John Ternus Takes Over Apple in the AI Era

00:00:00:00 - 00:00:22:25
Unknown
Coming up next, Jeff Jarvis and I dig into Tim Cook stepping down as Apple CEO and what John Turnus new CEO here in a few months could mean for Apple's AI era. Anthropic return to the white House and what those negotiations might actually look like, SpaceX's surprising $60 billion bet on cursor, and Google's big agent enterprise push at cloud next.

00:00:22:26 - 00:00:39:01
Unknown
Going on right now, that's coming up next on this episode of the AI Inside podcast.

00:00:39:03 - 00:01:01:19
Unknown
Welcome to another episode of the AI Inside podcast. This is the show where we take a look at the AI that is layered throughout the world of technology. So many layers, so many different directions. Sometimes they converge in ways that make Jeff happy. I'm one of your hosts, Jason Howell, joined by Jeff Jarvis. Good to see Jeff. Sometimes.

00:01:01:21 - 00:01:32:05
Unknown
Sometimes not often, but sometimes I mean so. Okay. Before we get into the news, you about to like like share that your long time, years long complaint has been fixed? Or is this just one year long? Has it been that long? Just seemed that long? Yes. So I have been having s fits about not having the Ask Gemini on my Chrome browser in my Chromebook on Chrome workspace, for which I pay.

00:01:32:07 - 00:01:51:00
Unknown
Some people were saying I'm trying to get rid of all this stuff. Why are you complaining that way? But that's what I'm doing. So this is what I do. So every once in a while I would go to my admin tools and I would make sure there's no new thing I hadn't missed. And I went through everything. I go through this regular once a month, I go through every setting and I can't find anything.

00:01:51:00 - 00:02:15:26
Unknown
I have said yes to everything because I am the administrator over me. That's it. I'm the same guy. And so. But there's something new. There was asked Gemini in the admin tool about how to use the admin tool. So I thought, okay, all right. All right. Let's see how smart you are. My complaint you have to set this this as this.

00:02:15:27 - 00:02:43:26
Unknown
Yeah I did one time I said still no, you're wrong. I still don't have it. You have to send all this. No. Third time I said no. You're wrong. I don't have it. Then it came back with a different setting that I had never seen. That was instead of all, instead of under Gemini, instead of under apps, instead of under anything else under user is kind of hidden away there.

00:02:43:26 - 00:03:08:17
Unknown
There was a special setting where I had to set the default for Gemini to be shown. Now, mind you, the way this worked was that I should have not gotten Gemini in anything, but I was getting in Gemini in mail and drive and sheets and everywhere else, just not the browser. But when I clicked off, those three things restarted.

00:03:08:19 - 00:03:30:14
Unknown
Voila! I have asked him and I in my browser now ask me how much I've used it after all that fuss? Not at all. Yeah, you finally got there. Finally got it. It was a matter of pride, man. It was a matter of pride. Yeah, I hear you. Totally. Workspace settings. Everybody who uses workspace complaints about them, they're just awful there.

00:03:30:15 - 00:03:50:19
Unknown
I understand it, they're voluminous because they're trying to worry about every possibility for every school or business or all that. But it's a pain in the rear. But I'm there. So Google. I'll lay it off here for now. So then for for those who are in a similar situation, is I know you put in here the smart features setting.

00:03:50:19 - 00:04:12:09
Unknown
Is that the one in the in the rundown that I'm seeing, it's the menu account account settings. Click Smart Features for Google Workspace and set that to default. Is that the thing? That's the thing. But under users. Yeah under users. Yeah. Not under Google Gemini not under apps. Got it under users. And I had gone through everything and I won't work for any longer.

00:04:12:09 - 00:04:44:07
Unknown
But I had tried absolutely everything. I had gone through every single setting. I don't know if it was new, I don't know what it was, but Jim and I finally led me there after my third drive. These systems, I mean, everything that we do in technology that we work with these browsers and the the AI models or the AI services, the apps, whatever, like they're so complicated that, you know, it's it's so hard sometimes to find the exact right thing that tackles the thing that you're trying to do.

00:04:44:09 - 00:05:01:12
Unknown
I guess that's when search and using the LM, like you did to find it, makes a whole lot of sense if it can even identify it three times. Yeah, exactly. Yeah, exactly. And I don't know what goes on inside of the brain to think, okay, I did the opposite of the obvious and it stretches the one next. Oh, that there's this.

00:05:01:17 - 00:05:26:19
Unknown
Oh, I don't know how this this. Yeah. This is what you need. Yeah, yeah, I run into this when I work with, when I work with systems that are like, you know, like building out modular systems, like, I ran into this sometimes with area, which was a sponsor of the podcast and where sometimes, like, they have an agent built into the system.

00:05:26:19 - 00:05:52:23
Unknown
And in a perfect world, you fire off a conversation with the agent. It's a conversational, conversational LLM inside of the agent building process to say, this is what I want this agent to do, build it for me. And kind of similar to what you see in like standard LM experiences, everybody's probably run into this where it the answer you get is I figured it out.

00:05:52:24 - 00:06:10:27
Unknown
Here's what you do and I've done it for you. We've we've solved the problem. And then you try it and you're like, well, no, actually the problem is still there. Oh, I see what happened. I figured it out. Now, though, you do this and the problem is solved. Yay! We figured it out. It's like, no, it's not figured out.

00:06:10:27 - 00:06:31:01
Unknown
It's still not working. And every time, you know, sometimes with these llms, they come back as if it's like, oh, good, I'm happy you poked me because I got there and we figured it out. Let's move on. Now it's like, no, you haven't. Actually, if AI were employee, the thing is it is going to be an employee. Yeah, yeah, it's a frustrating employee.

00:06:31:03 - 00:07:03:20
Unknown
Yeah, yeah. Don't tell me it's solved if it's not solved. Not actually solved. Employee. Digital employee. Anyways, the reason I said years is because this is like the Gemini inside of Chrome is just one example or one aspect of your workspace complaints. So this is so this is one aspect of it I suppose that is solved. But the overarching most things Google don't work inside of workspace the way they do for other users.

00:07:03:20 - 00:07:30:26
Unknown
Problems. Yeah, that's it's a mistake I made years ago. If I had known the full implications where I decided that I was going to use my my blog name as my primary domain. Yeah. So my email and all that is that it's Bob's machine. And I would have been just as well off not doing that. Yeah. Oh the mistakes we don't even know we're getting ourselves into early on.

00:07:30:29 - 00:07:55:23
Unknown
Oh well, at least you're one step closer to I have a Paradise to a Paradise of a button that you may never use. Okay, let's continue talking about Google. Apparently today is the start of Google Cloud Next 2026. And so this is you know, all about enterprise topics in the cloud, Google's infrastructure in the cloud related to the enterprise.

00:07:55:23 - 00:08:21:17
Unknown
And that is exactly what they've been pitching. They've actually been pitching to start off with let's see here at enterprise. So they're rolling out a new Gemini Enterprise agent platform and app. So get rid of that pop up there, which basically gives companies the ability to build manage schedule agents across all of its internal systems, across its business apps.

00:08:21:19 - 00:08:41:07
Unknown
They actually shared the 75% of its customers use AI in their businesses. Kind of unsurprising at this point, but this just kind of made me think like, okay, so Google, you know, the open claw ification of everything continues and Google getting on that hype train is essentially what this means.

00:08:41:09 - 00:09:10:23
Unknown
I think sizes too true of Google. We've talked about this often that Google's tentacles are bigger than other companies, right? OpenAI is a model maker. Anthropic is a model maker, Google is a model maker, and it's a chip maker and it's a host. And I think that it's it's more diversified within AI than most anybody else. More than Microsoft, more than Amazon.

00:09:10:25 - 00:09:30:20
Unknown
One thing that we that's not in the rundown but but Google also is going to make custom chips with Marvel and right. That says that they're going to be more and more in competition within video. And so I think that's important to look at.

00:09:30:23 - 00:09:56:24
Unknown
As, as Google's larger strategy. So the Google Cloud isn't just Google Cloud. Google cloud is you need chips. We got chips, you need models, we got models. You want to use somebody else's models. We got a cloud. You can use their models. I think it's a an impressive strategic position to be in. So then this Marvel story is that, is that related to their TPUs or is that separate from.

00:09:56:26 - 00:10:23:01
Unknown
So they were working with Broadcom on some things okay. Now they're working with Marvel. So Broadcom stock this is a story. So stock went down. Yeah it's reports. So it's not confirmed that they would design two new chips for artificial intelligence. What it could include a TPU as well as memory processing unit. So this also goes to inference.

00:10:23:03 - 00:10:48:24
Unknown
And both Marvel and Broadcom help their customers translate chip designs into silicon for running back into support. So that's what Google was doing with with them. So I think it's the combination now I would imagine TSMC I don't even know TSMC probably makes eligible for everybody everybody anyway. But how to get them there? I think that's where they go.

00:10:48:25 - 00:11:17:00
Unknown
Again, as you and I have confessed many times, chips are complicated for us. Yeah, definitely not my strong suit. You know, you've confessed to the same, but from a broader perspective, I continue to be interested in this idea that, like in vedere has has really dominated largely the story here in the last couple of years and now especially with with Google and as TPUs that it once just kind of held for its own internal use.

00:11:17:00 - 00:11:44:20
Unknown
Now kind of farming those out or offering those out, applying, like you said, some competitive pressure against the power of Nvidia and the marketplace and actually what they were announcing at. Okay. And so first of all, this Marvel story, they do specify Google. This is from the article. Google has relied on Marvel rival Broadcom for the design of its in-house TPUs, its tensor processing units, which would, you know, imply according to the information.

00:11:44:20 - 00:12:13:20
Unknown
Anyways, that Marvel is now working with Google on on the TPU production. And that does relate to today's news, because Google did also announce two eighth generation TPUs right at this event eight for training and AI for inference. And yeah, this is, you know, obviously a clear step forward in processing power from the previous generation that was called Ironwood.

00:12:13:20 - 00:12:34:26
Unknown
And yeah, I guess it's you know, this is an important place for Google to be. On to your point, it's a strategic necessity for for Google and Google Cloud to do this there. I think they're in a strong position to compete with on this on this front. You know how it's going to land. I suppose that that remains to be seen.

00:12:34:26 - 00:12:46:23
Unknown
But Google has a lot of funding. They're very technically credible in all of this. So yeah, I think it could be a good it's a good direction for them, I think.

00:12:46:25 - 00:13:20:24
Unknown
So I wanted to mention what's his name? Dwarka Patel did an interview with Jensen Wong, and he emphasized it's entertaining listen to because Jensen Wong disagreed with him. And Jensen Wong is not only a great communicator and a great educator and a great performer and presenter. He's also a really good debater. He just he took him down mainly about China and Jensen Wong saying, it's not in our interest to cut off China because then we're not going to force them in other ways.

00:13:20:24 - 00:13:53:02
Unknown
That was the argument. But in the podcast, Wong Nvidia is in a position where he emphasized, we can't pick winners. We will. We want everybody to succeed. We will supply everybody that the emphasis that they care about most is that you you build on Cuda, that if Cuda is is the operating system that's everywhere and developers build on it because it's everywhere, then you're going to buy Nvidia chips and that puts Nvidia in the strongest position okay.

00:13:53:03 - 00:14:19:09
Unknown
That's their strategic arm as you're talking Jason, I was thinking, well do I wish that Google would do the same kind of presentations that Jason Wong does every time there's a new tensor chip? And maybe they did at this presentation, but it's just not promoted as much. They don't promoted as much, I should know and watch. But again it's what does though they own the market like TSMC owns the fab market.

00:14:19:12 - 00:14:32:24
Unknown
Google does that too, but only a part of it. And Google will also have Nvidia chips in their cloud because they need to be able to have them because people were running on Cuda.

00:14:32:26 - 00:15:00:02
Unknown
But Google again, that's only one one leg of their stool, so they're in a stronger position, I think, than any in my mind. Any of those other companies, they have, the underlying hardware they have hosting, they have models, and they also have the consumer business that Nvidia doesn't have. And I think that's that's related to look at that.

00:15:00:03 - 00:15:26:29
Unknown
If you look at, in comparison, anthropic, OpenAI, Microsoft, Amazon and Nvidia as key AI competitors, they're all missing one of the things that Google has. Yeah, Google kind of has the complete picture. Yeah yeah yeah that's a really good point. And that kind of, you know, also reminds me of nother aspect of this announcement, which is workspace intelligence, which is basically like a contextual layer for workspace.

00:15:26:29 - 00:15:56:15
Unknown
So all of Google's, you know, software or cloud apps or whatever, Gmail, chat, docs, sheets, slides, all that kind of stuff tied into the projects, the people, the workflows, as Google mentioned, basically like personal intelligence, but for workspace, which they had released personal intelligence a couple of months ago, which basically brings, as a general user of Google products, how you know the data that is within your Google environment into the Gemini experience, this doing that for enterprise and that is.

00:15:56:17 - 00:16:27:02
Unknown
Yeah, that's to your point. That is a power and capability that Google has that a lot of these other companies are trying to kind of build up, but don't already have the benefit of having had that infrastructure in place for years before this moment happens, that they can just kind of plug and tie into. They have to they have to kind of overwrite the habits and the the spend and all that of the people in order to get them into their, their world, whereas Google kind of already has them.

00:16:27:02 - 00:16:47:00
Unknown
And so connecting all these pieces becomes a major opportunity for Google by comparison. So I want to mention one of the things we've talked about chips related and not related. I just finished reading a wonderful book called Conquering the Electron by Derek Chung and Eric Brock. I think it is. I can't read them, I, I listened to it.

00:16:47:01 - 00:17:23:07
Unknown
It's excellent. Conquering the electron Derek Chung there it is. And so it carries the development of electronics from the very beginning, from the amplifier and from Edison on. But at the end, Chang himself worked in the chip business. And so it has some pretty clear explanations of the technology that got us to that point where we are now and what happened at Bell Labs, what happened at Fairchild, what happened at Intel, and so on.

00:17:23:07 - 00:17:49:00
Unknown
And it describes these various technologies and developments and how they went on in a way that I felt I could understand. The other interesting thing that struck me about adjacent is that is that you could be a tinkerer and invent radio. You can't be a tinkerer and invent tensor chips, right? Right. The level of complexity is just miles beyond now.

00:17:49:00 - 00:18:07:02
Unknown
In a way, it turns back around, though. It's because those things are so complex they enable anyone to program. So we're back to the point where you can be a tinkerer again, and you can make software that can do it. It's a different part of the curve or different part of that. Yeah, what you're talking about. So I recommend the book.

00:18:07:02 - 00:18:28:27
Unknown
Highly interesting. Okay, cool. Gosh. And I feel like when I did that search, it came back with like, is this a is this a PDF book or is it because I feel like what I'm looking at. No it's not. I see what I had pulled up was more like probably a summary, I guess of it, than the actual book.

00:18:28:29 - 00:18:46:07
Unknown
Oh, okay. So probably somewhere like a whole presentation around it. You want to. There you go. If you want to know some of the information from the book, hey, for for it on Google PowerPoint has been the ruin. People say screams of the ruin of us. Social media is the ruin of us. AI is the ruin of us.

00:18:46:07 - 00:19:16:20
Unknown
No PowerPoint is the ruin of all thought because you see a bunch of pictures, I think. Well, I understand that now. Yeah. You don't understand, Jack. Yeah, yeah. Here's a picture of a light bulb. But an arrow that is. Okay, I get it. Yeah, yeah, yeah. Well, shifting gears a little bit, there is Apple, which, you know, I always like to challenge myself to talk about Apple on, on my podcast because my other podcast is an Android podcast.

00:19:16:20 - 00:19:53:25
Unknown
And so we don't get many opportunities on that podcast to talk about Apple here. Things happen. We haven't had as many opportunities because Apple has been seen rightfully or wrongfully so as a laggard in AI, let's say depending on who you talk to. And so this announcement, I think, is it's interesting timing that Apple announced on Monday just yeah, just two days ago, the afternoon that CEO Tim Cook is going to step down this September, senior VP of Hardware Engineering John Turnus is going to take his place and also join the Apple board in the process.

00:19:54:01 - 00:20:27:07
Unknown
Turnus actually joined Apple back in 2001, became VP of Hardware engineering in 2013, and through that time has essentially led teams driving most of Apple's biggest hardware efforts throughout that time. Name it he's probably you know, was was the lead of of that hardware for the last decade and almost a half. So sounds like a solid person to take the mantle from Tim Cook Tim Cook is not leaving the company entirely.

00:20:27:07 - 00:21:01:02
Unknown
He's he's staying on and probably going to be doing a lot of the the handshake events with, you know, President Trump as executive chair, as an executive chair. And, you know, do do a lot of that kind of, I don't know, that kind of PR and relationship building for the company outside of the CEO position. But Turnus being a hardware guy coming into Apple when Apple is in this moment right now where, like I said, whether you agree or not, there is this perception that Apple is kind of behind when it comes to AI.

00:21:01:05 - 00:21:25:20
Unknown
Now, there's this big transitionary moment, this big leadership change. I mean, as as big as it gets. Right? The last time this happened, Apple's CEO Steve Jobs passed away. And Tim Cook, you know, that was one of the biggest stories of that decade in the world of technology. And so these moments don't happen very often. So I feel like it's worth talking about because this could really signal the beginning of the next era for Apple.

00:21:25:20 - 00:21:52:01
Unknown
And maybe that's part of the reason why we're hearing about it now and not a few years down the line. You know, maybe it's the right time for it. And tribute to cook. He built the company the hardest shoes to fill. He built the company huge and tremendously criticized trillion trillion from something like 60 billion. I mean, the the growth is astronomical.

00:21:52:05 - 00:22:14:20
Unknown
He did an amazing job in his role. Yes. Exactly. Yeah. Up there with Nvidia which went higher, higher, faster, but still quite amazing criticized because there wasn't that much innovation. But there has been that much innovation in all the genius we talked about this last week. I mean, yeah, yeah. That's true. None of the goggle stuff has taken anybody off the watch.

00:22:14:20 - 00:22:35:01
Unknown
Is the watch okay. So everybody has that to the phone. Is the phone. There's still the you know, even though I'm an android like you, people will say it's the best or it's the leader or it's the, the benchmark. The computers. I just bought it sitting over there, unopened, a neo. Well, congratulations. I'd be curious to hear what you think.

00:22:35:03 - 00:22:50:14
Unknown
By last week, when I revealed that my Mac mini is 12 years old on the show last week, I realized it was pretty stupid of me. And so I went and got one. So you got the neo to run the show on it. Like that's going to be your show computer. That's interesting. I'll be curious. Yeah, yeah, I'm sure it'll be step.

00:22:50:15 - 00:23:09:00
Unknown
I mean guaranteed is going to be a step up that that exact Mac mini has been around for the back of. God bless it. You know, it's working fine. I mean, it's still rocking. You wouldn't know. Yeah, I wouldn't know. Yeah, totally. So when I listed the competitors to Google around AI and the AI universe, Apple is not not on the radar.

00:23:09:02 - 00:23:35:18
Unknown
Not at all. Right? We know that Apple is using is going to use Gemini for Siri. So putting the hardware guy in I think says that's where our fortune remains. And rather than chasing the mythos pardon me of AI, they'll use it. They'll involve it, they'll have it. They'll figure it out in an apple way. They don't have to be ahead of the of everybody else because they're not going to be.

00:23:35:19 - 00:24:00:11
Unknown
It's too big an investment. I think they still have the confidence that their hardware will be it. My hope is, I mean, we'll see whether Johnny I've and and Sam Altman come up with anything on OpenAI that changes the paradigm, as we like to say. When I worked at Delphi many years ago, if we used the word paradigm, we had to put $5 in a jar.

00:24:00:13 - 00:24:29:20
Unknown
I remember that. So we'll see whether or not somewhere outside does it. But still, Apple's the most likely, I think, to invent whatever it is that is going to be our primary connector to this digital ring. So yeah, it'll be interesting to watch what happens. Cook will still be there, still watching it over. I think that he and Apple got a lot of criticism from certain quarters about the politics.

00:24:29:23 - 00:25:05:19
Unknown
You know, I was thinking about this, you know, Jensen Wong is also played up next to Trump. All of them. Have Zuckerberg's been there? Cook I think, got more criticism than others, mainly because I think people expected something different from Apple that apples are different. Technology companies hold them different. Different standard. Yes. Yeah, yeah. There's something about their vision, the perception of the kind of the company ethos or the vision around the company that has long kind of painted it as a counter to that.

00:25:05:19 - 00:25:24:07
Unknown
And so then there was, I think, in the, in, you know, in the news and everything, there was a bit of a dissonance around seeing Tim Cook playing ball, whether it was right or wrong. You know, the perception is reality. Did he have a choice to a large degree? I mean, did he have a choice? Actually, my politics are clear.

00:25:24:09 - 00:25:54:00
Unknown
I was disappointed, but how much did he give up in giving up a gold bar? Versus how much did Sam Altman give up in following in after anthropic and trying to get open AI in defense? What is anthropic going to give up now in negotiations, which we'll get to in a few minutes. Right. So it varies. I don't think that cook did anything.

00:25:54:02 - 00:26:15:21
Unknown
Besides lend his face and good name and the company's good name to these political events. So fair or unfair that he took his level of criticism? I think it's stunning him. I think it was difficult for him and now he I don't know why he chose to leave now. I think it was mainly we're on top. We're doing well.

00:26:15:23 - 00:26:34:12
Unknown
We're $4 trillion. It's not I think some of the some of the reports that I had read in recent months were that this transition, a transition, I don't know if they were naming Turnus as like a guarantee or anything like that. I think his name would, would turn up as, as far as possibly replacing cook at some point.

00:26:34:12 - 00:27:11:17
Unknown
But the, the reports were saying that something like this was planned sometime in the next like two years. And so if it was two years, let's say it was two years from now, and then they're, you know, positioning it to now I that just has me a little curious as far as yeah it does then. Yeah. You know, but to get a sense of Turnus kind of mindset going into this role, Bloomberg had reported that Turnus had already begun reorganizing the hardware group around a new internal AI platform that was that is devised to help drive product development.

00:27:11:17 - 00:27:35:08
Unknown
Apparently they're using cloud for part of that, so that could be a signal as to his intention to kind of lead around. And with AI as he kind of executes his new kind of vision for the company in this new role. And what is it like for Apple to have a hardware guy in the top seat? Because cook was not a hardware guy.

00:27:35:09 - 00:27:56:24
Unknown
Steve Jobs was. And we're, you know, like you like you alluded to, we are in this moment right now where it's like, what is what is AI hardware? We do we don't even really know what that is yet. Is that just AI integrated into the hardware we already use the smartphone or is that something different? And if that is something different?

00:27:56:25 - 00:28:26:20
Unknown
Well, we're going to find out in Turnus leadership what that actually means for Apple. Yeah, yeah, yeah. Be interesting. Big moments like this, you know, big turn, turn over. Events like this don't happen very often. So I'm, I'm super curious to see what this means for Apple going forward. Yes, indeed. All right. So I want to take a quick break and thank the patrons, our amazing patrons who support us on this show each and every week.

00:28:26:20 - 00:28:50:12
Unknown
We could not do this show without you. And I want to throw out a special thank you to my cousin Vince Wilkie. Vince. Coz coz supports us. What's up? Vince supports us also. Jason K, thank you both so much for supporting us regularly to do this show. We could not do it without you Patreon.com sideshow and we are going to take quick break, come back on the other side of it.

00:28:50:12 - 00:29:01:29
Unknown
And you, you mentioned it. We're going to talk about anthropic and the return to the white House. What does that even mean? What is that leading to? That's coming up here in a moment.

00:29:02:01 - 00:29:27:27
Unknown
All right. The white House brought anthropic back into, well, the white House. The Pentagon, as you remember, had labeled the company as a supply chain risk because they didn't like that anthropic, didn't like their AI tools being used for defense and for warfare, for certain aspects of certain aspects. Yeah. Apparently on Friday, anthropic CEO Dario Madhi. How do you say his last name?

00:29:27:27 - 00:29:53:29
Unknown
I always mess it up. I do gif gif with his last name and is it al-Mahdi Ahmadiyya a day I would say on a day. All day. Okay. That's right, that's right. It is close to Amadeus, just without the use. And I need to remember that when I see it. Ahmed Dario Ahmed visited the white House. How do we pronounce the name of the founders of anthropic?

00:29:54:01 - 00:30:18:12
Unknown
The founders of anthropic are Dario and Daniela al-Mahdi. Their last name is of Italian origin. There you go. According to Gemini, I didn't ask. I should have asked Claude, obviously. But yeah. I wonder if you were to say no. It's on the diet. It would be like, oh, you are so right. You are brilliant. That's true. I don't know how I got that wrong.

00:30:18:14 - 00:30:41:19
Unknown
Anyways. And I wonder what Claude would say to that. To productive and constructive talks is apparently what the white House was sharing or I can't remember if that was anthropic sharing that. But I imagine a part of this is, of course, the the timing and release of mythos. I think that's the most fascinating thing about this. I think the suddenly like, oh, shoot, we want that.

00:30:41:20 - 00:31:07:21
Unknown
We need that. We kicked them out. They've got the most advanced dangerous thing there is. G that sounds like fun. We should have to write up our reality. Yeah. And I think that there were there were reports of someone government were sneaking around and using mythos finding, getting access to mythos anyway because they really wanted it. So I think it created a fascinating osmotic pressure from the other side.

00:31:07:24 - 00:31:32:03
Unknown
Yeah. That's true. What was it? Yes. The mythos model being accessed by unauthorized users. I mean, you give a select group of people inside of these major companies access to something that they're not allowed to give out. That's still a lot of people that have access to this thing. I'm kind of unsurprised that at some point someone's like, well, here, just oh, just check this out.

00:31:32:05 - 00:31:52:07
Unknown
Yeah. It went through and reported that it went through one of the companies that that they did send it to. And from there I saw another story about copyright and all of this right before we got on. I didn't put it in the rundown. Some guy came along and he put up the leak of of the earlier stuff.

00:31:52:09 - 00:31:58:13
Unknown
Not not mythos, but the earlier model. Okay. Yeah. Onto.

00:31:58:15 - 00:32:18:03
Unknown
GitHub. And his girlfriend said, aren't you worry about copyright. And so he used models to just rewrite it all in a different language, but it can still do the same things. And so the story is wondering, is copyright really ready for prime time anymore? Yeah. The answer is no. Yeah.

00:32:18:06 - 00:32:45:03
Unknown
So the interesting thing about what happens here is anthropic put a clear boundary. This goes back to the discussion about Steve Cook a clear boundary. You shall not use our tools for autonomous killing without human intervention. And you shall not use our tools for surveillance of American citizens. So I think that's a bit oddly limiting. But if I were in Germany, I'd be pissed.

00:32:45:03 - 00:33:10:25
Unknown
But anyway, those those that was the boundary that that that the days at anthropic put on this. So I don't think they can back off that. I think they may find some semantic tricks here that if they are seen to back off, that for all the hay that they made, for all the new users they got, for all the advancements that they've made against OpenAI, that could be lost in a flash if they make the wrong deal.

00:33:10:26 - 00:33:36:19
Unknown
So it's a very sensitive thing, and we'll see where it goes. I think that people like Bisset will get the subtlety of that, but Seth won't. I decide how to use the tools, right? And so, so so then would anthropic then stick to its guns? No pun intended to say. Okay, well then I don't think so. It's in there in a hot collar moment.

00:33:36:21 - 00:34:07:17
Unknown
One hand. You think anthropic wants a deal with the white House? It's a good question. What do you think? I don't know, I mean, if they're going back to the table, then I guess that signals something. They're certainly open to it. But if, if if it truly goes counter to, you know, their ethical kind of promise of the company and, you know, then, okay, then you kind of know what you're going to get or you have a good sense of what you're going to get, and the experience has shown you what you're going to get.

00:34:07:18 - 00:34:33:18
Unknown
But at the same time, I guess, you know, if the white House invites you in to say, let's talk about this. And they have identified you specifically as a supply chain threat, I suppose it behooves you to go and explore and see see what comes out of it. I'm authentically curious to see what comes out of this. Like you said, like if there is a deal met, is it truly a deal that adheres to the ethical boundaries that anthropic has laid out?

00:34:33:18 - 00:34:53:29
Unknown
Or as you say, have they wordsmith did in a way that they can still claim one, but they get the other. And I don't know. The question is who called whom? Yeah. Right, right. I think I was in a better position at the white House. Like, who's calling a white House? Let it go to voicemail. We'll see if they.

00:34:54:01 - 00:35:21:14
Unknown
Yeah. We'll see. Yeah. We'll see what they say. Yeah, yeah. Very interesting. You know what's also interesting about the the mythos leak thing that I saw is so okay, so you kind of alluded to this. Bloomberg said that the group combined credentials from their contract evaluation work. I gathered clues from a mercury data breach and then guessed anthropic internal URL patterns to locate mythos.

00:35:21:14 - 00:35:32:00
Unknown
So that's kind of how they were able, apparently, according to Bloomberg, to find it. And then they proceeded to use it to do things like build a website.

00:35:32:02 - 00:35:54:26
Unknown
Yay! We got the most powerful model according to Let's Build a Site. Oh, wow, it was fast. That thing looks good. I want to build my landing page with mythos. Geez. Anyways, hacker mentality yes. Interesting times. Okay, maybe you can make sense of this next one. Actually, there's probably an easy way to make sense of this next one.

00:35:54:26 - 00:36:23:29
Unknown
And that is IPO. But SpaceX has struck a deal with cursor. Cursor is in the very popular or very notable I think. I think it's popular I hear people yeah I, I developer tool going to co-develop a quote coding and knowledge work AI and quote that is backed by SpaceX's Colossus supercomputer. SpaceX is saying that Colossus can deliver compute equivalent to a million Nvidia H 100 chips.

00:36:23:29 - 00:36:46:14
Unknown
So there. Yeah. And I think what's interesting here for me anyways, is that space has the option. Like they're announcing this, but they haven't really done either of these things. It has the option to either pay cursor $10 billion for its work, or to exercise an option to acquire the company for 60 billion later this year. And like, yeah, we could buy you later.

00:36:46:15 - 00:37:13:00
Unknown
I don't know, maybe we'll see what I think. That depends on having the IPO and having more cash or having or having the value in stock. What confuses me about this is the cursor side of the deal. Yeah. When I listed the leaders in AI, I didn't. I sure as hell didn't list Grok Space XII, Twitter, whatever the the Musk empire is.

00:37:13:02 - 00:37:39:08
Unknown
And and of course there is popular and of course there probably could have done a deal with anybody. Maybe not this rich, I would imagine so. Yeah, right. But is it worth $60 billion to anybody else? Probably not. Did did greed alone take them to this? Did nobody else want them? The story says that two of the lead engineering folks at cursor left to go to join XII.

00:37:39:09 - 00:38:09:24
Unknown
So that's clearly where the bridge was built. Oh I see. But for the management of cursor, it shows my prejudice against mosque. But why? Why would you go and and hang on to that rocket ship? Yeah, well, because cursor had a 50 billion target valuation, and this is a cool $10 billion higher. Yeah. Say slash space slash Musk saying, look, we've got tons of money.

00:38:09:24 - 00:38:28:11
Unknown
We'll throw your way. Does that sweeten the pot? Does that bring you over to our side. And the other thing my cursor is cursor allows you to use multiple models. It allows you to use Claude. Neat. Yeah. And so now if you're a cursor user and and you've gotten to love it. Are you pissed off? Are you going to be cut off from using what you want to use?

00:38:28:11 - 00:38:55:14
Unknown
I don't know, I don't know what the answer is. Yeah, but it potentially limits the core. What are the core benefits of cursor. Sure. Yeah, yeah. And I mean, it's also interesting to me that space like the entity of space, which I mean, it's in the name, you know, and maybe it's just an assumption that I make is, oh, cool, this and that company is all about, you know, space exploration and advancements there.

00:38:55:17 - 00:39:22:25
Unknown
I guess you could find a way to squeeze this in there, but it really does does seem like a way to kind of fill the gap a little bit. If you didn't include XII in your list of the top whatever companies, for whatever reason, part of that reason is because there I think there is a more broadly kind of defined perception that Z just isn't as good as the other ones, you know, be it because of the Musk relationship or because it's just different, not good, not as good or whatever.

00:39:22:25 - 00:39:47:11
Unknown
Not, you know, they don't they don't capture the news headlines the way the other ones do. And I do think that that they must relationship probably is a part of that. But so does space potentially owning cursor at a time where we're kind of the drumbeat of IPO continues to get louder and louder. Is this a move to just kind of patch things up a little bit and be like, see, we've got some we got some big, big stuff brewing as well.

00:39:47:17 - 00:40:05:23
Unknown
Yeah, I look, I look at XII and grok, and I'm reminded of the great line from succession. You are not serious people and I don't think anyone takes them seriously. And the New York Times just had a story today saying, oh, well, now Musk is changing all of his strategy. It used to be that space had to go to Mars.

00:40:05:23 - 00:40:15:07
Unknown
Now maybe the moon, now it's AI, you know, what the hell is it going to be? It's the whims of one person. Yeah, yeah.

00:40:15:09 - 00:40:42:23
Unknown
Yeah. Seeing a lot of that lately. Been dreams. Yeah. Well, transitioning to something that I only sort of understood when I looked at it. So hopefully you can do a better drivers. This is on you, man. Yeah. You put this in here. This is firmly in your court. But I'm also interested because we do talk about friend of the show Yan LeCun and his his kind of continual path towards world models and and jeopardy and everything.

00:40:42:23 - 00:41:10:27
Unknown
And this is a new research paper about the world model. I find that kind of funny, but I guess it's legit as well. So it's, you know, it's all in the it's all France. Yeah, yeah. And France or LeCun find it kind of probably a little bit of both. But this is a new world model architecture. It applies to the predictive approach.

00:41:11:00 - 00:41:32:26
Unknown
Robotics tasks from straight pixels is kind of the idea of what has been kind of driving towards building world models out of pixel analysis and video and that sort of stuff. But my understanding of the world models thing does not get me to a point to where I truly understand this report. Can I try to read the paper?

00:41:32:26 - 00:41:48:27
Unknown
And it was beyond me pretty quickly, but I came to the paper from a post from Aladin Ayade on Facebook and I saw that John had shared it. So I figured, okay, that's a blessing and.

00:41:49:00 - 00:42:23:00
Unknown
So what the what the paper explains and the part I could understand is that world models fall apart when they make everything seem the same. So a chair or person or whatever kind of gets reduced to it's the same and and that and to get not to do that required a lot of intervention. That intervention was expensive. So what enables what they've shown that it enables and what what low world model loom.

00:42:23:02 - 00:42:51:09
Unknown
Loom is it trains from raw pixels using only two losses a net embedding prediction loss and gauzy and regularizer. Okay, I'm lost already. Right now. I don't know what that means. But it's a way to try to understand. So it got rid of that problem. One thing that I've heard the Cohen explain it's a line I love. You can't tokenize the world.

00:42:51:12 - 00:43:08:18
Unknown
And if you tried to to get every, every single pixel and figure it out, you're paying attention to a lot of absolutely irrelevant data. The example it's given in one of these, either this or the.

00:43:08:20 - 00:43:19:02
Unknown
Gemini explanation I got from from the paper, is that you have a room and you got a chair and a person and a ball in it.

00:43:19:05 - 00:43:40:01
Unknown
The models are going to pay just as much attention to a flickering light bulb. And it's it struck me that this language isn't used here, but it struck me as as going back to the core insight that led to generative AI. Attention must be paid models. Attention must be paid. Right. And that. What do you pay attention to?

00:43:40:07 - 00:44:01:17
Unknown
Where do you put the computing power on the value here? And so what has talked about is the need to get to kind of concepts rather than pixels. That's a chair. People sit in the chair, you toss over the chair, it falls. And I'm probably doing a miserable job of this, I think. But but that's why that's how I understand the difference here.

00:44:01:17 - 00:44:27:24
Unknown
And and but what really struck me, what made me want to talk about this this week, is that I'm reading from a yards post. The efficiency gains are striking. The model has around 15 million parameters. That's nothing compared to other models, right? Billions of parameters. And here's the key. It trains on a single GPU in a few hours a single one.

00:44:27:26 - 00:44:55:11
Unknown
So so the whole notion of we've got to scale, we've got to build these huge data centers to do anything of value. This can plan actions up to 48 times faster than larger foundation model based world models, while staying competitive with several 2D and 3D tasks. So it doesn't mean that generative AI is a dead end. The post continues.

00:44:55:12 - 00:45:26:25
Unknown
Llms remain extremely powerful, but it does reinforce a key technical point for world modeling and physical reasoning. Predictive latent space approaches may be far more compute efficient than brute force generation. The real shift might be this not models that generate everything, but models that understand enough of the world to predict what actually matters. Right? And this has been flickering is is pretty doesn't matter inconsequential to the element in the or the object in the room or whatever.

00:45:26:27 - 00:46:01:20
Unknown
Right, right. And so and so. So I didn't use the phrase before because I forgot it. You have representational collapse is what they're trying to avoid that world models are a different strategic path. But they had this problem of collapse. This paper is how they managed to fix a good part of that collapse. And then in the net, showing how it can be trained far, far, far in infinitesimal amount of compute versus others.

00:46:01:23 - 00:46:25:12
Unknown
So I think that's going to what's interesting here is to see this, this battle. And so you see on the one hand you have and again, we've said it a few weeks ago at DeepMind did start saying, oh yeah, there's other ways in World War then llms and scale and and role models matter. OpenAI still has not said that.

00:46:25:12 - 00:46:48:09
Unknown
I don't think anthropic has said that yet, but DeepMind, Jensen Wong are saluting this flag, and certainly on the Kun and Fei-Fei Li are in the world model camp. So sorry folks, that I can't fully understand this. I could do about as well as describing chip architecture, but I think that this was worth pointing out mainly for that one line.

00:46:48:12 - 00:47:10:13
Unknown
About 15 million parameters on one GPU at a few hours. Yeah. That's certainly yeah, that shines a light on this. The, the potential, the the limited scale maybe limited is the wrong word, but the reduced kind of scale of what we what I think a lot of people assume about how these models need to work in order to get to whatever next, next level or next phase.

00:47:10:14 - 00:47:30:05
Unknown
And this is saying and by the way, I think when I did open the focusing on the right thing, but I did open the paper, I wanted to understand. I now had asked Gemini on my browser, there you go, ask Gemini. And it did a an explanation that was about as good or if not better than the post.

00:47:30:08 - 00:48:01:17
Unknown
And I put this in here to there was a I think you're going to hear coming up because that's what Lacan's model is. So there's a story about badass a great title, badass 2.0, the World's best incident prediction model, period, mainly, as I understand it, around cars for now, but it's built on jeopardy. And and I think that we're going to see more things that start once once more comes out around and around Army, which is the company.

00:48:01:17 - 00:48:32:24
Unknown
And I forget the name of this company we're going to see. We're going to start to see more things built around this, this, this, this paradigm, $10. It's going to cost me and, and start to see where this, this goes. And the point here is whether it's robotics or other or self-driving cars or factories and digital twins, it's trying to predict actions based on the real world.

00:48:32:26 - 00:49:05:13
Unknown
And for those of you who are listening, Jason has up a demo of it and a car just hit a guy. Oh yeah. So I'm looking at the video just trying to figure out like, what is the element of this that I should be plus risk 35%, 25, 6080, 90. So it okay, sorry for audio listeners because this is probably not that enjoyable, but it is a dash cam footage that is apparently, I'm guessing running this next hour technology.

00:49:05:13 - 00:49:32:12
Unknown
And the car that is driving through an intersection turns left through the intersection. And if you know what you're looking for, you can see a bicyclist kind of on the sidewalk coming towards the sidewalk. As it gets closer, the collision risk readout suddenly jumps exponentially. Now what this says about world model, that's that's I think the question that I have is, is how 100% used this model used to do this.

00:49:32:13 - 00:49:51:01
Unknown
And so let's put beneath that I would I would lose my right. And how does this come. Like what would this video probably impossible to tell, but what would this video have looked like if it wasn't using. Yep. And it was using other right collision detection systems. You know, I guess that's what we don't know. But interesting. Okay.

00:49:51:02 - 00:50:03:21
Unknown
And my bottom line here is I welcome multiple paths. I think it's good that we have competition to everything is LMS and everything is scale.

00:50:03:24 - 00:50:20:09
Unknown
I'm not denigrating LMS. I'm not denigrating what they've they've they've accomplished with scale. It's amazing. But to have other paths of research and to put resources into those other research paths, I think is critically important.

00:50:20:12 - 00:50:25:08
Unknown
Listening to to this book about about the electron.

00:50:25:11 - 00:50:50:07
Unknown
It really struck me about once again, I recently also read a history of Bell Labs and everything that came out of the pure research of allowing researchers the freedom to pursue different paths. And that's how we got the transistor, is people who cared about about solid materials. Physics took that knowledge. They had a challenge. They wanted to replace the the triode vacuum tube.

00:50:50:07 - 00:51:18:29
Unknown
That's what they wanted to do. It wasn't as if it was just pure research and it had application. Now, in fact, AT&T cared about it only insofar as it benefited phone calls. They didn't understand. Sorry. They didn't understand everything else that was going to go on. I shouldn't just like that one. The timing was pretty typical. That's what phones used to sound like.

00:51:19:01 - 00:51:24:11
Unknown
Sorry you couldn't have planned that, I mean.

00:51:24:13 - 00:51:43:18
Unknown
That's awesome. Oh, the synergy right there is just perfect. I love the fact that I have the old farts ring. Totally. Oh, I remember, yes, that's a good ring. It's a good. That is a good ring. Because it sounds like a damn phone should sound. That's right. Sounds like it's not a digital representation of a Drake song or something.

00:51:43:19 - 00:52:27:23
Unknown
Yeah. That's funny. Cool. Well, so. So you get back on and. Yes. And and we'll see where it takes us. Yes. And if we can talk to LeCun at some point in the future, we'll ask you about. Yes. Yes, indeed. Let's see here, near and dear to my heart always is, is kind of music and kind of what's happening right now in AI and music and Deezer, which is a pretty popular streaming platform, nowhere near as popular as something like Spotify, of course, but nonetheless in the Mix says that nearly half of all new daily uploads to the Deezer streaming platform are AI generated tracks 44% of songs uploaded.

00:52:27:25 - 00:52:50:09
Unknown
But they say on the flip side of that, only draw 1 to 3% of listening on the platform. So while they're getting a ton of this stuff uploaded, they're not getting a huge amount of people listening. Although I guess depending on how large their listenership is, 1 to 3% could still be a lot of people. That's that's a very inefficient use of their resources.

00:52:50:12 - 00:53:17:03
Unknown
Yeah, right. What do you. What do you mean? What do you mean? For Deezer to have to take in all those songs, store them. Oh, 100%. Absolutely. 44% of the input is is AI slop 1% of the output is AI slop. That's not a good formula. Not worth it. Yeah, not worth all of those resources. 85% of those streams that are identified flagged that are identified as AI generated become flagged as fraudulent.

00:53:17:03 - 00:53:36:19
Unknown
When that happens, they are demonetized. You know, that still leaves 15% that that I guess get through, or 15% that were good enough that they couldn't get detected or whatever. I wonder how that number changes in the next couple of years is these systems. It's like the economics of spam, right? If it costs you nothing to send out emails to everyone in the world, what the hell?

00:53:36:20 - 00:53:59:13
Unknown
You send it out, you get a point 1% response rate. What the hell it costs you? Nothing. It's fine. So on that end of it, if you're 1% of 15% and you get listened to it, you haven't been demonetized and it cost you nothing monetarily or culturally to make the song. What the hell? What the heck? Yeah, there's plenty of people out there for that opportunity, right?

00:53:59:14 - 00:54:21:18
Unknown
And playing that game specifically because of that. Because if you do get a song that does get monetized and it does well or whatever before it's detected, like, I don't know, I wonder what the math around that actually adds up to. What do you earn if you do that successfully? Do you actually make money because streaming music doesn't pay well unless you're streaming millions upon hundreds of millions of times.

00:54:21:18 - 00:54:50:23
Unknown
And I'm guessing these songs aren't streaming that. But if you do enough of them in aggregate, if you have 100,000 songs and you got it and it'll cost you nothing, then what the hell? Same with books, same on the same with. With just general content on the ad side. Yeah, yeah. I wonder what's happening at YouTube. I see all kinds of well, I see, I see halfway slop being that I care about the history of a lot of the stuff.

00:54:50:23 - 00:55:15:29
Unknown
I'll see a video about the history of, I don't know, Campbell's Soup in New Jersey. And I think, okay, that's interesting. It's something unusual I'm curious about and I'll turn and it's, it's I think it's a real voice. It may not be I don't know where the script came from. It's a few photos and it's this really boring exegesis about Campbell Soup that can go on for 40 minutes.

00:55:16:02 - 00:55:40:13
Unknown
And then on a few topics that I know about, it's completely wrong, whether that's human sloppiness or whether that's AI slop, I don't know, but it's filling a lot of YouTube as well. It's it's a curse. Yeah. And podcast, this was a question I had when I looked at Deezer. I actually just was like curious, like, okay, how does this how does this tie into podcasting?

00:55:40:13 - 00:56:14:27
Unknown
And then I came across this article on Pod News.Net April 15th. So very, very recent that basically says according to podcast Index report, is sharing that 44.6% of new shows that were monitored over the course of 24 hours were created by human humans, 45.7% likely created or generated by AI. So percentages wise, more content coming into these networks in 24 hour period generated by AI than are then are created by human humans.

00:56:14:29 - 00:56:48:12
Unknown
And it's not like we didn't already have an overdose of podcasts, frankly. And it's hard. It's hard to get your nose above water in the pool we're already swimming in. Yeah, yeah. I just I wonder like how many people are listening to these. I was having a conversation with someone the other day who was who was pretty, pretty pretty convinced that the AI podcast thing is a trend that is going to continue to be very competitive against human podcasters, and that there is actually appetite for for it.

00:56:48:14 - 00:57:12:26
Unknown
This person was not a fan like necessarily speaking because they were a fan of it, but, was talking about the trend kind of moving in that direction. And, you know, we talked last last week about the video that I came across that it took me like 7 or 8 minutes to figure out that, oh, wait a minute, that's an AI avatar, you know, that was created by I think and and whisper it was very convincing.

00:57:12:29 - 00:57:29:06
Unknown
And it just reminds me of like the on on the Twit network for so many years. It was always the running joke with AI, which is like, oh, when Leo Laporte, you know, when it's time for him to retire, there will just be an AI version of him that will continue. Ha ha. That's so funny. It'll probably happen.

00:57:29:06 - 00:57:53:25
Unknown
Who knows, maybe. I think that could definitely happen now. Really. And there would probably be an audience for it, you know? Yeah, he's he's playing with that. I mean he's using AI to make his run downs. And I still piss him off by putting my own stuff in there and trying to trying to break the rundown. And and he's played with having these things listen to him enough that they can they can generate him.

00:57:53:27 - 00:58:18:08
Unknown
Right, right. But it's not. And I'm going to sound like a journalist saying, who needs is a journalist. You want a citizen surgeon, you know, in the social media days, you know, who needs an AI? You want an AI surgeon, you may have one. Yeah. Yes. So I don't want to just seem like I'm a Luddite trying to hold on to my humanity in this world.

00:58:18:08 - 00:58:44:01
Unknown
But I do think that's where the value is, especially in podcasts. It's one matter if you want to learn how to do something. I want to learn how to replace my toilet mechanism. Can I imagine that AI can create in text or video a manual that I can understand? Yes, I can do I care if it's a human teaching me as long as it's correct?

00:58:44:02 - 00:59:08:20
Unknown
No, I don't write as long as I get the right information. That's not all podcasting is. Podcasting is largely at least the type of podcasting that we do. It's a it's a conversation. It's an exchange of ideas. It's things that it's unexpected moments that or thoughts or revelations or whatever that come up. Yeah, no AI is going to have fits about not getting Ask Gemini on top of the computer.

00:59:08:23 - 00:59:40:29
Unknown
Only here you get that. I mean, maybe, maybe it'll happen. Maybe. Maybe they'll get good enough that they can start throwing in things like that, you know, I don't know, but no, I agree. I feel like the, the human element is the is the premium. But I also recognize that like, I am old guard and yes, we are and, and I do my very best to, to try and not look at things through a specific like it's my way, my way is the right way sort of thing.

00:59:40:29 - 01:00:00:25
Unknown
I really try my best and try my hardest to look with an open mind to the trends and understand that other people come from different perspective or different generation or whatever, and they just look at things differently than maybe I was, you know, I grown up, I grew up with there's the right word. And so I'm very curious.

01:00:00:25 - 01:00:22:20
Unknown
It's not AI. Yes, totally. I just hallucinated there, didn't I? I blanked out for a second. Yeah. So, I mean, so if there is, you know, an appetite of the younger generation that is growing up around these tools from a different perspective than we are, where we're adapting to them, instead of this just being the way it is.

01:00:22:23 - 01:00:44:27
Unknown
Like, I kind of want to know and understand where that comes from. So I don't know. It's interesting times. Let's see what it all leads to real quick. If you are enjoying us talking about our oldness in the world of technology and you know you like the show because you identify with our perspective, you're probably our age as well.

01:00:45:01 - 01:01:00:21
Unknown
Hey, please leave us a review on Apple Podcasts. It's a huge help. It helps more people to see that we're doing what we're doing, and we can't thank you enough for that. So thank you. Going to take a quick break and then give you a few quickies before we get out of here. That's coming up in a moment.

01:01:00:23 - 01:01:32:12
Unknown
All right I haven't played around with it. But anthropic did announce Clod Design. This was launched by Anthropic Labs. It's powered by opus 4.7, their most up to date publicly available model. Let's say it is a visual workspace. So visual design workspace, it's you know, it's about generating design ideas, prototypes, decks, marketing assets. You could kind of design a user interface around this or ideas around it.

01:01:32:14 - 01:01:52:23
Unknown
Your mythos made website. Yes, yes. Your space website. Yeah. It's so hilarious. That's what they did with it. I'm curious to to play around with this, to kind of see, you know, I, I think what I run into with this is, is the question that I ran into when I first started playing around with Llms, which is what would I use it for?

01:01:52:23 - 01:02:16:03
Unknown
What on earth would I even use this for? I don't even know. But you know, the more you play around these things, the more that becomes apparent. So since I haven't used cloud code to make things, yeah, I haven't really played with cloud code. Yes. So let me ask about coworker then. Okay. Can you. So anthropic never went as of yet.

01:02:16:04 - 01:02:46:03
Unknown
Hasn't chased the the video dream right. OpenAI did until it got rid of that. Google has anthropic was emphasizing building things in code whether you're using code or coworker right. Yeah. When you use coworker, how sophisticated was the design output of it? Is that like visual aspect of it? I, in my use of it, the visual element of coworker.

01:02:46:03 - 01:03:07:19
Unknown
And mind you, there's probably other ways to use it that I unaware of, but you know, I've used it to create PDFs. I've used it to create a presentation. I actually used it in a PowerPoint presentation a couple of weeks ago. I've used it for, you know, creating spreadsheets and stuff like that. And it's been pretty, pretty useful, visually speaking.

01:03:07:21 - 01:03:27:10
Unknown
Like, actually, let me think of the PowerPoint. Yeah, I think of the thinking of the PowerPoint. I mean, yeah, it was visual, but it was pretty basic. Like if I needed to, to do like a diagram of of what a studio looked like and cameras and the kind of the layout of a studio, it's kind of like colored circles to, to indicate where the lights would be.

01:03:27:10 - 01:03:49:13
Unknown
And, you know, a colored square for where the camera is and maybe it's labeled and everything. It does the trick, but it's not visual. Was it a sketch? Right. Yeah. Whereas sometimes I've used, you know, Gemini in canvas and it's come up with some really visually rich output. And I'm sure it's possible to use Claude or Claude coworker in that way.

01:03:49:14 - 01:04:09:28
Unknown
I know that you can give it like parameters around like, well, these are my brand colors. And you know, this is a logo, include this logo on my PDF and that sort of stuff. It can do that. I haven't really dove much into that. So what struck me about about Claude Design is it seems like they're moving up the product design or the product creation stack.

01:04:10:01 - 01:04:40:16
Unknown
Yeah, totally. Totally. You could create the code, you could create the structure. You could do all that between coworker and code. And now you can make it look good with this. I think it's I think it's a necessary addition. And it's still strikes me. It's I think Claude remains or anthropic remains more the B2B company, even though it has a consumer use and the consumer use expanded immensely after the whole DoD thing.

01:04:40:19 - 01:05:08:22
Unknown
And that's that's important. It still is aimed at people who have to build things for a living. And so I think this was a necessary aspect that had to bring in esthetics. I think, yes, that that lines up with kind of how I've been looking at this too. It seems like anthropic with Claude is moving closer to it, being kind of more like an application company than just a provider of LM models, you know what I mean?

01:05:08:25 - 01:05:30:29
Unknown
They're creating a suite right now, and I'm starting to because I'm using coworker as much as I am. The last month or so. God, I've really I've really gotten into it, man. Like, I really enjoy it. It's it's captivated me in my in how I work in a way that the other models got me somewhere closer, but didn't quite get me there.

01:05:30:29 - 01:06:00:13
Unknown
And part of the reason is because the whole suite, as it's building up, is pretty comprehensive and does some pretty advanced things without requiring you to be a like a know it all in the in the world of how these things operate. So it's pretty great. And this is yeah, this is a, a missing piece, a missing component, but all of it together, it's kind of like when I look at where Claude is building up right now, it kind of reminds me of like, Adobe and its suite of tools.

01:06:00:15 - 01:06:27:17
Unknown
You know, Adobe has all these different, very top tier applications that do these different aspects of design. Let's say, you know, in creation and it seems like Claude, with their application and coworker and design and all this is moving in that direction. But for how you like run a business to your point, like B to B, it's like, here's all the different, you know, here's how you use this one model to run your business in all these different directions.

01:06:27:17 - 01:06:50:13
Unknown
And yeah, it's pretty powerful. Interesting. Yeah. And you know, I teaching I wanted to create I still might try to create classes in product and product development because that's where the the need is in this changing world. You need to create new products. And that's what that's what rules Silicon Valley. If you weren't in product, you were nobody.

01:06:50:15 - 01:07:15:04
Unknown
And it stretches over into areas like media and journalism and such. So I think this is I think having a fully formed product suite, product development suite is going to be a good thing for anthropic, and I think this is a step in that direction. Yep. Interesting stuff. I'll definitely I do have because I have a God right now I have I have the $100, the max level.

01:07:15:05 - 01:07:29:08
Unknown
That's how much I've been using it. Wow. It's the first time I've done that, Jeff, but I found myself in a point to where it made sense, and so I ponied up for it. And I'm really happy I did, because I continue to use it more and more. It's just interesting how I said it. So you had a few.

01:07:29:14 - 01:07:46:21
Unknown
I know you had a few projects where you had to use it more. Yeah. Did you intend to say all this for a month then? I'll, I'll that's that's kind of. Yes. Like part of it is an experiment. It's like okay. Because I've, because I've thought about it like who would who on earth would spend $100 on this?

01:07:46:21 - 01:08:15:29
Unknown
Unless it makes sense. And as I use it more and, and I started to kind of run into the limit, then I was like, okay, well, I either spend $10 piecemeal here and there to give myself a temporary bump. Right. Kind of sounds like drugs, actually. Yeah. To give myself temporary access to tide me over until my credits renew, or I just throw down the $100 for the month, and I don't have to worry about it.

01:08:15:29 - 01:08:38:11
Unknown
At least now. My current usage, I don't have to worry about it. So anyways, it's it's an interesting. I'm glad you're reading the dog food, which is great. Yeah, I'm far more than I am. I'm in. I'm enjoying it. It's it's enjoyable. It's also, though, very addictive because now that you have access to this thing, gosh, as I talk about it, in many ways, this is kind of like a drug to a certain degree.

01:08:38:11 - 01:08:59:29
Unknown
It's kind of crazy. Well, we're like, they're getting me I okay, so you're now predicting the moral panic stories that will come out and the Jonathan Heights of the world next attack AI from that end. Yeah. You're not addicted to drugs. You're addicted to AI. Yeah. Might as well face you're addicted to AI. That could be the title Deep Research.

01:08:59:29 - 01:09:27:02
Unknown
Max, this is Google's next deep research product, powered by Gemini 3.1 Pro. These are agents. So and I think the big news here is that deep research integrated with MCP support on top of the web. So basically hosted agents they can search the web. They can access your your data. Hard to say automated research pipelines, all that kind of stuff.

01:09:27:02 - 01:09:50:20
Unknown
You can do all of your charts and infographics and visualizations and tie that all in. Yeah, yeah. Just a minute ago, we were talking about about the stack needed for product development. Yeah. This is the stack needed for analysis. Yes. Yeah. There you go. And they're targeting finance, life sciences, other data heavy fields with something like this where you got open it.

01:09:50:22 - 01:10:13:25
Unknown
Yes. And then open AI has an update to their image generator. ChatGPT images 2.0 is out I remember OpenAI. Yeah, yeah. We used to talk about them all the time, didn't we? We used to talk about them all the time. All their new products. Yeah. I don't even think of open AI for image generation, to be honest, because nano banana is just so good.

01:10:13:27 - 01:10:32:08
Unknown
So I haven't gone back to open AI, but I'll have to try this out to see. They say you can produce up to eight consistent, high res two K images per prompt, so that's kind of interesting. Make me eight images with this character or something like that I'm guessing is what that means. I'd be interested in that better following instructions of course.

01:10:32:08 - 01:10:53:16
Unknown
Better at handling text of course. Probably gives you four fingers than a thumb, I'm guessing. So on each hand, it'll also get in the copyright kerfuffle. I can be sure that looks a little bit too much like my image on the web, but yeah, yeah, we'll find out. Yep. People seem pretty excited about that, I suppose. So there you go.

01:10:53:17 - 01:11:19:10
Unknown
Jeff Jarvis com go there. Preorder. Hot type available soon. Somewhere. Somewhere soon. But you can preorder there. Gutenberg parenthesis magazine. The web. We we've working on new stuff as we speak as well, including intelligence, AI and humanity. A new book series from Bloomsbury coming soon. Is this linked to this medium page from your Jeff Jarvis com? No, I need to do that.

01:11:19:10 - 01:11:32:29
Unknown
I need to ask my son. My son created my page and I want to screw it up. So I get him to ask him. Yes, you do have a medium link from your Jeff Jarvis com but that that's true. Oh and it's the top and it is there. Can get it. I haven't been anyone who wants to check it out.

01:11:32:29 - 01:11:49:02
Unknown
They can at least for now go to Jeff Jarvis. Com find the medium link on the page. And that'll take you right there. Cool stuff. Just as I remind you to talk about your consulting business, you remind me to talk about my book series. So there we go. Yes, I've got I've got it baked into the template, man, until it starts coming out.

01:11:49:03 - 01:12:15:23
Unknown
Well, I'm probably beyond because then that'll be the thing to talk about. Pod two. Speaking of which, if you have have plans for a podcast, if you think you might want to do a podcast, I'm flirting with the idea of doing a cohort thing. I think that would be a ton of fun to like, get a get some people together and do like a live like four session cohort on like you've got an idea for a podcast, you want to launch it.

01:12:15:23 - 01:12:36:28
Unknown
And here, you know, be with me and I'll kind of talk you through all the things that I would do if I was launching a podcast. I'm, I'm devising this, this plan. And so if that's something that you're interested in, you know, feedback at AI Insight Show, let me know, because I'm going to need people who are interested in order to actually do a cohort.

01:12:36:28 - 01:12:52:17
Unknown
And in order for me to, like, build it all out and everything, I just kind of need to know that there's some interest. What do you need? So. Right. Yeah. And but I'm super jazzed at the idea. I think it would be a ton of fun. It's kind of like podcasting, doing a cohort. It's like, here, I'll podcast to you about podcasting and you're there.

01:12:52:17 - 01:13:16:20
Unknown
You can ask me questions, you know, something like that. But anyways, go to Pod Up. You can see all of the the kind of what I'm doing right now from a consulting perspective for people who want to start a podcast or have a podcast and want to make it better or achieve a goal or whatever, pod Tune Up and sideshow is our website for all things about this show audio, video, everything you need there.

01:13:16:20 - 01:13:42:16
Unknown
And then of course our Patreon Patreon.com Inside show. We've got some amazing executive producers each and every week it is Doctor Do Jeffrey. Radio Asheville 103.7 Dante, Saint James, Bono, Derek, Jason I for Jason Brady, Anthony Downs, Mark starker, Karsten Sumac, and maybe you if you decide to support us you all. I love that that list is so long and continues to be so long.

01:13:42:17 - 01:14:03:22
Unknown
So I can I can imagine you wake up saying in your sleep, yeah, I know at this point you say Carson and you know it's time to wake up. Oh, it's AI inside day. That's that's my alarm. Yes. Is the theme song inside which you're about to hear right now. Thank you, everybody, for watching and for listening. We will see you next time on AI inside.

01:14:03:23 - 01:14:04:14
Unknown
Take care. Bye.