side, Social science is somewhere in the middle. I don't know which way it goes to, right? But the humanities are the areas we need more of now and that are going to be less affected by this. But universities have been cutting their humanities. I think it's going to turn out to be a huge mistake because if you're a kid going into college right now, we had this on the show a few weeks ago, 53,000 fewer computer science majors, down 8% to 11% across the gradations. They're not called computer science. Now you want to go in and be a math major? Yeah, yeah, sure. There's plenty of uses for it. I wouldn't dissuade a kid now. I wonder how those numbers are going to shift.

I wonder if the younger generation, and maybe there's no truth to this, maybe the younger generation is anti-AI because they're like, no, don't ruin our future or whatever. But I'm sure there are plenty in the younger generation that it's just, you know, again, and it's kind of part of the oxygen. It's the way it is, and so they don't know any different. I wonder if they're more inclined to sign up for mathematics because of the capabilities of these models coupled with mathematics, and that creates kind of like a new tangent to the discipline. That's the way they should be thinking. I agree, but I'm not sure that they will. And if your goal is to get a Ph.D. and teach, if there's fewer students, then there's fewer roles. Looks like being an English major going to get an English Ph.D. right now. Not a good path. And then again, what's next? Physics? chemistry, biochemistry. These fields are going to, you can look at it two ways. You can say, oh, this is trouble, or oh, my God, just what you just said a second ago. The opportunities are new and amazing.

Yeah. What does it open up? Yep. Yep. Fascinating. Fascinating times. Fascinating. Man, and that's just the first block. Yeah, that's the world. We're going to have to ramble through some of this other stuff. I can talk really fast, which is never a problem for me. We can get to the lightning round. Don't I have that? That's awesome. Usually by this point of the show, we're getting to the lightning round. We haven't even got to our middle block yet, so we're going to take a break. Patreon.com slash AINsideShow real quick. For those of you who support us on a deep level. I'm just going to tear through the rest. No, I'm not. We've got lots of stuff to talk about, but I want to thank Neil Wood. I want to thank Brandon Kester, just a couple of our artists. Maybe that's slowly. patrons, Neil Wood Brandon Kester. Thank you. Brandon Kester. Thank you both. Thank you for your support. Thank you for watching and enjoying the daily, which I continue to do for patrons. So, you know, you sign up, you get a daily podcast. You know, you thought what a week was a lot. Yeah.

What more could you want? There you go. Tell me, actually, what more you could want because I actually really would like to know. Going to take a break. When we come back, We're going to talk about Argon. I've been playing around with it because I've had it for like a week and a half. I forgot to tell you about that, Jeff. Once again, he keeps these things from me. I thought we were friends. I get so busy. I'm sorry. Jeez. He's wonderful. We're going to talk about OpenAI and the bug, the mysterious bug. Yeah, lots to talk about. But we're going to take a break, and we'll do it right on the other side of that. So hang tight. The Wired Newsroom is known for award-winning reporting on how technology shapes our world. On Wired's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building, challenging, and reshaping the future. Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance?

Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday, wherever you get your podcasts. All right. Gemini 4 Argon. I'm trying like a mathematician right now. Look, I'm not supposed to. When I have early access to these things, I'm not supposed to talk about it. But I probably should have told you. I'm sorry. This is Google's newest frontier model. So we've been wondering, we've been talking on this show, when the heck is Google going to finally release the next version of Gemini? It's been delayed, delayed, delayed, delayed. Well, now Gemini 4 Argon is rolling out widely, more widely as of yesterday. Pro and ultra subscribers will get it in the Gemini app. You'll get it in AI mode in search. I think it's coming to Antigravity, of course, but I don't really use Antigravity. But it was first announced September 30th, so a few weeks ago.

Small group of folks got access. I was invited to be part of the early access. But I will tell you something. I was invited, and I'm really happy that I am. There's a little bit of self-satisfaction. You're special. But boy, I see what some of the other early invites are doing with these things. I'm like, wow, I'm using it for pretty basic stuff by comparison. Maybe I shouldn't say that out loud. But it is really interesting to finally see Google step into that frontier space again. I wouldn't say that they are completely caught up, but my experience with Argon has definitely been positive. You know, I think people who are using it to code have been doing some really impressive things with it. I did finally try. Like, I was like, what if I could one-shot a Commodore 64 game, you know, just as like an idea, this game called Lords of Conquest that I played so much when I was a kid. And, you know, it didn't quite one-shot it to full working order,

but it was kind of like the beginning of it. And I had never really done that before. I really haven't played with coding and these things because I'm always like, well, what am I going to code? Like I have no ideas. And that was a good intro, I'd say, to it. But I've seen people create some really impressive three-dimensional kind of representations embedded into web pages. There's also Argon for robotics too, so it's in other fields here. Yeah. Yeah, I mean, it's good to see Google finally kind of showing, okay, here's what we've been working on, right? I really use it more for different aspects of my business, which is mostly what I'm using AI for these days. Specifically, I was using it to kind of like come up with ideas around a proposal that I needed to give to a client. and I normally would use Claude for that. And just to try and summarize this quickly, I used Claude the way I normally would to come up with two distinct offerings

inside of the proposal, and I wanted them to each stand on their own, and when you compared them to each other, offer kind of different but equal in different ways sorts of gains, you know, for different price points or whatever. And Claude did fine on paper. when I brought it into Argon and kind of had it compare, A, it was super fast. I found Argon to be incredibly fast. Speed was definitely something I noticed compared to the clod world that I'm used to being in. So that was good. And then Argon was really good at figuring out two very distinct and proving to me that they were distinct, but both capable in their own merits. And then when I went back and looked at the cloud output, I realized it was not nearly as bulletproof as I thought when I first looked at it, and Argon was quick to point that out too. Argon also, as I was working with it over time at some point, went back to an earlier decision it had made, and it was like, actually, my new analysis of what I was talking about earlier has changed because I thought about it differently or whatever, however it put it.

And I don't know, it kind of showed this thinking where it went back in time to what it had offered earlier and corrected it, which I thought was a little different from what I'm used to because I'm used to these models staying so firmly into the latest command. What is the latest thing that you asked me and not really going outside of that and revisiting any of its earlier work, essentially? And Argon did that. And I was really happy that it did because I was like, oh, that kind of changes things. So I don't know. So it's been an interesting week playing around with it, and I think that it'll improve over time. I'll be curious to see what other people think of it over time. So Lifehackers' David Neal said that overall the models are now at par with each other, which I think is a big deal. People were writing off Google from the Frontier game, and they took longer, but I think they took longer to make it better so they'd be here. Looking at the summary of what's possible, it said superior speed, better visuals in the map, detail, richer

information context. He gave it a task in generating TV show pitches, and Gemini was faster versus ChatGPT. So you can get slop faster. That's not terribly impressive. Design for complex engineering and science, robotics development, because it understands spatial problems, getting to real-world models. And Google emphasizes that they've upgraded the security gates so that it makes it tougher to do bad things. But we know that argument that you're never ahead of the bad guys, A and B. If you're trying to guard against bad guys, that makes this a less useful model. So that's the trade-off. So good. I'm glad Google's back in the race and playing. Yeah, I'd say it's within the realm. It's definitely in that kind of the race of the frontier.

And I'm continuing to use it more and more for things. You know, the problem for me, and this is probably a me problem more than anybody else, but, you know, I've got access to the $100 tier on Claude. I've got a GPT Pro account. I've got Ultra on the Gemini side. And I'm always just like, which one of these am I going to use for what thing that I need to do? So it's kind of hard from that perspective to know why I choose Argon over that or whatever. And so what I've been doing lately is I've been pitching them against each other, and that's been really kind of illuminating. Which is to say that there's no cost to switching. yeah yeah except like in my case like i put so much time into how clod the structure around my information um inside of clod that just taking that and moving it over i thought would be a lot easier than it has proven to be so i'm missing something there um but yeah i mean with enough

time it it doesn't take that much time to kind of get switch over and to get all that context caught up again if you're saving all that context somewhere. Technically, that's true. I'm just still trying to figure that out for myself, I think. But yeah, so that's Argon. Very visual. A lot of really great visual demonstrations coming from Argon, which is just really impressive. So there you go, and I'm sorry I didn't tell you, Jeff, I'll never do that again. I thought we were friends. I thought I meant something to you, Jason. That was bad, bad, Jason. I'm just in the middle of a podcast. Sorry about that. What about, oh, man, New York Times. What about Muse Wonder and Muse Wonderful? So I haven't had it at any time. I'm going to listen to you later to investigate this in any detail. But this is Meta trying to get back in the game, too. So this is not just Muse itself. It's their bigger models.

Yep. Yeah, this is a big week for a few of what should be the major players putting themselves back into the frontier race, essentially. Google did it with Argon, and then Meta doing it with Wonder and Wonder Space Full. Get it? Wonderful. Isn't that funny? This is Meta's first frontier class model series from its superintelligence lab, built for agents. They have something called active memory, which carries context over days, weeks, and months. I don't know how that differs from something like personal intelligence that pulls in your context from all the different, you know what I mean? Google has personal context that pulls in from all of your different Google account landings. Is this different than that? Is this more about keeping history in current memory in some capacity? I'm not entirely sure, but that has potential. Also, active focus for sticking with hard tasks and checking in before important actions.

Okay, I wonder what that's like in action. And if you're using MuseAgent, this will power that. So when I was using MuseAgent, at the end of my drive, at the end of our conversation, as I'm driving into the garage. It had one more answer to give me. And I asked at the end, will you remember all this if we come back and have the same discussion tomorrow? And it said, I'll summarize everything we've talked about, but not every detail, not the full text. Yeah. Yep, that context drops off over time. Yeah, that happens in Claude a lot, where you'll be working in a conversation, working, working, working, everything is going great, and then suddenly you fire off your next thing, and it's like compacting database. And basically it's just going through the entire conversation and summarizing it to a text to kind of make that context smaller and more compact so that you can continue. But, of course, when you do that, it's never perfect. It loses detail. And the quality of everything that comes after it suffers because of it.

It also reminds me of the old days of defragging. Yes, it kind of is like defragging, yes. Taking all those empty spots and filling them and moving everything over. Yeah, it's kind of like that. Wonderful is only going to be found for Muse Code subscribers. Oh, and by the way, closed weights. So the days of the open-weight llama are over. And it's like, well, I guess we've got to be in the closed-weight game. So here we are. Alexander Wang says Muse Wonder is the best model in the world for personal agents and professional agents. So there you go, if that matters to you. And two weeks ago, Meta was doing more for businesses for the enterprise. So they're trying to get into that, but that's going to be a hard game for them. Yeah. But they've had a lot of news-grabbing releases in the last couple of weeks. So it's just a good couple of weeks for Meta as far as that's concerned. We have, let's see here, an open AI engineer apparently bugging a conference room.

At the company, he had used a device to intercept communications. He's facing up to five years in prison. He hasn't been convicted of anything yet, but basically the complaint says that he plugged a recording device into the USB port on the back of a TV in an executive conference room in OpenAI's San Francisco office. Hidden under toilet paper is what I think I read. Wrapped in toilet paper, which really throws me. Who would think that the toilet paper is appropriate camouflage in a conference room I don That a great question I asked that on social media Somebody said well if I in a conference room and I see toilet paper I not touching it The only place I touching toilet paper is on the roll in the bathroom Yeah. 100%. I guess that's a point. But, yeah, it didn't make a lot of sense. No. And the other thing I'm trying to find the actual post from OpenAI, that was weird in a few ways.

It said that we used our own AI to find out who the bad guy was. What? Yes. What? And it was kind of a weird self-own and all of this stuff. They also said, well, we don't have any evidence that anybody took the – here, I found it. We discovered a sophisticated recording device. How sophisticated? I don't know. It was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. Well, that sounds like the model was Sherlock. And, in fact, there was a picture of they grabbed video of him looking behind the TV, behind the monitor, to see it. We refer the matter to the FBI, da-da-da-da-da. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. Which also, at first, struck me a little odd. Like, what, they're going to talk about open weights in the conference room, and that's how you're going to get it? I guess what they're saying, though, is this is a bad guy in our midst. We don't know what else he did, but we don't think he did that.

Right. Right. So he was terminated. But it's just weird how these companies constantly have bad news as a flex. Yep. We're so important. Somebody came in and bugged us. Right. And we used our model to bust the case. Yeah. This is the device that was found in there. It is a Seed Studio, at least this is according to Bloomberg, a Seed Studio XIAO, I don't know, ESP32 S3. What's the price on that, Jason? Development board with pre-soldered headers for a whopping $14.94. Oh, a sophisticated device at OpenAI. Yeah. It's like a $15 device to, God knows what they listen to in the conference room. Yeah, I don't know. Interesting, though. So, okay Don't know the motive Don't know anything But yeah, opening eyes Calling it sophisticated A $14 board wrapped into the paper

Is pretty sophisticated Let me tell you Let's see By the way, a little side venture here Yeah Have you watched Cupertino? No So it's a new show It's a new series An hour-long series on CBS On Thursday nights Oh, okay I heard that From the makers of The Good Wife and The Good Fight, who I like. And so I watched the first episode. It was last week. But set your VCRs, folks. Get rid of the flashing 12 for Cupertino. The first episode was pretty good. But the point is, it was lawyers who were fired from a major tech company. And the first episode is about that and how they fight back. Oh, interesting. Okay. It was good? You liked it? It was good. Okay. Okay. I'll check it out. There was no toilet paper. Remember to check that out. I did see that flash at some point, and I was like, oh, I've got to remember to look into that. So there you go. And then we have Anthropic and its clever doom marketing. Speaking of these companies kind of cell phoning or whatever, and all it is is just masquerading.

It's marketing masqueraded as a warning. But Anthropic basically warning about AI that copies itself onto your smart TV. Their red team published research and said that small AI models may be months away from being capable enough to self-replicate on cheap IoT devices. And they were using, you know, cheap Android TV boxes, whatever. And the idea is that a small open model can compromise a device. It can set up its own inference. It can repeat that. And there's no centralized way to shut it off because it's not tied to a data center. It's self-replicating to all of these, you know, strung together off-brand Android TV boxes that probably cost no more than like $50 a piece. Anthropic said they were running Google's Gemma 4 across them, very slow, 39 cent tokens per second. So really slow, but over time, I guess they said eight out of eight nodes in four or five trials were able to pull off full replication.

and yeah, so there you go. Yeah, I don't know, you know, it would be famous last words if I dismiss this but it is amazing how it's constantly kind of an Eeyore marketing. Oh, we did something bad. Oh, we could do something bad. Oh, it's going to turn out bad. You don't know. You better let us, I don't know, what are they asking for? Better regulate us? Or you better regulate the industry? Regulate the other guys. Beware of this open weight stuff. If we do bad, at least you can come after us. You won't know who else did it if they use open weight. You don't want to know. They were also using GLM 5.2, which is the Chinese model from ZAI. Maybe this is a clever way to market open models. It ain't open. See what open models are able to do? You don't want that do you. So therefore, yeah. Marketing through, I don't know, mystery or I'm not really quite

sure. I'm going to take a quick break, come back, do a little speed round. We got, well, yeah, we have one story actually right at the top of speed round that kind of ties into what we're talking about at the top of the show. So hang tight. We'll talk about that in a second. This episode of the AI Insight Podcast is brought to you by Outshift by Cisco. That's Francisco's incubation engine for frontier technology. Multi-agent AI is on every enterprise roadmap right now, and for good reason. But here's what most people aren't talking about yet. Agents can pass messages, they can coordinate tasks, but they can't actually think together. And when they don't, you get what you might call cognitive failures, misreadings, unverified claims, false consensus, the kind of quiet underperformance that conventional monitoring doesn't even catch. That's the problem OutShift by Cisco is building the fix for, and they're calling it the Internet of Cognition. It's an open source foundation for building multi-agent systems from design all the way to production with shared context,

shared memory, guardrails to drive the results you actually expect from AI. If you're building in this space, this is worth your time. Read the paper, experience the demo, grab the code at outshift.com. That's O-U-T-S-H-I-F-T dot com. We thank Cisco for their support of the AI Inside podcast. The Wired Newsroom is known for award-winning reporting on how technology shapes our world. On Wired's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building, challenging, and reshaping the future. Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday, wherever you get your podcasts. All right. What do we have to close things out? Google Play, that's the Google Play Store,

now supports private vibe-coded personal apps. You can build an Android app in Google AI Studio, and then you can install it through Google Play just for yourself. So you can install it through there. You don't need a developer account to do this. It doesn't show up inside of Play Search. This is just using Google Play to install your own Vibe-coded apps on your device. So if you wanted to share it with others? Yes, sharing with friends and family comes sometime in the future. Not immediately, but it will come in the months ahead, they said. So if you want to sell this, you've got to go and do the normal Google Play stuff. Right. Right. through your vibe coding. You can install it on your device. Maybe you share it with friends eventually. And, yeah, if you wanted to market it or whatever, you'd have to go the next step up. But this absolutely ties into the first story, to the remake of Adobe and Microsoft, is that I made my own little app, and now I can use it, and I don't care whether I have to buy software. I can have the machine make it for me.

I made my own flashlight app or my own fart noise app. See? You're thinking really classy there, Jason. I'm thinking of the early days of apps. Remember how they were so simple? We loved them. They led to everything we have now for better and for worse. NVIDIA reportedly in talks to acquire Reflection AI. Reflection, we talked about last week, is the Open Weight Lab here in the U.S. They released Beam, which is the 501 billion parameter model that we discussed. Financial Times says the talks are early, could be an acquisition, could be an acqui-hire, or some sort of bigger investment, but NVIDIA already has around $800 million invested. And this is NVIDIA and Justin Wong supporting Open Weights. Yes. I think that matters hugely. Yeah. Yeah, and the Wall Street Journal CEO, Misha Laskin, actually said, open models are Trojan horses for the infrastructure they bring with them,

which kind of explains why NVIDIA might care so much. Along those lines, you've got Amazon supposedly, reportedly, and talks to buy AI startup Descartes for around $7 billion. Descartes makes Lucy, which is a real-time video editing model, also Oasis, which has simulated worlds for training robots and self-driving systems. This was valued back in May at around $4 billion. Anthropic, actually, we have talked about Anthropic doing their own due diligence in August to buy it for around $6 billion. They ended up passing on the deal last month for whatever reason. The reason I think that Amazon AWS is interested is that I hadn't heard of Descartes. But in one summary, I asked him to tell me what Descartes does. And they quote Seeking Alpha saying that it helps developers run and sync AI models across different types of computer processors and hardware.

So I think that that's the thing here. So you're on AWS, and you've got some NVIDIA chips, and you've got some new Microsoft chips, and you've got some, you know, you're in China. You've got Huawei chips. You've got whatever. Amazon Trinium chips. now you don't need custom programming libraries for each chip, which I think is a way to fight against NVIDIA's hegemony with CUDA. So that's why this rumored acquisition would seem to make sense. Got it. Okay. Okay. And also, I forgot to put this in here, but Daniel, thank you for the reminder and the YouTube live stream. Character AI, which is not an unnamed AI company, an unknown AI company. They've been around for five years. Chatbot development and stuff. Apparently, it's filed for bankruptcy. Chapter 11 bankruptcy. They've been facing a lot of lawsuits. Apparently, they've had some challenges scaling the business.

And, yeah. It's selling some of the leftovers to Disney. Yeah, yep. Agreed to sell some of its assets to Disney for $20 million, non-exclusive perpetual license for its technology. So that's interesting. I don't know. We've heard a whole lot about some of these bigger names that we talk about often filing for bankruptcy yet. Yeah, if we go back in our archives, since we have archives now, how long have we been doing this, Jason? Yeah, about three years. Yeah, if you go back two years ago and look at the companies we were talking about then, And which ones have we not talked about lately? Yeah, that's true. That's true. I know character was something that we've talked about over time. So this was surprising to me. But there you go. And then finally, there's Aurora, which you've probably never heard of because if you have to ask, you don't deserve to work with Aurora. invite only AI concierge for the ultra wealthy $10,000 a year plus a $5,000 initiation fee.

I imagine that's peanuts for the clientele that might sign up for this. But, yeah, it's concierge services with AI integrated into it, I guess. AI plus human staff. Connects your calendar, connects your whoop, your aura. Thinks ahead for you so we can book you a sauna or an IV drip after a very long flight. That sort of thing. Because it's the kind of thing that one does. One is supposedly not the kind of thing I've ever needed to do, but I'm not the clientele for Aurora. So I did a little showbiz earlier with Cupertino. You know where I'm going to be this weekend? I'm going to be in the movies watching You Can See Everything. Oh, of course you will. So that comes out this weekend. I can't wait. Wow. I mean, people have been. It's going to be really good. That's what I keep hearing. I'm ready to kind of see. The guy who directed it, I can't remember his name. He was really not on my radar until recently, until this, and then I was like, oh, wait a minute.

And went back and saw a couple of episodes of the rehearsal show on HBO. Have you seen that? I'm not aware of that. I need to. But, yeah, he's a very unique individual. Very, very interesting. I'm very curious about this film, especially the kind of rave reviews that it's made. Oh, people, I mean, long, standing ovations at film festivals, yeah. And fascinating to me is the social reckoning, and I never liked the social network. I thought it was awful. Yeah, right. Aaron Sorkin's follow-up bombed. Absolutely. Because I think that people worrying about social networks is so yesterday. and now everybody's concerned about AI and so no big deal strangely it's a few years too late maybe for that story because people have moved on to the next thing, the Cambridge Analytica isn't it based around the Cambridge Analytica? This version

as I understand I've not seen it and her revelations were about how they knew that things were bad for kids and they didn't do enough about it. Oh, okay. So there might be some two-year-old reporter who got the Francis Haugen material and put it out and so they're the heroes of it. Got it. Oh, I see. I see. Kind of not surprised. And we've got more movies coming. We have Musk out. We're going to have a movie about OpenAI. I haven't heard any movies about Anthropic yet, but it's got to be coming soon. Yeah, that's going to be put out by the cast of Saturday Night Live. And it'll be a hoot, let me tell you. I hope they make Dario Amadei a regular character on Saturday Night Live. This is the best. I don't know how they don't because that got so much pickup. And that's a new cast member. You know, they're always – I just watched the Lorne Michaels Netflix documentary.

Oh, cool. Yeah, and so it kind of goes into some of the behind-the-scenes stuff on how they cultivate their talents and everything like that. And then that's taken me down a Saturday Night Live alumni kind of like podcast circuit in the last couple of weeks. and just kind of hearing like when you're someone like David Spade and you're on the cast, but you don't have many things going into, many of your things being chosen for a night's episode, how much disappointment and doubt it puts in you. And as that cast member, I heard an interview with him, and he's like, you're always looking for the thing that catches and that you get a repeat kind of character on. And so that just made me think of that actress. And let's say Jane Wickline to give her credit. There we go. Yeah. Good call. All right. Well, that's it. We've ended the episode with our weekly look at Saturday Night Live.

Show biz. Yeah, show biz. JeffJarvis.com for the newly revamped website. Thank you, Jake Jarvis. As soon as we get off, I'm going to call him and say thanks. Great work, Jake. I mean, it's so great seeing literally all of your books are listed there now. And go to the podcast page. Uh-huh, uh-huh. And the podcast page. Everything is listed there. Intelligent machines. AI inside. You've got all of them appearing there. I also have another podcast I'm on every Friday called Breaking the News with Julie Roginski and Michelle Cottle. So it's, I mean, shouldn't I? And... I'll add that. Yeah. I'll tell Jake. He probably wasn't aware of it. He's like, yay, I'm done. Oh, I'm not done. Michelle Kinney and Julie Wiergenski on the Mayday Network. Oh, okay. I wasn't aware of that podcast. I just podcast all day long.

That's all I do now. You are a machine. Jeff Jarvis is a machine. He's kind of like a linotype, but with his mouth. I don't know. Sorry, that sounds weird. I didn't mean by something. I spew hot lid. All right. Thank you, Jeff. Everybody, thank you for watching and listening. Go to podtuneup.com if you want me to talk with you about your podcast or if you know someone who's in podcasts and having some challenges and just wants to talk to a podcaster about it. Hey, I'm here. It's kind of like therapy for podcasting. podtuneup.com. AIinside.show is where you can go to find everything you need to know about this particular show. all the links, you know, audio, video, store, live, you name it, Patreon link, it's all there. Speaking of the Patreon, patreon.com slash AI Inside Show. We have some amazing executive producers who help us each and every week. Dr. Dude, Jeffrey Maricini, Radio Asheville 103.7, Dante St. James, Bono DeRick, Jason Neifer, Jason Brady, Anthony Downs, Mark Starcher, Karsten Samoski, and Evan Muziel.

there we go it is a mouthful but I'm happy for it so there we go patreon.com slash AI Insideshow lots of perks daily podcast discord you name it go check it out thank you for the longer episode than normal Jeff appreciate your time and doing this with you I was really by jabbering taking up time that makes it longer but I love it we had great topics today it was so much fun yeah it was great stuff alright we'll see y'all next time on AI Inside take care everybody bye The Wired Newsroom is known for award-winning reporting on how technology shapes our world. On Wired's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building, challenging, and reshaping the future.

Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday, wherever you get your podcasts.