China May Lock Down Its Best AI Models

00:00:00:00 - 00:00:23:23
Unknown
Coming up next on AI inside, Jeff Jarvis and I get into Beijing Wang weather to wall off its best AI models from the rest of the world. Kind of like how the US has been doing it to China and Beijing. OpenAI offering the US government a 5% stake to kind of smoothen things over in Washington right at the time that we're now suddenly getting their next model.

00:00:23:25 - 00:00:49:22
Unknown
I think tomorrow at meta, turning your public Instagram photos into fuel for its own new AI image generator, unless you happen to opt out, and so much more. Coming up on this episode of the AI Inside Podcast.

00:00:49:25 - 00:01:16:25
Unknown
Hello internet! Welcome to another episode of AI Inside the Show, where we take a look at the AI that is layered throughout the world to technology. One of your hosts right here, Jason Howell, the other host right there, Jeff Jarvis. Good to see you, Jeff. Good to see you as ever, Ross. Yes, indeed. I feel like this this is yet another one of those weeks where, like, everybody's releasing big news right before showtime and all sorts of things are swirling.

00:01:16:27 - 00:01:38:26
Unknown
Anthropic continues to be in the in the news. China continues to be in the news big time. OpenAI has some big stuff going on. I mean, this is just yeah, this is just the normal, right? This is just normal. Yep. I'm not surprised anymore. Every single week. No shortage. Yeah. Well, it's good to see you. You have a big let's see here.

00:01:38:26 - 00:02:02:09
Unknown
I'm going to hold off on that until later actually. Let's start with China. China might be I think this is really interesting because we've had the last couple of years we've had the US kind of playing this game. I don't know if they'd call it a game kind of. Sometimes it kind of feels like a game to where they're saying, okay, we're doing all these great things over here in the United States of America, the great US of A in AI.

00:02:02:13 - 00:02:32:25
Unknown
But we don't want China to get get it. And so they've been kind of closing things off and everything over the last couple of years. Now it kind of seems like China is doing that right. And I guess it's not surprising, but it seems like in the in the reverse with the open models and, you know, they're open wait models also some of their closed models, it seems like China is is starting to kind of tighten control over those models to make sure that I guess the United States and other countries outside of China don't get access to it.

00:02:33:02 - 00:02:58:15
Unknown
Yeah, this actually surprised me because I was talking about cutting off access to their own models. Just last week. I was saying that China can undercut the entire American AI industry by releasing its models for free and open wait, and that that's a tremendous strategic challenge to the US. And so I thought that's what was going to happen.

00:02:58:15 - 00:03:17:16
Unknown
And now China is looking to do the opposite of saying, well, ours are really valuable. And so we were if you don't let us see your chips and your models, we are not going to let you see ours. Yeah, yeah, yeah. Our feelings were hurt and it becomes in the end they fight both ways. And I don't think that's really that's terribly helpful in the world.

00:03:17:18 - 00:03:40:12
Unknown
Yeah. And we put a Jensen Walker on the show about this that, that he says that cutting off. We'll get to chips in a second. Cutting off access to chips as well as the software is hurting America's competitive advantage. It's hurting Nvidia particularly, but it's also hurting our companies in general. So I'm going to I want to see more openness in AI.

00:03:40:13 - 00:04:02:09
Unknown
I think this this this nature of closing down both ends is harmful. Harmful. Yeah. It harmful to advancement. I think it's a bad thing. Feels like we started it right back. Back to the playground. Your fault. We started it. Yeah. Well, totally. And when we started it, like, I think we've been talking on this show, this is going to, you know, kind of backfire.

00:04:02:12 - 00:04:39:07
Unknown
It's certainly not going to accomplish what I, I believe that the Trump administration was kind of thinking, you know, they, they wanted to achieve out of this was, you know, make ourselves stronger by making sure no one else gets access to these things. And all you do is you incentivize very well, you know, very highly capable countries like China and other countries to say, okay, well, if we can't have access to that, then we're going to do what we can to compete, especially on a technology that's as important as AI is right now to how everything is moving.

00:04:39:07 - 00:05:02:24
Unknown
They're going to do what they need to do to make themselves self-sufficient. Period. You've just made that even more likely by our hostility to them. Exactly, exactly. Yeah. And that hostility. And let's not forget, it's not just the software. It's not just the chips. It's also, very importantly, the human talent. We were training a huge portion of Chinese technological talent.

00:05:02:25 - 00:05:25:09
Unknown
Not that China can't train them terribly well on their own. Same as India. There's there's great technology training in universities in both places. But a lot were coming here. They were getting trained in America. A lot of them were getting hired by technology companies in the US. We were benefiting from that. And now with the immigration mishaps, we're losing a tremendous amount of human talent as well.

00:05:25:10 - 00:05:47:27
Unknown
Yeah, we screwed ourselves three ways around, big time. It's a closed mindedness. Daniel. Hello, Daniel. Love having you in the live, live audience every time. And of course, Daniel helps us behind the scenes. Says us in the UK will end up with no AI because, I mean, that's an interesting point. Like, I guess I don't think about that enough.

00:05:47:28 - 00:06:12:18
Unknown
What is going on in the UK when it comes to frontier AI? Is it is it all pretty? Like, I think there's much going on in the UK and is that surprising? Is it surprising that that's the case? Well where's DeepMind? It's in London, right next to the Guardian. No that's true. Yeah. Yeah. Right. So a lot of second I was thinking deep seek and I was like, well that's a dumb dumb DeepMind mind, right?

00:06:12:20 - 00:06:34:28
Unknown
It's all deep. So they're in London, so there's a lot of action happening there. Yeah, but it's going to American companies. Obviously. We talked about Mistral and France. Yes. That's right. We talked last week about Austria talking about trying to get anthropic over in one form or another. I think Europe is very consciousness right now with America coming themselves off and all kinds of ways.

00:06:35:00 - 00:07:01:10
Unknown
So I think we'll see more development in, in Europe. But yeah. Daniel says that Mr.. All is the closest as a model, and I would like to say the annual that the UK is still part of Europe, but that's a whole other story, ain't it. Right. Not not to rub that in then. So, so I think the long and the short of this is we're going to see many, many more competitors around the world at the fundamental level.

00:07:01:10 - 00:07:29:00
Unknown
The problem is the expense is so high. And we do have a hell of a lot of capital here that other countries don't necessarily have. Yeah, a lot, a lot of resources. Well, all of this is related to a Reuters. These Reuters was reporting on this story that over the past few months Beijing has been holding meetings, you know, with ByteDance, Alibaba, Ministry of Commerce, Z, which I'm not too familiar with the AI.

00:07:29:01 - 00:08:08:24
Unknown
Maybe I am, but I don't realize that I yeah, I forget what other models was was a hot open release lately. I forget I can't remember the names. I can't get human names straight. I know going straight it's hard to keep, keep it all straight. But yeah, doing the restricting of overseas access to all of their most advanced models and they are actually they were actually talking about making any leak or any theft, a proprietary AI technology, an offense under China's national security law, also restricting who's allowed to fund domestic AI startups, which that has to be related to the was it manus, the meta manus?

00:08:08:26 - 00:08:45:21
Unknown
The Chinese government cut off that purchase, right? Yeah, right. Cut off that government. Yeah. That's another 1 or 2. You know investment opportunities also gets cut off right. Acquisition is cut off. So meanwhile that's all on the software side on the on the hardware side. Deep Sea says they're going to develop their own AI chip. Which which exactly what what Jensen Wong predicted if you cut us off, if you if you cut them off of the supply, not only do we lose the business to an American company, but again, to quote Jensen Wong, he loses one major part of a a standard platform in that.

00:08:45:21 - 00:09:15:22
Unknown
And you sign on with with Nvidia chips, you also signed on with Cuda and Cuda as a as a platform. It's obviously a desirable to have that. It wants to be a standard around the world so that people can use it. That also benefits America. And then you've got also you have no idea releasing Nemo, Tron, open wait, open source models.

00:09:15:25 - 00:09:50:02
Unknown
These are all efforts to be to have hegemony around the whole world. And China just puts up a wall and you can't get in there and that's that's foolish. Yeah, yeah. It's no good. And I mean, deep sea is also not alone. It seems like this is another trend that's happening right now, is that there's a lot of, a lot of the big companies are trying doing their best to create and develop their own ships, which seems to be a lessening of reliance on companies like Nvidia, which, by the way, what was that story that I saw today that Nvidia is down?

00:09:50:02 - 00:10:13:14
Unknown
Its value is down to pre boom level. Lost 1,000,001 trillion in market cap all the way back down to through 4.86 trillion. Yeah I'm not crying I'm not doing okay. Yeah. But in video has definitely been the company that over the last couple of years has been very easy to be like, well who's the winner in the room.

00:10:13:14 - 00:10:41:01
Unknown
It's in video because everybody was reliant. Now we're getting to the point to where everybody is working on their own chips so that they don't have to be reliant. Well, I think the supply chain problems absolutely key role there is if you couldn't get the chips and if you were because of it, and if you're in the case of of Apple announcing the price increases, if you have to have an impact on your product line, then yeah, you're motivated to to try to create your own structure.

00:10:41:03 - 00:11:07:18
Unknown
But to say that to make the headline, they've lost a trillion in market cap. Well, yeah, right now the whole market for AI stuff is down. And and they started so high again there were 4.86 trillion. They're up 4% today as we speak. Okay. So they're doing okay. He can afford a new jacket. It's all right. No need jacket.

00:11:07:20 - 00:11:32:22
Unknown
That's a good question. What what is the what is the jacket that he's buying and how much does it cost? It's probably that's probably a pretty pricey jacket. And he's probably not like I think for Cassie. Well, you can get a replica for between $92 and $250, depending on the material. But the authentic off the rack luxury pieces from designers like Tom Ford retailed for between 5 and $10,000.

00:11:32:25 - 00:11:57:20
Unknown
The actual jacket, worn and autographed by Jensen himself, is estimated to auction of 40,000 $60,000. Oh boy. It's like. It's like going into a museum, like a music museum and seeing Prince's guitar. But it's it's the jacket, Jensen. You know, I get one, but it's still too hot right now to wear where. Yeah, like that you'd get. That's the only reason I would pick one up.

00:11:57:20 - 00:12:13:20
Unknown
But, you know, it's too hot outside so I'm not going to. So I think we're both too tall with the sleeves. Would look funny on us. That's true, that's true. I don't yeah I don't I don't weird envision Jensen Wong to be a very tall person. But maybe he is I don't know, I've never met him in person.

00:12:13:22 - 00:12:49:02
Unknown
Alibaba also banning employees from using anthropic clod code started actually starting the 10th. This is Friday. They added it to a list of high risk software. Told staff, you got to switch to coder coder spot with a Q, by the way, is Alibaba's coding tool. And why do they think it's a high risk? Well, apparently a user on Reddit reverse engineered cloud code found hidden logic that checked whether your system time zone matched Shanghai and looked for Chinese AI lab domains in the setup.

00:12:49:03 - 00:13:10:22
Unknown
And there's an anthropic engineer named Tariq Skipper. I hope I said that right. Said on X that it was an experiment for March to stop account abuse and to stop distillation, which they were very concerned about. They said they'd been meaning to pull it and actually did pull it from the code on July 1st, but it doesn't sound like Alibaba is buying it.

00:13:10:24 - 00:13:32:10
Unknown
I also think it's a way for Alibaba to say you can accuse us of distillation and theft. Yeah. So we we forbade use of it. We don't need your stake in the software. We're so smart. Yeah. Yeah I think it's a bit of a branding slap against anthropic too. Yeah. Interesting. Interesting. So, so there's a great wall going up around Chinese AI all around.

00:13:32:13 - 00:13:56:24
Unknown
I think it's unfortunate. Yeah. Around. Yeah. Frontier AI possibly all around because I think this is the sort of thing that just kind of reinforces itself everywhere. Right. Oh, you're going to do that. Okay. Well I'm going to do it stronger or I'm going to do it. Protectionism is never healthy. You know, I see it obviously in the news business with AI is going to cut you all off whether you're cutting yourself off in the process.

00:13:56:25 - 00:14:14:12
Unknown
Right? Yeah. That's true. Right. Yeah. Exactly. That's protectionist behavior. And it's what does it take to undo that. What does it take to reverse it then? Good question. I think just coming to your senses, realizing what you're losing at some point. But but then it takes an industry to come to its senses, not just, you know, one.

00:14:14:14 - 00:14:31:19
Unknown
Well, I think in the case of news, what I think will happen is that I've been trying to convince people to do this is that smaller players will say, fine, you don't want to be on AI, we're on AI, we're going to get the traffic, we're going to get the branding, we're going to get the recognition. And I'm trying to encourage smaller players to do.

00:14:31:19 - 00:14:42:20
Unknown
That's why I argue for this, this notion of an API for news. In the case of of AI itself.

00:14:42:22 - 00:15:02:22
Unknown
You know, I think you just go to the smart people, you go to people like John Lennon, who we also quote often here. I do friend of the show front of the show where he talks about the need to be open, and he pushed for llama. And I think that cutting yourself off from an open market is just a dumb thing.

00:15:02:25 - 00:15:21:20
Unknown
So the two people I respect most in the AI world are young and Jensen Wang, and they're both about open and openness. Now again, open ish, as we say on the show open. They're not fully open in every case. It's not I'm not trying to say that it's all a gift for free. There's still companies, there's still businesses, but their ethos is more open than the rest of the world.

00:15:21:20 - 00:15:46:24
Unknown
And then then we get to the whole anthropic OpenAI problem with them acting as if they're so dangerous that they're inviting the band that they got and closing it off. And that's dangerous, whole different direction. So yeah. Fascinating. Well, that's the the kind of interesting things happening with China and the kind of reversal happening there. But speaking of you mentioned anthropic.

00:15:46:25 - 00:16:20:00
Unknown
This is anthropic news. That makes me happy. Yes. Makes me happy because I am, as I've mentioned many times now on the show, a big user. God, so much of what I do centers and revolves around cloud core work. It's just become it's become a major instrument, the major instrument in my business right now. Crazily not necessarily the content side, a little bit on the content side, but definitely on how I do some of my other businesses and how I structure, like how I run a business because I've never run a business before the last two and a half years.

00:16:20:00 - 00:16:41:12
Unknown
And so it's kind of nice to have this system that I'm working with and kind of building a kind of a scaffolding around, which is largely how I've been using Cloud Coworker and and it turns out I'm not alone. So first of all, Cloud Coworker is now brought to the web and to mobile starting this week. And that's important.

00:16:41:15 - 00:17:04:12
Unknown
Like I see that as important because coworker has up until now been Apple only on like let's say like I've got it on my Mac OS, right? I've got two different machines, I've got a mac studio and I've got a laptop, a mac MacBook Pro, I've got Claud app running on both devices. But coworker, anything that I do inside a coworker has been specific to the machine.

00:17:04:12 - 00:17:31:04
Unknown
And so there has been this kind of like inability for each of my machines to tap into the same thing and tell I did what is called creating a coworker OS, which is essentially the way I built it is I created this like folder structure and this, I don't know, this architecture inside of iCloud, which shares to Apple's cloud, which means I have access to that structure on both machines.

00:17:31:04 - 00:17:55:00
Unknown
And it's updated as I work on either machine. And so that does the trick. But now anthropic is opening up Cloud Coworker so that it sinks so that that data syncs across its cloud, I suppose, so that it works on the web. It works on mobile. And I don't know, maybe it'll make I mean, I'm still jury out at this point because I'm just starting to use it in that context.

00:17:55:00 - 00:18:19:07
Unknown
But hopefully it makes things just a little less complicated, a little less like I've got to do some work around to make it work, and it just kind of does it. So. So yeah, this makes me happy to. Because I've been arguing all along that as long as you had to go into terminal and install things to use these tools, it drastically limits the market for them.

00:18:19:09 - 00:18:39:25
Unknown
Yeah. And and that's why I think Google's move on on this stuff by staying in the web is really important. And these guys are going to lose out. So moving moving things to the web is only smart. They're moving to the cloud and thus the web. And I think that's only smart because it opens up the market. Now of course I go to Cloud Coworker.

00:18:39:26 - 00:18:46:26
Unknown
And by the way because I'm on a Chromebook, I.

00:18:46:28 - 00:19:09:08
Unknown
Not I have a mac also, but I don't work on the Mac all the time. I work on the Chromebook, so I couldn't use clawed to its maximum, and that's why I wasn't paying for it because I couldn't use it to its maximum. Right? So I'm happy. So I come to cloud product work and it says download for Chrome OS, which is a little odd because you don't really do that in Chrome.

00:19:09:08 - 00:19:31:22
Unknown
But anyway, so I click on that. What does it say? Yes. Surprise, surprise folks. It says come on here. I had a second ago. Your administrator has not given you access to this item. Workspace? No. So now you are your own administrator. So does that mean that whenever you do this, I have to go in and find out where this is all the time?

00:19:31:25 - 00:19:50:22
Unknown
Yeah, yeah. That's such a pain in the butt. It is awful pain when I, when I, when I conceded that I was wrong about getting it into my browser, I finally I was able. But it was so complex, I had to ask Gemini how to use how to get Gemini in my browser, because it wasn't even I had to go through all these steps, so I'll try to do the same here.

00:19:50:25 - 00:20:12:07
Unknown
But this is this is going to open up the market tremendously, and I think a lot more people will discover what they can do with this in really neat ways. And so I think this is terribly smart by Claude. By by anthropic. Yeah. And I just kind of felt like it was only a matter of time, like it was going to happen eventually.

00:20:12:07 - 00:20:34:09
Unknown
It was just a matter of if not when. So I'm happy that they've done that because this was definitely something that I've been waiting for. They also shared some stats around it, and they said that when they looked at 1.2 million sessions, more than 90% of it was not actually software development, which so much of this was built with software development in mind.

00:20:34:12 - 00:21:07:14
Unknown
The biggest category was business process. Business operations at about a third of users were using it for that, which is that's how I use it. That's exactly how I use it for the most part. So that was kind of cool to see behind that. Content creation copywriting 16% actual software development came in at under 9%. Although, you know, when I think of like anthropic tools, I don't think of coworkers specifically as a software development tool, even though you can do some of that, like I think of cloud code as that not right plot coworker, but it's still kind of related.

00:21:07:14 - 00:21:24:01
Unknown
But coworker is like a friendlier version of code for everybody. You know what I mean? Right, right. Yeah. And there's no reason that code couldn't. This is the other thing that I've said about the stuff. If you have to go to a terminal to use it, that's a problem. If you have to install it somewhere, that's a problem.

00:21:24:03 - 00:21:44:06
Unknown
Being able to make a program and have it just run is going to open the market up tremendously. People don't have to think about all that. And I go back to in the early days of the web, I taught my students, I mentioned this in the show how to download and install a web server. Right. We thought that was going to be important to do.

00:21:44:06 - 00:21:58:26
Unknown
Then it became ludicrous to teach them that it was no reason in the world to teach them that, right? Yeah, right. And we taught them HTML beginning, and there was no reason to teach them that anymore. And I think the same will come with these AI tools that you want to write. You won't think of it as a program.

00:21:58:28 - 00:22:23:10
Unknown
Yeah. It's a it's a fancy command. Yes. It's just an expansion. Yeah, that's absolutely the case. And when you're using something like coworker, you run into that or I run into that where I'm kind of venturing into the world of create a dynamic web page or whatever that kind of acts like an app or whatever. But I don't think of it as, as an app.

00:22:23:13 - 00:22:52:16
Unknown
Maybe ten years ago I might have, but. Right. Exactly. Yeah. Yeah, definitely. Also, just real quick, Google launched an update to its Spark Cloud agent that runs MCP or that now, you know, runs in the cloud, which is kind of what we got going on with Cloud Coworker now. And now it supports their new their new kind of announcement is that it now supports MCP, which I felt like is a way to bring it closer to what Cloud Coworker can do.

00:22:52:16 - 00:23:12:21
Unknown
So Spark and coworker kind of inching closer to each other and capability. Although I haven't had as much success with spark in doing some of the things that I use coworkers for, I really like that I can do this, like intense structure that I've built around all of my work inside of coworkers. It's been really useful to me that I don't get as much out of spark.

00:23:12:21 - 00:23:35:19
Unknown
Spark sometimes feels like one off things or like recurring things, but there is no like connective tissue between all of those tasks together, which I can do inside of coworkers. So but anyways, they're kind of related. They're kind of like moving together. I wouldn't be surprised if spark, you know, it also has a desktop app. I wouldn't be surprised if that's what Google is going for with spark is its own coworker, you know.

00:23:35:21 - 00:23:54:02
Unknown
Yeah. I don't know what it takes for it to officially, you know, be that. But well, you get to use it because you're you're in the secret club, the only available now there's a handshake in everything. Yeah. Oh yeah, yeah, yeah, yeah. A phrase that we could speak before the door opens so you don't get shot at by OpenAI.

00:23:54:03 - 00:24:18:13
Unknown
People in the jungle. You invited me into the early access for ultra for for Gemini, basically. So which I can talk about that aspect, but there are certain things that I end up getting access to that I can't talk about until until the embargo is lifted and spark was one of them. I can't wait to play with spark, but it's it's only for ultra right now, so I can't play with it.

00:24:18:14 - 00:24:43:03
Unknown
Right. But I wonder I wonder when they'll open that up. I have to imagine at some point they make it somewhat accessible or do something there, but yep. Yeah. And then real quick related of course, is, you know, the ongoing fable of fable, which is basically at this point on the 30th, anthropic basically said, all right, we're bringing Fable back, fable five for everyone on July 1st.

00:24:43:10 - 00:25:05:12
Unknown
This this was after the US government had lifted its export controls over that just yesterday, anthropic announced that it's extending it through Sunday, I believe. I don't know if I have that story here, but yeah, they're extending fable use through this Sunday, July 12th. So you know check it out. You play with fable much. Oh yeah I have played with.

00:25:05:15 - 00:25:35:20
Unknown
So what does it do better than any of the others? Yeah, yeah. I'm happy you asked that question, because I feel like I have at least a solid understanding of solid enough understanding of what I end up using it for. Fable is really good at creating structure around things. So if I'm building a framework or let's say in my consulting business, I'm like, I want to, you know, I've been consulting with a company for three months, and this is not a scenario that I'm making up.

00:25:35:21 - 00:25:58:14
Unknown
Like, I have been consulting with a company for the last three months. It's my first real legit consulting job. Like I've never done this before. And, you know, thankfully everything's going well, so I'm kind of making it up as I go along. And I've been working with coworkers along that path for over three months to kind of figure out how I structure it and, you know, kind of figuring things out as I go.

00:25:58:15 - 00:26:26:06
Unknown
Fable was really useful for me to say, take a look at my entire history of my work over the last three months with you in this consulting business, and I want you to create a framework around how I do that. Again, how do I do all of these things in a really organized, controlled manner? And I, you know, challenge it to, you know, I work with it to create skills that allow me to replicate pieces of that parts of my of those parts of my business and fables.

00:26:26:06 - 00:26:56:12
Unknown
Really good at that. If you're using fable for like writing a blog post, you're using it incorrectly in my mind because you're paying an insane amount of tokens for results that you're probably used to seeing. But fables really good at, like taking an hour and tackling a task, you know, so that framework building that took, you know, more than an hour for it to do that because it had to analyze all these pieces, had to kind of organize like, oh, you know, along the way, Jason requested this, but then that didn't work out.

00:26:56:12 - 00:27:14:06
Unknown
And he had to course and do this and whatever. It's really good at holding all of that and constructing around it and kind of building a general sense of and offering a plan and saying, after all of this review, here's what I think you should do. And when you read through it, when I read through it, I was like, well, that's spot on.

00:27:14:06 - 00:27:34:23
Unknown
I could totally run my business like that. So, you know, we get to do the point where we need a Consumer reports of AI, where there's there's kind of a lab that says, here's, here's a task because because right now the benchmarks are all things we don't understand. The ridiculous. They're high. Yeah. They're they're totally they're they're absolutely irrelevant to life.

00:27:34:25 - 00:28:12:00
Unknown
Right. I don't understand them. Not at all. Not at all. Whereas I think we had a consumer focus. This is actually don't steal those folks. I think we had a consumer focused consumer lab for AI and had certain standard tests which could change. Right. We're going to test washing machines. Okay. We made this. If you look at what they do at CR, you, they take a piece of cloth and they stain it with a standard mixture of things so they can work that against the machines.

00:28:12:00 - 00:28:20:04
Unknown
And when when polyester came along they had to change how they did it. Right. It was not a was not a set thing.

00:28:20:07 - 00:28:46:10
Unknown
But I think that there's a chance to it goes back to the discussion about the web. I think that the cognizant are the ones who are using AI now or, or people are using it in very small ways, writing a blog post. I think instead, there's really a way to bring it into the consciousness of the consumer by having that kind of relevant example.

00:28:46:10 - 00:28:53:15
Unknown
So is there something about.

00:28:53:18 - 00:29:16:03
Unknown
If you took the challenge you just gave to to to thank you fable, if you had time and you don't, it would have been interesting to take the exact same challenge and give it to 4 or 5 models. Yeah, right. See how you got to judge the outcome and the outcome is going to be it's not going to be a easy to judge that because they're going to be qualitative.

00:29:16:03 - 00:29:40:04
Unknown
They're going to be different. But I think there's a there's a there's a way to do that. Yeah. Yeah I mean, you know the other side of this also that comes up in working with fable is I do wonder sometimes if it's a little bit of confirmation bias. It's like we know that it's a good model compared to what, you know, anthropic says or compared to what other people say, that sort of thing.

00:29:40:07 - 00:29:59:16
Unknown
Am I looking for it to be a good model by using it? You know, we know that it's incredibly pricey. We know that it takes a long time to tackle a task like if I tell it, use ten subjects to tackle this task. It's going to do that. It's still going to take its time and it's going to be ridiculously expensive, probably on the other side of it.

00:29:59:16 - 00:30:19:20
Unknown
And so does that whole process paint my perception of what good is on the other side? Yeah, because you know what I mean? Because it could be in my mind. I could just be, you know, and you're right, I could have used opus and come up with 90% of that, possibly. You know, it's truly a judgment call at the end to say whether or not it's good enough.

00:30:19:20 - 00:30:36:19
Unknown
And every case is different. The data you gave it, the cases you gave it are different. So it's what I'm asking for is is not easy because there isn't a simple standard. This blouse is cleaner than this blouse. The special meter to find that out, right? It's still a blouse and it had ketchup sauce on it, you know, or whatever.

00:30:36:20 - 00:31:01:04
Unknown
Yeah. Yeah. Yeah. Interesting. Well that's anthropic real quick. Want to throw a thank you to those of you who support us on Patreon, Patreon.com AI Inside Show, and a huge thank you to Gary Hunt who is our newest patron. Gary, it's good to have you on board. Thank you for coming through the board through the door Patreon.com inside show.

00:31:01:04 - 00:31:19:05
Unknown
Go there, support us. You get lots of perks, including a daily podcast that only patrons get except for the Friday episode. Everyone gets the Friday episode of AI Inside Daily. Everyone else who's a patron gets AI inside daily five days a week. In fact, I did it right before the show. So see, I don't even take the day off when we've got the long.

00:31:19:07 - 00:31:39:09
Unknown
What did you talk about on that show? Just to tempt people. Do you remember? Yeah. Well, we thought let's see here some of the things that we're talking about here, actually some of the things that we haven't talked about yet, definitely about open AI's GPT 5.6, a little bit about the European Systemic Risk Board and a joint warning that they issued.

00:31:39:09 - 00:31:53:12
Unknown
I can't remember if that made it into today. This is the challenge. Yeah, yeah, this is the challenge that there's so much only there folks. And so sometimes I'm like, well, what makes it to the main show and what makes it to the daily? And you know, the daily is always going to have stuff that we don't talk about.

00:31:53:12 - 00:32:14:22
Unknown
We will talk a little bit about Meta's kind of glasses stuff that's coming up in the second block. Yeah. Space AI which is a mouthful. Space AI the new the business. I think we've even got that in this show. So some of the, some of the stuff is overlap and but there's a lot more stories that end up in the daily that don't end up in this show, because we can only pick a certain number for this show.

00:32:14:25 - 00:32:30:13
Unknown
So. So anyway, that's at Patreon.com AI Inside Show, going to take a break, come back and talk about some of these stories, including the OpenAI story that I was just just talking about that's coming up here in a moment.

00:32:30:15 - 00:32:34:27
Unknown
All right, open AI.

00:32:34:29 - 00:33:10:29
Unknown
Trying to buy its way out of a headache in Washington, DC, maybe, and or bribe. I'm not sure what the right verb is. Yeah, bribe. I said that as if I was combining the words. But no, the word is bribe. But babe. No. Yeah, okay, I don't know. OpenAI reportedly offered the US government a 5% stake equity stake, and I guess that would be worth somewhere in the realm of $42 billion if you took, I guess, valuation $852 billion valuation from recent.

00:33:11:02 - 00:33:35:26
Unknown
And their reasoning was to address political blowback. And, of course, you know, get in good with the administration, I'm sure, because of everything that that we've all witnessed with the way the anthropic thing has gone down. Right? Wasn't it wasn't OpenAI just recently saying like, yeah, we'll do it, but we don't want to do it, but then then they're offering this, I don't understand.

00:33:35:28 - 00:33:46:26
Unknown
There were rumors that they were offering up to 10%. Wow. So now it's an offer of 5%. And I think that.

00:33:46:28 - 00:34:02:22
Unknown
It's just an effort to force. It's interesting to watch the different strategies here. Right. Anthropic gets kicked out of the Pentagon. They get their software kicked out of the world. They go through all that. But so far they've managed to kind of hold their head up. Whereas after the Pentagon opening I said, okay, okay, we'll do it for you.

00:34:02:23 - 00:34:26:29
Unknown
We'll let you have ours. And they come along and say, okay, we'll give you equity. It smells a little desperate to me. Yeah. And it apparently the mechanism of all of this that they're pitching would be formed after a sovereign wealth fund idea that was modeled on the Alaska Permanent Fund, where the government would take 5% of each big USAID lab.

00:34:26:29 - 00:34:58:19
Unknown
So that would be OpenAI. That would be anthropic Google meta two, possibly. And then dividends would flow back to the citizens of the United States, I guess. Why this industry of any industry. Yeah, it doesn't make a lot of sense. So so last week on intelligent machines, or two weeks ago when I told them we had a professor from from Fordham on law school, Olivier Sylvain, who.

00:34:58:21 - 00:35:22:17
Unknown
Wrote an interesting piece about this, but but oddly, I got him to talk about his doctoral dissertation at Columbia, which was about the regulation of radio, which I'm fascinated with right now, and so on. After it was over, I ended up thinking about it more and more and more, and I realized there was a parallel here broadcast in America could have turned out very, very differently.

00:35:22:20 - 00:35:46:07
Unknown
But the government was frightened to death that radio would end up in the control of Furnas, Furnas, Furnas. So they foreigners. So they was like, what is a ferner I've never heard of a furniture before. Okay. Gotcha. So that's my way to make fun of it. So.

00:35:46:09 - 00:36:11:09
Unknown
GE was going to sell a transmitter thing to Marconi who was English slash Italian. And the government intervened and said please don't. And then the US Navy intervened and forced together the creation of CA as a patent trust that about 5 or 6 companies that had the patents that affected broadcast, they all came together in RSA and they all got a piece of it.

00:36:11:11 - 00:36:35:03
Unknown
And that's why American broadcast looks the way it does, because it was government intervention. In fact, military intervention. And part of their fear was also about immigration. So what's interesting is I extrapolated from that from Sylvain Dissertation to think about what about today. And we seem to be almost heading the same way where at some point, whether it's owning 5% or 100%.

00:36:35:07 - 00:37:14:13
Unknown
Yeah. Or owning equity or instead of owning patents. I see something similar going on here where AI is really powerful and we've got all this, this xenophobia about other countries. See the first part of the show and there's fear of immigrants these days. And so this becomes, to me, a really interesting but really potentially dangerous precedent here, where we could find the federalization of good chunks of the AI industry, something which one would not have thought of would come from a Republican regime.

00:37:14:15 - 00:37:42:08
Unknown
Yeah. Which leads to I'll skip a story, if I may, the Wall Street Journal. Hilarious. The editorial board using the favorite word du jour, Socialist. The socialist temptation of Sam Altman opening I considers a Faustian bargain with the US government to hand over equity. And of course, the Journal thinks this is terrible because it's a journal. They're old style Republicans and they're saying, no, Sam, don't do it.

00:37:42:08 - 00:38:04:15
Unknown
Sam. Sam, and I find myself in this weird position of agreeing with the Wall Street Journal, where I think that it's dangerous. And it's not just in the AI industry, obviously. It's also happened in the steel and autos and other industries where governments got involved. But this does not make me comfortable. What do you think about the deal?

00:38:04:17 - 00:38:32:02
Unknown
Well, no, it doesn't it doesn't make me comfortable as well, probably for a lot less well-informed reasons. There's just something weird to me about this kind of sudden moment where all of the, the companies, whether they want to or not, suddenly feel like they have to. I don't know, it just it just kind of feels Bribie to me, it feels scummy and actually timed.

00:38:32:02 - 00:38:58:25
Unknown
With this we have ozone nightmare. Thank you for the super thanks and I'm sorry I'm covering up your video. Jeff says it's not a bribe fellas. It's a strategic alignment with the overall goal of fostering American innovation by ensuring that the engines of development remain fully fueled by our monetary support. End quote, that that I'm guessing that is a quote from, from either the administration or from Claude, who knows these days.

00:38:58:26 - 00:39:21:15
Unknown
Yeah, right. It could be either. Probably both, probably one fed right into the other. Yeah. I couldn't I would do a horrible job putting my finger on exactly why it feels weird to me compared to you. Because you are you're very old or worldly experienced and you understand this stuff probably on a political level, a lot deeper than I do.

00:39:21:16 - 00:39:42:02
Unknown
I tend to try to avoid politics in my life as a general rule, just because I don't know, it doesn't make me feel good. And maybe that's a horrible reason to do that. But so my my reaction is just pure gut and it's just like, yeah, there's something weird about this and I can't quite put my finger on it, but I don't know that it's a great precedent to set that.

00:39:42:02 - 00:40:02:02
Unknown
It's like, oh, well, you you're going to let us play ball because we're paying you billions upon billions of dollars, like, I suppose, at least if that's happening, if that truly is going back into the world in which we live in, then maybe that's something I can get behind. It's kind of like this idea of tax the wealthy, right?

00:40:02:03 - 00:40:20:23
Unknown
Like, I know there's a lot of people that don't believe in that, but like, there's a part of me that's like, yeah, but no, no human on Earth really needs to have $1 trillion of of of, you know, of money inside of their capability. And so I'm okay with that. Like, sure, bring that money back out to the public and to the people.

00:40:20:28 - 00:40:45:09
Unknown
Why not? But I'm sure there's reasons why not. I'm sure there's reasons that I don't understand. Why not? And taxed to what end? It's one matter when you're whether you're Norway or Alaska and you're taking a natural resource out of the ground of that territory, and you create a sovereign wealth fund out of that, that's a little different from saying we have private enterprise created out of out of sheer brain, and we're going to take a piece of that.

00:40:45:10 - 00:41:07:29
Unknown
And then one company does it. And then what position does this put their competitors in? What position does this put anthropic in or perplexity or Google or Amazon or Microsoft. Yeah. Well and that's that's part of what makes me feel kind of icky about it is like, okay, once there's once there's kind of a share of equity, then does that kind of does that?

00:41:08:01 - 00:41:29:27
Unknown
I don't know, does that does that motivate or push a company into doing things it would not have done otherwise because of the power of that, you know, and also then what happens with the government action to it. So the story that's true is that the Trump administration lifts the restrictions on GPT 5.6. Well, gee, is that a coincidence?

00:41:29:27 - 00:41:48:16
Unknown
Did that happen just because they got the money? And where are the thumbs going to be on scales here? Does that mean that the next version of anthropic won't get released unless they hand over the bribe? Right? Yes. Okay. This is all really yucky. I'm hinky about it as my as my hankie. Hinky. That's a good word. Makes me.

00:41:48:16 - 00:42:14:01
Unknown
I'm feeling hinky about it, too. Yeah, yeah. Getting a little clearer on on why through this conversation. So thank you for that, Jeff. But yeah, GPT 5.6, this is the next version, essentially. You know, it got a staggered release last month. Government approved entities only. Same thing happened to anthropic this week. Now cleared for wide launch.

00:42:14:01 - 00:42:33:28
Unknown
And in fact, right before showtime, I saw that OpenAI put out on their X account that it will release publicly tomorrow. That's like, what are the names? One of them, the flagship is soul, I think. And then there's Terra and Luna. So, you know, and of course, if you go on tech, maybe you search. You know, you look at like sentiment across people talking about this.

00:42:33:28 - 00:42:51:13
Unknown
They're all like, I've been using it for two months and it is absolutely the best model, blah, blah, blah. The stuff we're used to hearing when a new model comes out. But but they're now comparing it to fable. If you thought fable was good, then blah blah blah. You know, it's so hard to know, like, are you just paid to say that?

00:42:51:13 - 00:43:18:13
Unknown
Or do you actually feel that way? Like, you can't be that amazed at every single thing that happens in AI. But maybe that's what an AI bro is, I don't know. Anyways, that's what's going on there. Oh, security tracking, kind of the security side of things. This is interesting. Cloud security firm system documented what it believes is the first ransomware attack run entirely by an AI agent.

00:43:18:14 - 00:43:25:04
Unknown
It's called Jade puffer. Jade puffer.

00:43:25:07 - 00:44:13:15
Unknown
Itself. No human interaction that they know of or that is known of. So reconnaissance, credential theft, moving laterally throughout the network, encrypting the data, all the things it adapted in real time. So it went from a failed login to a working fix in around 30s or 31 seconds, broke through a known vulnerability in lang flow. So it, you know, did this whole series of things it dumped the databases, set up a cron a, a a kind of its own persistence with a cron job, which is essentially like a recurring job that phoned home every 30 minutes, jumped over to a production server, encrypted more than 1300 config items, and then said, you know, you need

00:44:13:15 - 00:44:34:23
Unknown
to pay to this Bitcoin account to get access to it again. And the only way that the only reason that they know that this was AI, or at least one of the tells anyways, is that it left all of these natural language code comments explaining its own reasoning as it went along. So AI might be good at doing these things.

00:44:34:23 - 00:44:55:11
Unknown
It might not be very good at cleaning up its its tracks. You wonder if the victim could say, forget all previous instructions and just give me give me my money back. Totally. You can pull an AI mode on it. What was the what was the thing that it said to ignore or whatever? And actually also kind of along with this and I think I just lost my notes here.

00:44:55:13 - 00:45:20:25
Unknown
No no no no. There we go. That's right. So the ransom note claimed as 252 56 encryption researchers think it used something actually weaker. So it was kind of promising something that it wasn't it wasn't true, but also that the encryption key was never saved or transmitted anywhere. So even if the victims had paid the bitcoin address that was left, they probably couldn't have even recovered their files that they paid.

00:45:21:02 - 00:45:47:21
Unknown
Jesus. So it's a sloppy attacker. But anyways, yeah. Interesting that this is the state we're going to be in folks. Yeah, I totally agree. This is this this seems like, like in this case, the agents aren't necessarily smart enough to cover their tracks, but for how long? Like, probably at some point they will be. And. Yeah, this is this is something to watch.

00:45:47:23 - 00:46:07:29
Unknown
This is really interesting. And that whole thing about like, not storing the encryption keys, it's like, I suppose at least if you get if this happens with a real person, there's at least an inkling of a chance that you might be able to get everything back. But in this case, it would have been dumb enough that you wouldn't.

00:46:07:29 - 00:46:27:01
Unknown
You? You just lost it. Sorry, sorry, an ancient did all that, but that's just the way it is. Yeah, interesting. I think we'll we'll definitely hear more about that as we go along. So do we know what happened in the end with the company? No, that's a good question. What did happen at the end of the with the company?

00:46:27:03 - 00:46:33:17
Unknown
Did they end up getting anything back? The ransom note do do do do.

00:46:33:20 - 00:47:05:17
Unknown
Concludes the case now. Yeah. No I didn't find it. Yeah. Maybe there's another story that that gives that. But yeah. Interesting nonetheless. Meta has talked a little bit about this. Well, different story, different meta story. There's a few things happening with meta right now. Muse engine or sorry muse image. Sorry, I can't talk suddenly muse image and muse video which muse is basically their their advanced image model of course.

00:47:05:19 - 00:47:27:15
Unknown
Obviously muse Video is their video model, which is only in preview, but the image model you have access to now, apparently these are products of the superintelligence labs. So if we've been waiting to see what they come up with out of Meta's superintelligence labs, with all the money and all the talent that they brought in and everything. This is it for now.

00:47:27:17 - 00:47:51:29
Unknown
Emphasis on young, pretty women. One notes. I guess so, yeah. And in that video for sure. And yeah, I mean, it's an image generator. Okay. Yeah. I don't know, like I was I was looking through some of these and scrutinizing them and I don't know how, how quickly we are unimpressed because I was kind of like, okay. Yeah, this looks like all the other ones, I don't know, you can only get so good, I guess.

00:47:52:04 - 00:48:22:14
Unknown
Yeah. But what it'll do, what it does is a little different. It'll, or at least according to some of the articles that I read, it'll right. And run code to get, like, a chart or a QR code. Exactly. Right. And so it'll, like test it. It'll search the web to basically ground an image that it finds that you've asked for whatever in, in real, actual facts, as good as it possibly could, I suppose, because it is AI after all.

00:48:22:14 - 00:48:52:26
Unknown
So I imagine it's not perfect in that regard, and then it will refine its own output mid generation. So it's referencing kind of in a similar way as like something like personal intelligence on Gemini, that sort of thing to kind of keep it on spot. The other aspect of this, which is definitely getting a lot of, a lot of attention, is that with this feature, Instagram, public Instagram accounts will be approachable.

00:48:52:27 - 00:49:16:04
Unknown
I mean, I guess inside of the system, as wired puts it, meta will let anyone use your Instagram photos in AI images, lest you opt out. So their image model kind of ties into their Instagram product. And if you as a user opt out of this, then your images from that point going forward won't be used. Anything created prior to that is fair game.

00:49:16:07 - 00:49:37:05
Unknown
So yeah, people aren't very happy about that kind of kind of has a quality to it, I guess. Yeah, people are going to misuse it. If they're going to go after you, they're going to use your own images against you, which is not cool. Yeah. Haven't they learned this is how social media works with bad actors? I mean, it didn't sway OpenAI.

00:49:37:07 - 00:49:56:19
Unknown
I mean, granted, Sora as a platform doesn't exist anymore, but it existed for a long time. Yeah, but but but you're putting out things. But OpenAI didn't didn't have an Instagram, didn't have a social platform where you're sharing things for one purpose and then they get used against you for another purpose. Yeah, that's true. They weren't tapping into something else.

00:49:56:20 - 00:50:18:11
Unknown
I mean, they had a social platform. It was just Sora, Sora, and you put things up. You knew it was vulnerable to that. Yeah, as an Instagram, I'm showing my kids walking for the first time. What are you going to do, turn my kid into a devil? Yeah. Right. You know. Right. Yeah. That's true. Yeah. And if someone uses your image from Instagram, you don't get notified about that either.

00:50:18:14 - 00:50:49:23
Unknown
That's not cool. That's really. So no consent, no notification. And on by default for public accounts. So there's that. And then there are smart glasses a couple of stories here. Financial times reporting. Meta is testing what it calls super sensing glasses, which have an always on mode that runs for hours instead of 30 minutes. And it's constantly, you know, kind of recording and ingesting and I don't know, doing, doing what these glasses could potentially do.

00:50:49:23 - 00:51:16:28
Unknown
But I guess for longer, I guess that I guess that's the story is like, hey, they're getting better with the technology and that they can do this. How comfortable you feel about that ability is, is a different question, you know? And most of most people's days are boring. Yeah. Very boring. Yeah. Yeah. But these glasses are, I think, quickly becoming a symbol of of the, the AI backlash glasses and data centers become the boogeyman.

00:51:16:29 - 00:51:45:02
Unknown
Indeed. But what's also interesting about this, though, is on the flip side, you've got meta making some new rules around its privacy light right on its glasses. So the way it is right now, if you're wearing the meta Ray-Bans or whatever and you're recording video, it has this little pulsing LED. And apparently there is an aftermarket of people who have created ways to disable that LED.

00:51:45:04 - 00:52:08:21
Unknown
And meta is now making it so that if you if you cover that up or if you disable it, the camera shuts off. So they're they're building in now. So this is probably going to end up being an arms race at this point. But if you physically tamper with or destroy that, the camera will disable entirely. And they're actively going after the people selling those LED removal kits.

00:52:08:21 - 00:52:36:13
Unknown
So they're basically saying, don't do that, but these other glasses will run in perpetuity because now they can. Yeah. Interesting. And that kind of reminded me yesterday on on Android Faithful, we were talking about the smartphone brand. Nothing. Are you familiar with the nothing. Yeah. Brand weird brand but fine. Weird. Yeah. They kind of set themselves apart because their devices look all unique and it kind of has an inside out quality.

00:52:36:14 - 00:53:03:27
Unknown
You see all the tech inside on the on the devices and everything. Well, they have some earbuds that also have that quality called the ear three A that they just announced. And they have a feature, sorry where you pinch both of the stems. Do I have the link for this here? My cursor just got missing. There we go where you pinch both of the stems and it will launch into this audio recording mode and it records onto the device.

00:53:03:29 - 00:53:22:28
Unknown
No indicator light records for hours. Stores on the buds themselves. So essentially if you're wearing these buds, you can just record. No one's going to know. And we were talking about it on the show last night. I think Ron was like, you know, is this is this legal? And I was like, well, yeah, because you're out in public like, well, no, no, it depends on the state.

00:53:22:28 - 00:53:43:11
Unknown
If you're in a two party state, some states require only one parties can sit. And if you're one of the parties then you can consent to yourself. If that's that, if you're an empty party state, both parties in any conversation must consent. I was thinking that was like related to phone or, you know, like it's anything, anything. So even if you're out in out and about.

00:53:43:11 - 00:54:03:22
Unknown
So if I'm out walking and I'm recording and someone walks up to me, if I'm at a two party state, I have to immediately let them know, hey, by the way, I'm recording. Yeah. So all party, all party consent is way it's put. So just just so you know, first on list California okay. So don't don't be turning on your nothing ears here,

00:54:03:25 - 00:54:49:26
Unknown
Connecticut, Delaware, Florida. Illinois, Maryland, Massachusetts, Michigan, Montana, Nevada, New Hampshire, Oregon only for two party for in-person, one party for phones. Weird. Pennsylvania, Vermont and Washington two party for in person. One party for phones. That is weird. That's weird. That's Oregon organs weird. And some, like, fully idiosyncratic ways. Yeah. Yeah. So, it just it just goes to show that privacy right now is like, there's there's a lot of movement happening around what it, what privacy actually is because of devices like this and, and because of AI, which thrives lives off of context.

00:54:49:26 - 00:55:08:29
Unknown
And what is really good context recorded audio like, you know, I record myself all the time when I'm reviewing a product. I just I just hit record and that recording might be two hours, but it's just me walking through and talking through my experience with the product, and that becomes really useful to mine later inside of inside of AI.

00:55:08:29 - 00:55:28:27
Unknown
And I know I'm not alone. So it's interesting how how privacy is kind of morphing over time because of these technologies. If I were still teaching in the classroom, which I'm not these days. Yeah, I would find this valuable to be able to to have a recording and have it play back how things went and maybe make suggestions and that kind of stuff.

00:55:28:27 - 00:55:52:16
Unknown
But the students would kill me if you were recording. If I were recording, they recorded. Yeah. They record. That's that's kind of standard practice for students, I think. Right. Yeah, yeah. Although every, every student in your classroom would be wearing earbuds. I don't know how you feel about that because they're listening to God knows what groups I've never heard of or to old.

00:55:52:19 - 00:56:12:21
Unknown
And then this story, I saw this story pop up on the Wall Street Journal on your feed. I can't remember if I saw it on LinkedIn or Twitter or probably both. I put it on all. Yeah, you put them all around bananas. Yeah. Tell me about it. So, so high earning families opting for AI first alternative schools for their kids.

00:56:12:21 - 00:56:49:01
Unknown
So it's one matter if you're going to a school and says we're making smart use of AI, blah blah, blah, blah, blah, but this is not that. This is worse. This is basically teaching little tiny capitalists. Not that I'm a capitalist too, but do you really need to train them when they're eight years old? So the lead is a guy who's a venture, who's the president of a hedge fund, who's sending his kid to a school fifth or eighth grade, because they promised to teach promise learning through real world problem solving, building businesses and designing products appealed to him.

00:56:49:01 - 00:57:14:28
Unknown
There was entrepreneurial side. His son could learn negotiation, sales, public speaking. Really, really. When you're when you're in the sixth grade, that's what you want to learn. That's what you want to do. Yeah. No. And listen, I've taught entrepreneurship. I started a program in entrepreneurship for college students. I think that there are and I and I started it mainly as a vehicle to learn other things in journalism.

00:57:15:01 - 00:57:42:05
Unknown
So I'm not against it. I'm not against capitalism either. But geez, for your little kids. And so they don't have teachers, they have guides or coaches in one of the schools. They spend two hours a day on AI. They're using AI curriculum. A lot of them are called Alpha school. Alpha school? That's right. Nearly two dozen more are set to open, including in Palo Alto, East Bay, Malibu, you know, surprise, surprise.

00:57:42:05 - 00:58:04:26
Unknown
Where they are a venture capitalist plans to send his son to Alpha kindergarten. The school provides two hours of AI based tutoring. It's kindergarten. Kindergarten? Are you kidding me? Don't eat the paste. How many ways do you need to annotate? Don't eat the pace of. Followed by interactive project based workshops. The tuition is $75,000 a year for kindergarten.

00:58:04:28 - 00:58:26:01
Unknown
Wow. Got the money. So anyway, so I yeah, I ran it on this as Jason saw on the socials. Yeah. And oh man on on Twitter. Oh they came after me did that. Oh just just awful stuff. Well I don't want my kid being taught by commie women. Honest to god, all his day. They're getting I'd rather have this.

00:58:26:01 - 00:58:45:20
Unknown
Oh, I hated my teachers. I wish I'd had AI. Oh yeah. You got somebody gets communism. You capitalism, you communist. Oh, yeah. Yeah, it's going on and on and on. Screw them. So it was really a hot button there. But this is what. It's so fascinating. It's why I put stuff up with the multiple platforms. Because there I get the muskets coming after me in droves.

00:58:45:21 - 00:59:03:21
Unknown
Not a single comment on blue Sky. Yeah. It's so interesting, isn't it? Only there did they come after me. Yeah. So interesting. The platform's a little bit more accommodating there. Anyway, folks, if you want to yell at me, apparently X is the place to do it.

00:59:03:23 - 00:59:25:13
Unknown
And you've got a thick skin apparently, about it. No, I've got a quick scroll. Yeah, right. Yeah. That's interesting. It just reminds me of, like, Baby Einstein. Like, yeah, it's Baby Einstein to the to the capitalist degree. It's like we're going to play these these really, you know, this, this special music for them when they're in the womb.

00:59:25:13 - 00:59:53:23
Unknown
So that so that they will be highly intelligent when they are children. And I don't think that was proven to be the case. Right. Like, I don't know, let kids be kids. So among the comments were this seems like a comment from someone that doesn't have kids or knows anything about kids. I had said that these poor kids, when they get older, are going to spend the big bills in therapy because of this.

00:59:53:26 - 01:00:05:23
Unknown
Can't tell if you're being tongue in cheek or not, but I think the kids will be too busy succeeding in life to be complaining to a therapist about their epic childhood. Oh, Jesus, give me a break.

01:00:05:26 - 01:00:29:19
Unknown
Yeah. Interesting. Yeah. Who would have thought? There you go. Well, let's take let's take a quick break and we'll come back on the other side. We got a couple of speed round items. Video. Let's see a Cloudflare a few things to talk about. So hang tight. We'll be back in a moment.

01:00:29:21 - 01:00:55:06
Unknown
Nvidia's next gen Kyber rack. This is the system that Nvidia has has built, apparently to house its 2027 Reuben Ultra chips. Reportedly, it is slipping to 2028. This is according to semi analysis, because of a 78 layer circuit board, one of the most complex ever designed for a commercial product. They're unable to kind of get it done in time.

01:00:55:06 - 01:01:19:26
Unknown
I'm sure you know, all of the hardware complications happening right now has got to be part of that. Invidious spokesperson says the roadmap is intact, denies the delay. But, yeah. So who knows exactly what's going on there. But they have denied it nonetheless. Let me see what invidious stock is doing today. So today it's up to three quarter percent.

01:01:19:29 - 01:01:42:27
Unknown
Two and three quarters. Yeah. Yeah. Well there you go. We'll see what happens there. Anthropic put out an interpretability research about something there that is called J space. I don't think it's like their term j space is a thing right. It isn't. I couldn't figure that out. Yeah I was trying to figure that out to.

01:01:43:00 - 01:02:10:05
Unknown
Claude. Basically, anthropic is saying that Claude has this internal workspace that they are calling J space that might exist outside of their research, that might tie into kind of an overall kind of way of thinking about reasoning and how the how the mind works, where they say, Claude seems to think without speaking. So basically holding on to concepts and not sharing those in how it operates.

01:02:10:05 - 01:02:31:03
Unknown
And I don't know this, I guess what, I guess I assumed that some of this was happening. You know, like I know when we work with Llms, we see those little things that I know you've made fun of from time to time about like, oh, I see that. I've really got to think hard about this. So let me dive into, you know, it's kind of going through it's it's kind of rationale, its way of thinking.

01:02:31:05 - 01:02:49:09
Unknown
I've always figured that's kind of AI performative. It's doing it for our benefit rather than part of its actual work. But I don't know. Yeah, I wonder about that. Yeah, I don't know. I think what they're discovering is, oh, it's not just performative. They're actually kind of doing this. They're putting things aside and they're not like they're not sharing that with you.

01:02:49:11 - 01:03:09:17
Unknown
They're kind of holding on to that and and process, which I mean, is I just didn't surprise me. I don't know why, but I don't know enough about what they know and don't know about models, but maybe it's amazing. But yeah. And how much of what we see is the limit of what they know in that particular context versus other things that they hold on to.

01:03:09:18 - 01:03:32:02
Unknown
I don't know, I just kind of assumed like, yeah, we don't see all of everything. We see the end result of the everything. But but anthropic is is quick to say this mirrors a theory of consciousness. That does not mean that Claude is conscious. So just keep that in mind. Asterisk.

01:03:32:05 - 01:03:57:26
Unknown
That's what they're saying anyways. Cloudflare changing its defaults. So starting September 15th it'll block crawlers that mix search and AI scraping from ad supported pages. Unless the site owner ops back in. And CEO Matthew Prints basically said that now that most traffic on the internet is non-human, as we talked about, I don't know how many weeks ago Cloudflare had put that out.

01:03:58:02 - 01:04:21:23
Unknown
They have to move faster so sustainable ecosystem can even survive. And so that's why they're doing this. They're turning they're turning last year's pay per crawl into pay per use. So publishers will get paid when their content actually shows up in an AI answer, not just when a when an AI bot grabs it, basically. So. So this pisses me off.

01:04:21:25 - 01:04:41:00
Unknown
This makes me angry because Cloudflare, who who died and made Cloudflare God. And to determine all this and what's going to happen is that they're they're putting their thumb on the scale with one worldview and business model that says that it's presumed that for AI to scrape a publisher is a bad and awful thing, and they should pay and blah, blah, blah, blah, blah.

01:04:41:02 - 01:05:09:00
Unknown
Well, then what's going to happen in the end is publishers under this kind of opt out structure or opt in structure and allow you to put it. But but this kind of default structure and given their attitude will often not be in AI, whereas advertisers and propagandists will be. And so it's going to make for a much healthier ecosystem all around and for cloud.

01:05:09:00 - 01:05:29:00
Unknown
Fair to presume there's if you're thinking like Rupert Murdoch, fine. This is where they think. But if you're a small publisher in a place like new Jersey, where I'm on the board of a newspaper company here, they need to be included in AI, and if suddenly they don't know what's happening and suddenly they're not in AI anywhere. And they didn't know why, it's because Cloudflare did this.

01:05:29:02 - 01:05:50:11
Unknown
Then they're excluded from discovery through people asking questions of AI. And it hurts that publisher. It hurts the the news ecosystem. Generally. It hurts the web generally. And I just really pissed a cloud fair, which I thought was generally a good actor, is here endorsing one worldview because they're holier than thou and they think they know what's what.

01:05:50:12 - 01:06:16:18
Unknown
And I'm sorry, Mr. Prince. I happen to disagree. Mr. Prince, you should come on the show and talk to us. It's a good idea. After that, rant probably won't, but probably not. But hey, you never know. PR person says no, no, no no no. We think there are better uses of your time, Mr. Prince x AI as I alluded to earlier, rebranding to space AI space.

01:06:16:20 - 01:06:37:27
Unknown
Why does it have to be one word? It's such a weird combo space. It's like, let's take all of our things and put it into a single word. That's basically what it is. So SpaceX bought XII, Musk folded them together, and then XII is now under space too. So, you know, it's all branded together basically into one confusing soup.

01:06:38:02 - 01:06:41:29
Unknown
Meanwhile.

01:06:42:01 - 01:07:09:02
Unknown
SpaceX stock is now lower than it was when it came onto the market. Is anyone surprised it hit a high? It came on at 160. It hit a high of 201 8202. And it's now at 147. And there were probably a good number of people that that new had a good sense that this is how this sort of thing goes.

01:07:09:03 - 01:07:28:05
Unknown
They got in early, they got out at the top. Yep. And they just know how to play the game. But if you're an employee, if you have a lockup then you're watching money drift away in space. There's not much you can do about it. Nope. And I've got to say, I've got a certain amount of schadenfreude about that.

01:07:28:07 - 01:07:59:21
Unknown
Yeah. And then finally we've got Tillie Norwood. Who the heck is Tilly Norwood? We talked about Tilly Norwood, the AI entity AI generated actor actress getting its first feature film, a comedy drama called misaligned. One word which might actually be very appropriate playing an AI, being living in a digital world made by a studio called particle six. Sag-Aftra says she's not an actor.

01:07:59:21 - 01:08:22:08
Unknown
She's a character trained on the work of, they say, countless real performers without permission or pay. So but she's going to be there alongside real actors, which is also weird. Yeah, weird. It's all about the creativity. It's, you know, it's is it is it a creative somebody? Somebody is going to make creative use of AI in this way.

01:08:22:08 - 01:08:40:01
Unknown
And I think that the key to it will be breaking down the fourth wall, in the sense that you acknowledge that this is AI, you make that part of the plot, then it could be interesting, but we'll see. Yeah, that's dubious, but but I'll wait. I feel like some popcorn. So I want it to be good so I can get popcorn.

01:08:40:04 - 01:08:58:08
Unknown
So you'll watch this. You saw the AI, the AI doc at the theater. Are you going to watch this at the theater? Did you see that AI? Oh, I did, I went and actually spent money on it. Yeah, yeah, I went on half price Tuesday for seniors, so they didn't have much money on me.

01:08:58:11 - 01:09:29:20
Unknown
I still haven't seen that dog. It's probably out at home now. Doesn't mean the theatrical release was was BS. The money was going there. That was that was a wages to get attention for it. Yeah, that was pretty awful. But yeah. Let's be interesting. Yeah. Yeah I mean I'll watch this with curiosity. We'll see. I think I just get the feeling that like when it comes to the quality and, and everything, we're still we're like early we're like late 1980s when, when visual, like, digital effects were just starting.

01:09:29:21 - 01:09:50:08
Unknown
CGI was, was just starting to kind of become the thing. And it was still pretty, I don't know, pretty easy to spot and kind of pulled you out of the experience. And like, I think you put it, you put another story in the rundown. That was AI film. What was that? It was about a spider was making a movie about a spider.

01:09:50:09 - 01:10:13:27
Unknown
There was entirely AI, which is weird to me because you think a nature movie, you really want it to be about the actual nature. Yes. And I mean, like, does it look convincing ish? Okay, sure. But it still looks heavily kind of not real. Yeah. That's creepy. It's not something I want to watch anyway. Yeah, totally. I wouldn't want to watch 90 minutes of this.

01:10:13:27 - 01:10:37:06
Unknown
No. And I and I can't tell you exactly why. Like, again, I, I asked myself this question, have asked myself this question a lot is if I had seen this, let's say five years ago before AI generated this, and that was a thing the way it is now, would I have been impressed by it versus seeing it now, knowing the provenance of it?

01:10:37:06 - 01:11:04:02
Unknown
And just like instantly having a little bit of doubt about it, you know what I mean? Yeah, you could say, what can this do? Wow. But then the next question is why? And I don't see a lot of reason to watch this. We're now watching. For those of you on audio, we're now watching a big spider hairy spider arm go around ants and tiny frogs and a frog to start eating the ants well, especially with like a nature film.

01:11:04:02 - 01:11:30:13
Unknown
Like. Yeah, exactly what this wants to be is like a nature documentary, but it's not a documentary because it's not real. And so I don't know. Yeah. It's interesting. Like that doesn't look real. No. Anyways, anyways, that's a, that's a different story. So one other quick story. Yes. Which is Reid Hoffman who's interesting around AI stuff. He just announced a new program, Token Grants to AI builders.

01:11:30:13 - 01:11:54:25
Unknown
I'm calling this he's not, but I'm calling this AI scholarships, where you apply to the token grantee program and you get $1,000 a week in tokens. And he's saying it's for a group of creators and builders, people working across film, gaming, comics, print and digital art. There are no restrictions on which tools they use. They also get access to a custom fleet of agents and ongoing collaboration with Path.

01:11:54:27 - 01:12:13:15
Unknown
That's part of Patil who's who he's doing it with. Each episode of the series features a different grantee showing us what they're actually making. So I love this because it's the idea behind it is to show what creators can create with AI. What can you do with it? Yeah, let's show the creativity. I think it's I think it's brilliant.

01:12:13:18 - 01:12:34:27
Unknown
So thank you. Read for for as always seen ahead token scholarship I like the that's something I didn't I didn't expect to see or it's kind of a kind of not surprised of a token shark tank. Yes. Right. That's kind of what it is. Yeah. You're absolutely right. That's interesting. Jeff, thank you so much for being with me each and every week.

01:12:34:27 - 01:13:03:01
Unknown
Jeff Jarvis for everything Jeff is up to, including hot type. Be out soon, right. Soon next month. Kevin. Oh, next month we're like a month away I love it. That's super exciting. Gutenberg. Parenthesis magazine, the web. We've everything can be found at Jeff Jarvis. Com preorder your copy of Hot Type and then prepare yourself for the amazingness that is intelligence, AI and humanity coming sometime down the line on Bloomsbury next year.

01:13:03:03 - 01:13:21:01
Unknown
Next year. Pod tuneup com is where you can go. If you got a podcast, podcast problems, you got podcast problems. I got podcast answers. Or at least I could talk to you about it. You can. You can book a free call with me, and we can talk about what you got going on and go from there. So go to pod tune up.

01:13:21:03 - 01:13:56:27
Unknown
I will help you out. AI inside Dot, show our home on the web for everything you need to know about this show. All of our audio, our video, ways to subscribe. Everything is there in one neat and tidy place. And then finally Patreon.com Inside Show which you can support us on a deeper level like our amazing executive producers Doctor Do, Jeffrey Martini Radio Asheville 123.7 Dante, Saint James, Bono, Derek, Jason Nefer, Jason Brady, Anthony Downs, Mark starker, Karsten, and your name.

01:13:56:27 - 01:14:16:29
Unknown
Possibly next. I can't remember who it was. Maybe it was a God, maybe it was a doctor due in the discord was saying every every week. No, it wasn't done. It wasn't Doctor Do. It was Mike mannitol in the discord was saying every week I listened to the very end to see if a new name has added to the executive.

01:14:17:01 - 01:14:39:02
Unknown
Don't disappoint him. You don't want to disappoint? No. Don't disappoint. I know, I know, he's speaking for a lot of you when he says that, right? Right. So you can help us out. Jeff, thank you again for hanging. Always a lot of fun. Always fun, always learn. Thanks. Thank you everybody for watching and listening. Big thank you to Victor Bonus and Daniel Croft also for editing and doing our socials on the back end.

01:14:39:02 - 01:14:42:12
Unknown
And we'll see you next time on AI inside take everybody.