AI Is Eating Its Own Tail

00:00:00:10 - 00:00:27:09
Unknown
Coming up next, Jeff Jarvis and I dig into the wild and wooly week of AI all over the map. We're talking open source, Kimmy, K3, Nvidia's tendrils and all of it. Washington you know, accusing China of stealing from anthropic OpenAI's own employees, urging the government to slow things down. And I feel like I'm touching on a 10th of what we're talking about.

00:00:27:09 - 00:00:55:19
Unknown
It's so interconnected. It gets confusing really quick. But thank God for Jeff Jarvis because he makes so much sense out of all of it. Also, smart glasses, a little bit of a backlash going on there and a few other things. Coming up next on the AI inside podcast.

00:00:55:21 - 00:01:16:13
Unknown
What's going on, everybody? Welcome to another episode of AI Inside the Show, where we take a look at the AI that is layered throughout the world of technology. Through all the AI stories are layered throughout each other. I'm telling you, this rundown was hard to put together because I feel like every flippin story feeds into every other story, but we're going to get through it because there's a lot to talk about.

00:01:16:13 - 00:01:36:14
Unknown
I'm one of your host, Jason Howell, joined, as always by Jeff Jarvis. How are you doing, Jeff? Hey, good to see you, boss. Great to see you too. Great to see you too. Podcast and machine. Yeah, it's it's another one of those weeks I do another podcast that I do is called Daily Tech News Show with my long time friend in podcasting, Tom Merritt.

00:01:36:14 - 00:01:58:13
Unknown
And so I'm very lucky and fortunate to be able to be one of his writer producer hosts on his staff. And when he goes out of town, I end up taking his slots. But that but it's a daily show and so that's like five extra days. And yeah, I'm feeling a little mentally kind of like not drained, maybe a little drained.

00:01:58:13 - 00:02:22:06
Unknown
But you know, podcasting like it's a very mentally challenging it really is to do one after the other after the other. It's yeah, it's like my brain needs to take a nap. And as much as I love to hear myself talking, oh, I do. Yeah. You know, it's it's challenging. Sometimes you get to the other side of those really challenging podcast days and I'm just like, I don't want to talk to anybody.

00:02:22:08 - 00:02:42:22
Unknown
I'm gonna, like, turn on a movie and shut up. And you weren't going to get very much for me. I'm sorry. That sounds nice. Yeah, it does sound nice. Hey, I want to I want to congratulate you, first of all, because that Nate B Jones interview was actually nice of you on intelligent Machines. People should definitely check it out.

00:02:42:24 - 00:03:06:02
Unknown
So the funny thing that made him laugh most was you and I were trying to get hold of Jones. They were trying to get Old Navy Jones. We're all trying to hold Navy. Jones finally, Leo asked his AI and that came up with the email and who knows where where that AI. You know that AI went down into the the ocean and came up with a license plate that was sitting on the bottom of the ocean and it happened to have have the like information.

00:03:06:06 - 00:03:31:24
Unknown
He's really very he's so knowledgeable and a very nice guy. Yeah, it was cool. It was cool because because what we're used to seeing is the the TikTok slash YouTube slash Substack writing side of Nate very directed and everything. And it was kind of cool to see him a little more relaxed. And, you know, not that he's not relaxed in his videos, but it's, you know, it kind of humanizes a little bit more.

00:03:31:25 - 00:03:55:02
Unknown
That's what podcast is really good. I might have mentioned last week. It's fascinating to his wife, whose name I'm suddenly forgetting, is an author of science fiction author, nonfiction author. She also happens to be legally deaf and blind. Yeah, and so her interaction with AI and their interaction together with AI is just a fascinating side angle here. Somebody should do a feature story on that because it's it's amazing how AI becomes a tool in life like that.

00:03:55:07 - 00:04:20:08
Unknown
Well, yeah. What I can't see for it that others do or the frustrations for sure. Yeah. He was he was saying in that, in that interview that it's just such an accessibility, jump, you know, improvement for her life, you know, and yeah, it's an interesting angle of, of AI how how much it enables for people who are less than abled in other ways and throughout their life.

00:04:20:08 - 00:04:48:28
Unknown
So pretty cool stuff. Yeah. Anyways, great great interview. Folks should go find out or find intelligent machines. You can find its way out TV. You can find last week's episode, and it starts with about a 45 minute interview with Napi Jones, who is just a really cool guy in the world of AI. So you'll learn something. And actually, I got to the end of that and I was like, I need to check out Open Brain Have you, which is not yet not states kind of thing.

00:04:49:03 - 00:05:21:28
Unknown
Yeah. And I haven't spent any time I did. I did sign up again for his Substack because I wasn't on it. So now I have at least the next month to kind of go through some of his content, and he has some real deep dive content on how to set up an open brain. And I want to do that because, like, I've been going the more of the the like clod coworker os the carpet karpathy approach, the like wiki everything that's been kind of my approach.

00:05:21:28 - 00:05:37:00
Unknown
And then he was talking a lot about Open Brain in that interview and I was like, God, I wonder how compatible those two things are. Can I just move the brain that I've been building into open brains that actually work better for me? Yeah, I think that's what the power is here. And we'll get into this in a second here.

00:05:37:00 - 00:05:58:04
Unknown
But that's where the power of having so many open weight models out there, and the choice among models is so important, because if you just have the nucleus of your stuff and these models can call upon it, it leaves you so much more in charge. Yeah. Totally. Right. Go ahead. Yeah. Well, I was just going to say, because right now I'm doing it all through Claude, through one thing.

00:05:58:04 - 00:06:16:14
Unknown
I'm really kind of relying on that one thing. It might not be the best at all the things that I need to do. Exactly. And this goes to the Palantir model of saying, hold on to your own stuff as much as you can. Yeah, right. And by the way, the little gag here is it's open brain one, so it's Obi-Wan.

00:06:16:16 - 00:06:39:20
Unknown
I did catch that. That was good. That was good. That was that was well caught off the show as well. I would I would not have put that to, you know, those things together. So very cool. Where do we even start man. So before we get into it. Yeah, because Jason and I were commiserating about this, there's so much news and so much of it is intertwined this week.

00:06:39:21 - 00:07:11:00
Unknown
Yeah, it's a whole story of open weight models of China, of the reaction of the frontier model companies to all of this, what they're doing that's a little bit different. The all the other companies that are behind open now and the pressure to do that. And then the regulation talk on top of the OpenAI rogue AI goes, goes on a on a trip across the internet trying to make it look like it's a Bonnie and Clyde.

00:07:11:01 - 00:07:31:22
Unknown
Let's just do what I'm supposed to do. So you put all that together, and that's kind of the story of the week. It's about ten stories in one. Yeah, that was that was assembling kind of trying to find some sort of order because I am a very in downs, as you know, I'm a very orderly person. I'm like, then we'll do this one and this one and this one, and it just works for my brain.

00:07:31:22 - 00:07:48:00
Unknown
And I was like, God, I just don't even know. There's so many different directions that this this goes. There's, you know, like I have up at the top and I don't even know why I put it up at the top, but kind of like all the invidious tendrils that are, that are kind of making their way through a lot.

00:07:48:01 - 00:08:13:19
Unknown
Well, I think, I think go ahead and start there. And I think we can use that to beat the picture. Well, yeah, I mean, Nvidia basically leading off with this open secure AI alliance. Right. Kind of collecting 50 companies, more than 50 actually. But you mentioned Palantir Microsoft hugging face of course Salesforce Cisco Dell is a number of GitHub SAP.

00:08:13:21 - 00:08:34:09
Unknown
What's interesting about this though is that you don't have OpenAI on that list. You don't have you don't have anthropic. Well, you got you got a couple different you got three different things going on here. You have the defense, the letter that was the open the defense of open. Wait. That's 50 companies that did include Google finally include yes I did include OpenAI though.

00:08:34:10 - 00:08:58:21
Unknown
That's kind of weird. It did not include I don't think it still has been signed by anthropic. Then Nvidia put together a separate initiative, and that doesn't include Google and doesn't include a lot of the other big companies. Right? Right. That was kind of the what was on, on on Little Rascals, the Superman girl hating man's club or whatever it is, you know, they decide who's let in.

00:08:58:21 - 00:09:22:26
Unknown
And then there was a third kind of initiative that they started. So it's like Nvidia there's so many initiatives. Yes it is. There's a letter and there's the initiative and then there's the response to the letters. But and I mean are they they're kind of all kind of pointing in a similar ish direction. Right. I think what I think what they're recognizing is that there is a fear going on.

00:09:22:28 - 00:09:41:04
Unknown
There's a tech clash, which we've talked about a lot. Right. Yeah. After the opening, I story, which I think was handled very badly by OpenAI, the hugging face story. Yes, it it went into hugging face in a way that was unexpected, but it just was doing his job. It was doing it was told it was in a security test.

00:09:41:07 - 00:10:04:16
Unknown
It didn't have guardrails on. So it wasn't that bad. In my view, it was a big deal. It was very clever about how it got in there. It was amazing. But I don't think it was anything, you know, rogue and dangerous and badly motivated because it has no motivation. All right. Nonetheless, that still sets off a panic in media and in policy.

00:10:04:16 - 00:10:22:15
Unknown
And it opens the door to politicians. We'll get to one case in a little bit trying to regulate. So I think when Nvidia comes along and says we're going to do two things at once, we're going to defend open weights because he believes doesn't want believes in it. And the more people use AI, the more chips he sells.

00:10:22:18 - 00:10:43:10
Unknown
His open weights is good for Nvidia. It's good for them. It's fine. Yeah. Yes or open models are good for Nvidia's business. And but at the same time he's saying we have a security, we have a way to become secure as an industry ourselves and don't do that. On the contrary, you have another story in here. We'll have to get to whatever order.

00:10:43:10 - 00:11:05:28
Unknown
But I think we're going we're going to do as we list them. So you have open AI and anthropic both say saying on the one hand, no, no, no, we're okay with open wait. No no no no we're fine. Yeah. But but then you have 1000 employees from the two companies saying regulate all of this. So they throw their own employees kind of out on stage saying, yeah, we're dangerous.

00:11:05:28 - 00:11:27:14
Unknown
You should regulate us because that's regulatory capture. And they want to decree how that goes. And you have anthropic saying, don't sell trips to China. And you have Justin Wong saying, yeah, what? I sell chips to China. And it's all getting this really weird spiderweb here. Yeah. I mean, if you if you go to tech meme, it's like half the entire page is all stuff that's intertwined and interrelated.

00:11:27:14 - 00:11:47:20
Unknown
And, you know, maybe that's inside baseball. Not everybody goes to tech meme. I spend a lot of time there just because it's a quick and easy way to kind of get a quick sense of of what's happening. And I'm used to things being super nested and they're they're all scattered, but they're all related. It's just really hard to kind of keep up with things.

00:11:47:21 - 00:12:08:06
Unknown
Did we mention last week? I don't think that we did that. The hugging face. Because now I'm getting a confused because I listened to your interview with Nate and that kind of touched on some of this stuff, so I can't remember if it was our show or that show. That hugging face tried to use a model to kind of address models.

00:12:08:08 - 00:12:28:13
Unknown
Oh, so it was three different US models that they tried to use. None of them would work because of the guardrails that were in place upon them, and so they ended up opting for an open weight Chinese model, GLM 5.2, and they were able to run that on their own hardware. Of course, they were able to reconstruct the entire attack and get the job done essentially.

00:12:28:14 - 00:12:55:28
Unknown
Exactly like yeah, yeah. So so all this talk about guardrails. Is false comfort I think. Yes. Yeah, I think that's that's right. For two reasons. I always argue on the show that it's false comfort in the sense that you cannot anticipate everything bad that would happen. And so to think, oh, this is safe now. We're okay now, whether on purpose, whether a malign actor or whether accidental, you can't anticipate everything and you can't then say we're guarded against everything bad that could happen.

00:12:56:01 - 00:13:19:09
Unknown
Point one. Point two is exactly what Hugging Face learned. The guardrails prevented them from using it defensively because the the bad guy of OpenAI's model that didn't have any guardrails on it, or that's the wrong word in this case that was given no limiting instructions. Right, right, right. So it could go. And it was search in the internet.

00:13:19:10 - 00:13:36:14
Unknown
It's always searching the internet and found. Oh yeah. Well Hugging Face has all this stuff. How do I get an hugging face? Okay, I'll figure out this way, this way, this way. I get hugging face. Right. It wasn't told not to do those things. So it put hugging face at a disadvantage with American models.

00:13:36:16 - 00:14:05:18
Unknown
So complicated. So so the the Nvidia Open secure AI Alliance includes company Adobe Cisco, CrowdStrike I'm looking down here Microsoft. But who's not in this list. There is no there's no open AI but there's open claw. There is no anthropic. There's no Google. There's no Amazon. So I don't know what the politics is here where they asked and refused.

00:14:05:19 - 00:14:24:01
Unknown
Did Nvidia say, no, you can't be in a clubhouse I don't know. It's weird. Comprised of invidious said you can't be in our clubhouse. I would think so too. But Microsoft Dell, space I mean if you if you let Elon in, for God's sakes, you let him.

00:14:24:03 - 00:14:50:02
Unknown
Pardon me Linux Foundation and so on. So this is an effort to it argues that the best defense against attacks in the age of AI models like mythos five is that security researchers need access to both open and closed frontier models. Okay, so they cite the attack on hugging face and say that they've got to create these tools in a way that you can use it as defense.

00:14:50:04 - 00:15:23:28
Unknown
So that's what that's about. The letter, which first 25 companies signed, then 50 companies signed. Google didn't sign at first, but they did is went on OpenAI signed. But anthropic my knowledge still has not signed is a rousing defense. Did we talk about this last week? I don't know if we did any more rousing defense of open weight models and open, open AI for the reasons of more commerce, more innovation, more competition, more control.

00:15:24:01 - 00:15:46:21
Unknown
Defensive use, and so on. And I think it's really I think it was a wonderful letter. I think it was a great, great letter. And I agreed with everything in it. Then you had that would be this, is that this? That's the only way that American leadership, okay. I've got so many links. I'm like like filtering through them, trying to find the things that we're talking about here.

00:15:46:21 - 00:16:03:28
Unknown
And of course, if you go to the bottom, does it have the 25 or the 50 in that that's the big list. Yeah. So that's yeah. That's all big list. Yeah. Yeah. But I don't think anthropic is there. I don't think they've signed up. So Dario came out and said well no no no I'm not against open. Wait I'm okay with open wait.

00:16:04:00 - 00:16:23:09
Unknown
That's right. But I'm not going to sell. I don't want you to sell to China. If you scroll up to his to his bullet points there. Jason, my primary concern is the risk of authoritarian governments. My secondary concern. No. Back up. I'm sorry. The one, two three. Oh. The 123. Yeah. My secondary concern is the risks of powerful A.

00:16:23:09 - 00:16:40:24
Unknown
It might be misused to carry out cyber attacks and biological weapons and so on. So as a result, he what he wants to do is one not sell powerful chips and shipping equipment to China like nothing they can't figure out in their own two how did that work out? So for us already like that, haven't we already tried that?

00:16:40:25 - 00:16:59:00
Unknown
So now they're going to make their own chips and be better in the long run than ours. Are all sufficiently capable models open and close should go through a mandatory safety testing. This is regulatory capture. This is him saying, well, trust us, but all your open and stuff, maybe it fails. And if you could scroll up to the next one.

00:16:59:02 - 00:17:27:21
Unknown
I'm sorry to make you and define sufficiently capable. Exactly. And who's going to decide what should go through. So so that was the reaction. Then you have the letter from the thousand OpenAI and anthropic employees. I don't know if you have that one available. I didn't have that one. Where yeah, there we go. 100 1100 call for the US to slow down tech development.

00:17:27:27 - 00:17:54:03
Unknown
This is a classic case of pulling the the ladder up into the blimp after you're flying right there saying, oh well, everybody else is slow down, but we're okay. And I forget, was it Taylor Lorenz or someone said on the socials? No, I think it was Adam Tier. I think that's in the rundown, actually. He said, I'm messing all around the rundown.

00:17:54:03 - 00:18:15:22
Unknown
Jason. I'm sorry. No, you're going you're going just fine. I, I figured today was going to be like this because one of those way to, to have any sort of. So Adam here who's a libertarian technology analyst who I respect greatly, he's at RSI and the firework. He said OpenAI and anthropic are free to slow down their own AI development efforts.

00:18:15:22 - 00:18:46:28
Unknown
All they want. They can cap their compute spend and cut back their own capabilities in various ways. That would be a huge loss for America, but that's their own business. It is absolutely outrageous. However, for America's two leading labs to ask our government to advocate global pacing, constraints be imposed on the entire section sector, calling for a global gatekeeper for AI that will cripple our entire nation's computational capabilities has obvious anti-competitive effects, especially for open source, which makes ongoing fears about regulatory capture all the more credible.

00:18:46:28 - 00:19:10:14
Unknown
I think he's spot on. Yeah, spot on that. Yeah. Cosign on all of that. Absolutely. I think he's great at this. And so I had the pleasure of meeting him at some conference once. So then so Sam Altman has also said that he wants to decelerate. Sam Altman is out there saying that the singularity has arrived, which I think is pure and utter, which I saw that I was like, oh, of course he did.

00:19:10:15 - 00:19:32:03
Unknown
Of course he brought this over to the singularity. And are you kidding me? Just in time for our IPO. We're already there. Yes, a defined it. It's it's a loose, loosey goosey term and be. No, this is the merging of human and computer brain. No, no, it hasn't happened. No, I do not believe that that is what we were looking at.

00:19:32:06 - 00:20:00:27
Unknown
So pretty far. Clip away from that. So wired did a story about how Silicon Valley is completely divided over Chinese AI. And I think open wait, AI is just a. Chinese AI is a placeholder in this argument for open wait AI because the open weights are coming from China. They're trying to say Chinese AI is dangerous. What they're really scared of is open wait, open source free AI.

00:20:01:00 - 00:20:32:25
Unknown
Yeah. There was I wish, I wish I had written it down once again, thinking back to the Nate interview that you had, where he put it, he put it into terms that had to do with with Napster and music that that really connected with me. I'm a very, you know, music driven person and from the perspective of, you know, music, whether it wanted to or not, suddenly found itself in a position where it it became free to all for the most part.

00:20:32:26 - 00:20:53:19
Unknown
Sure, we still pay for it, whatever. But the music paradigm before Napster versus after is very different. And because of the internet, because of that movement. Music, as I think, as he put it, want it to be free, whether that's true or not. But, you know, that's that's just kind of what ended up happening. And he kind of put it in terms of AI kind of going through a similar thing.

00:20:53:19 - 00:21:18:16
Unknown
We started we started off with this, oh my goodness, that's amazing technology. It can only be done in the cloud by these behemoth companies that know what they're doing. You know, we they're the ones that have the the capital to invest and to support and to build the infrastructure to do all this stuff. And now comes, come walking in the room comes the open wait and open models, which have been around for a while, but now.

00:21:18:16 - 00:21:42:07
Unknown
But what used to be the limiting factor is that open wait and open open source models tended to be pretty far behind what you can get out of current frontier models. And now that gap has shrunk a lot, or it's seemingly it has shrunk a lot to the point to where you get enough of them, enough out of them that it makes a whole lot more sense.

00:21:42:08 - 00:22:04:01
Unknown
Like if given the choice for me, if given the choice between paying for the rest of my life to a cloud provider, a subscription service to access a model that's capable of all these things, most of which I'll probably never need to touch because I'm just like running my business versus a model that does all those things that I control, I manage on my machine.

00:22:04:01 - 00:22:26:26
Unknown
Maybe it's, you know, over time it's more it's it's easier to kind of set up something like that. In that sense, maybe AI is going through the same transformation that music went through. Then AI wants to be free. And I'm fully behind the idea that everyone should have access to these tools, and not just the people who can afford it, not just the people who are deemed worthy of having it.

00:22:26:27 - 00:22:46:03
Unknown
Like, if these tools exist, everybody should have access. And I think that comes through open, open ways. I thought I thought it was a really good, really interesting framing from that interview that I hadn't thought about it in those terms, but that makes a lot of sense to me. So, comrade, you. Oh, there we are. Now we're in the communism.

00:22:46:06 - 00:23:04:21
Unknown
Yes. So last week we talked about Dean ball from OpenAI, wrote a piece that was just kind of shrill, screaming, oh my God, this is AI communism with open weight models. And so you and I talked about this last week and you kind of said, well, hold on a second. What's wrong with that? What's wrong with what you just said?

00:23:04:21 - 00:23:27:16
Unknown
That AI can be in everybody's hands. Yeah, and I thought about it more afterwards as I wrote this post. Nerds of the world unite. And responding to what ball said, I think it's pretty funny what he did, and I think he realized that he was a little shrill in it afterwards. But it's embarrassing because then turns around.

00:23:27:19 - 00:23:58:20
Unknown
OpenAI signed the letter that he was objecting to, so it was a little bit a little bit difficult there. But imagine all the good that comes from open wait and open source and locally run AI right now. Researchers and startups and entrepreneurs and creators can use it. Without this constant cost, we can have more control over it. There's more transparency, there's more competition in the market, which is a good thing, potentially privacy.

00:23:58:20 - 00:24:17:12
Unknown
You're not spraying exactly that, or trade secrets and business secrets and everything into a place that who knows if they're going to end up in Google search results. We'll get to a story on that later. Right. And so and the thing is, it's not as if you have to imagine any of that. It all exists today. You can do all these things today.

00:24:17:15 - 00:24:38:10
Unknown
The other thing I think it points out is the folly. And I've argued this often of the AI people trying to build the general machine, artificial general intelligence, the machine that can do everything. That's not where the value is going to be. As you just said it, Jason, the six months behind model can do 99% of what people need 99% of the time.

00:24:38:13 - 00:24:58:06
Unknown
There will be times when you want this more powerful model. That's fine and you'll be willing to pay for it. That's fine, but that's likely to be science and medicine and high finance and other areas that are going to want more powerful models. They're going to want system integration. They're going to want other services around it. They're going to want uptime guarantees and all of those kinds of things.

00:24:58:07 - 00:25:22:00
Unknown
And there's plenty of money to be had there. And I think we end up looking also at a WordPress model where WordPress was in a in a fight with what was a movable type. Right? Yeah, I believe so, yeah. And Movable Type had put out their model. There, there, there. What do I say anymore. It's not a model.

00:25:22:01 - 00:25:42:22
Unknown
They've not their software. And you could you could run a competitive host with it. But they limited it. They constrained it because they didn't want competitors to their own hosting business. Matt Mullen said no, the way to make this work is that we're going to make a WordPress open source. Everybody can use it. Anybody can run hosting with it, no limits whatsoever.

00:25:42:22 - 00:26:08:01
Unknown
But because it's open source, they have to contribute to the code pool. When they improve it, we all benefit. And I'm going to have the first mover advantage of building a company on top of that. So you can host your WordPress blog with WordPress.com or plenty of other places. And that turned out to be the winning model. And I think we end up with something similar here where the Giants, I don't think they're worth all the money that they think they're worth right now.

00:26:08:03 - 00:26:29:14
Unknown
But there's, there's there's a business there. They've they've been pursuing the wrong businesses. I think they've been pursuing. Yeah. And they're far down that path right now. And so hyperscalers we're going to do everything. One model for all I think that's so wrong. Whereas the smart people are going to specialized models. And this is where John LeCun is to and fail and others as well on the specialized model path.

00:26:29:14 - 00:26:53:14
Unknown
And Yann also argues that if you have a specialized model, it is constrained by its very nature and creation. This is all I do, right? And so you're more confident that you can limit what it does. And when I get my job done, I stop not trying to do everything right. So yeah, this is fast. And then Zuckerberg comes out.

00:26:53:16 - 00:27:14:01
Unknown
So so are you in Zuckerberg in the same room. I find myself agreeing with the Danvers people these days. I mean, green with Alex Karp at Palantir and Mark Zuckerberg. And it was an okay column he wrote for the Wall Street Journal defending open source. And there was some fear that he was going to retreat from this, that when John left meta, they pulled back lava.

00:27:14:03 - 00:27:41:22
Unknown
I forget lava was the primary open, open ish, open access, not open, not fully open. Wait model that was out there. So Zuckerberg came up with a rather clear endorsement of open weights. And then he makes his argument was not unlike ours, comrade, where he says that everyone should have access to superintelligence. Well, I'm going to quibble with this notion of superintelligence.

00:27:41:25 - 00:27:58:10
Unknown
And he has this weird argument that what happens if one person has the best virtual lawyer there is and everybody else doesn't have that? That's just not fair. So everybody should have the best lawyer AI lawyer they can have. What, so we can all sue each other? I think it's a little weird as a metaphor, but. Okay, fine.

00:27:58:13 - 00:28:23:07
Unknown
I wouldn't have that one. Yeah. One thing to look at. Maybe a teacher. You're the best teacher. Everybody else gets less teachers, I don't know. Yeah, but. So he's coming out this way now. And so they're all arguing the other piece, which is I forgot these two stories in your rundown rundown, is the argument that the frontier companies were making and OpenAI was it's no fair.

00:28:23:07 - 00:28:42:21
Unknown
They stole from us. They distilled us right around Kimi to me. Right. And so there's two arguments we've made here. One is, well, of course, they distilled the whole world. They distilled tons of books, they distilled all of our language online. So it's a little hard for them to argue that we're all evil if we're just doing what they're doing.

00:28:42:24 - 00:29:08:01
Unknown
What a point, B I'm really glad you found this story. Experts say that exploiting fable was not how community three K3 got so good. You want to talk about that? Well, I think I think they're just kind of you know, I think everybody is looking at this and getting clues around some of the outputs that they're seeing from Kimi, and understanding that it isn't always just about distillation.

00:29:08:02 - 00:29:36:08
Unknown
Right? They are kind of all moving in a similar direction. And yeah, I mean, it's it's it's interesting like the story from a year ago with, with Deep Sik, it's easy to kind of jump back into that pool and say, that's what's happening here. But I think just capabilities all around have improved tremendously in the last year. Kimi is just one example of that.

00:29:36:09 - 00:30:05:26
Unknown
I'm not sure I entirely understand how how you look at an output from a model and you determine other than like I think I had seen earlier reports of like output from Kimi identifying as Claude, and I don't know how specific or you know, how much that actually proves something around distillation or not, you know, because like, I've never distilled distilled a model, but I think, I think my understanding of this is that the capabilities are getting so good, it's just really not as necessary.

00:30:06:01 - 00:30:26:10
Unknown
It's a commodity. And and they have access to the hardware that they need to do what they're doing, whether the US is trying to limit that or not. You know, that's it kind of doesn't matter. So so as a societal battle, this is I'm going to start singing The Internationale in a minute. As a society, if we believe that intelligence is.

00:30:26:14 - 00:30:48:00
Unknown
Well, first I want to caveat I don't like describing this as intelligence. Sam Altman has this vision for kind of a vending machine where he sells you intelligence, and he says, I sell it to you as if it's electricity or water. I think that's anthropomorphic and competitive with human beings. And I don't think what you're buying is intelligence, per se, but I'll leave that aside.

00:30:48:00 - 00:31:17:21
Unknown
I'll use it for the sake of this argument. So if intelligence is a good that could improve our lives, our jobs, our education, our economies, all these things, then you and I, comrade, are kind of saying, yeah, everybody should have access to it. And Zuckerberg is saying, have access to it. And what's wrong with that? But there are other things that then fall out of that, which is that if we see it as a good, then if it's not a public good, is at least a public utility?

00:31:17:24 - 00:31:33:03
Unknown
Yes. What does that say about control of it? I don't think I don't want it to be like the internet as a public utility always bothered me, because that yields to control. In this case, we're saying open source is what makes it possible. It's not as if the government says there's one pipe and everybody has to have access to it.

00:31:33:04 - 00:32:01:18
Unknown
No, there's tons of open tons of I to be had. But but there's almost a I think we come to we came to a belief that people have a right to the internet in the UK by the year 2034 or maybe 2044, they're going to turn off broadcast towers for television there. There'll be no broadcast TV because their presumption is everyone will have broadband AI by then, and we can cut off this huge expense and we can use that spectrum for other things.

00:32:01:20 - 00:32:25:16
Unknown
So at that point, it's saying everyone has a right to broadband internet access. Will we get to the point that everyone has a right to AI intelligence? Yeah, I think that's I think that's kind of that that kind of ties in with how I feel like it should be. But but then we've got ozone nightmare, which by the way.

00:32:25:22 - 00:32:57:28
Unknown
Thank you for the super chat. Once again. You are always so incredibly supportive. We appreciate it. Said this a little bit earlier. I admire the idealistic idea of democratized AI, but I think it's a fantasy. It assumes a tremendous amount of financial privilege and technical knowledge, neither of which is widely available yet. For everyone to have access to AI, you also have to kind of have certain things that, you know, we might take for granted because we have access to the internet and we have a computer with resources and all this kind of thing, you're going to need somewhere, somehow.

00:32:58:00 - 00:33:16:06
Unknown
Yeah, not not everybody has that. I mean, it's kind of the same with the with the, the the TV over the air broadcast. Not everybody has access to the internet. We like to believe that that's true. But there is a digital divide of some sort. And so we are kind of all the more so here with absolutely nothing right to run.

00:33:16:06 - 00:33:40:19
Unknown
Kimmy K3 you can't do that on your laptop, man. Yeah. You have to have multiple high powered chips to do that. It's meant for corporations. None of us is going to run that, nor do we have the reasons to run it. But the fact that it can be open, a company can. It says a lot. Yeah. So so I don't think it becomes a public good in the sense that who's going to provide it for free, I think goes on nightmares.

00:33:40:19 - 00:34:10:19
Unknown
Exactly. Right. You need to presume more economic beneficence than exists or even can exist. But if we see it as an open public utility, that people come to be able to use it because the marketplace does that through openness, then the next thing that falls out to me is interest to me is, well, then, if we're all using this intelligence, we all think it's a good thing to use it in our control without the evil companies we've come to dislike.

00:34:10:19 - 00:34:37:01
Unknown
And we can do this with open source and open weights and all of that. Wonderful. And we can hug. And it's like Mozilla of AI and that's all great. Then the next part is, well then do you want stupid AI or do you want intelligent intelligence? Yeah. Right. Yeah. And so then we get into the questions about copyright and is it evil for open source, open wait AI to read these books so they can give us all free AI.

00:34:37:03 - 00:34:44:08
Unknown
That's a different equation from evil anthropic stole books.

00:34:44:10 - 00:35:04:28
Unknown
And then it gets it gets next. Beyond things like static media, it gets into things like journalism and updated media and so on and so forth. And so I think we get to a very different discussion about the information and intelligence ecosystems of society. This really starts to tear all that apart. That's why that's why this week is so, so beautifully complicated.

00:35:05:01 - 00:35:25:12
Unknown
Yeah. All of this stuff comes together to raise really big questions. Yes, indeed it does. It does. We got regulation to come up. And I'm going to guess you want to do a break. Yeah. Definitely need to jump into a break. Let's take a moment and recover. Recover from the 20 stories we've already talked about in the first block.

00:35:25:14 - 00:35:41:11
Unknown
Whatever. I don't even know what we're coming back out to. Come out to the kill switch. The. Should we be able to turn all of this off? How's that? Love that as a tease. All right. We're going to talk about kill switch. Coming up. Stand by. We'll be back in a moment.

00:35:41:13 - 00:36:03:07
Unknown
So, Jeff, should there be a kill switch? And if there is a kill switch, who should have their have that at their disposal? Is that President Trump? President Trump should have a button on his desk that says killed the AI. I don't like it, I got confused, was that the button to bring a Diet Coke, or is that the button to kill the I get which ones which what's the red button?

00:36:03:09 - 00:36:28:11
Unknown
So the hack Coca-Cola with the AI agent. Yeah. Ted lieu, who's a Democrat from California who I believe he represents Silicon Valley, and Nathaniel Moran, a Republican of Texas, had proposed the AI Kill Switch Act. This scares me mightily, but but this is what this is the fallout of the entire discussion we just had. Oh my God. AI is powerful.

00:36:28:11 - 00:36:49:18
Unknown
It can do all these horrible things. Look, it just went rogue on hugging face. Somebody has to be able to stop this. Something must be done. So it's a reflex from politicians to say, I'm going to fix that. I've got a bill. Okay. So the and companies that refuse to to go along with this can be fined $20 million a day.

00:36:49:20 - 00:37:13:08
Unknown
Here's the rub. The Department of Homeland Security would receive the unilateral power to order any of them to shut off. Wow. Yeah. So so, Mark Wayne Mullen or Kristi Noem before him would have had the power to say, no, no, no, we don't like that woke AI. I order you to shut it off because I'm going to say that it's a danger.

00:37:13:11 - 00:37:31:23
Unknown
Yeah, yeah. And it doesn't refer to everything. It only refers to the big companies so far. But it's not hard to see this now. Now, if that sets that. Oh, the big companies. This is regulatory capture okay. We've got to kill switch. Okay. We're we're the safe ones. Now that puts them in a position to demonize open. Wait AI all the more.

00:37:31:25 - 00:37:55:00
Unknown
All. It's a wild west. There's no control of them. You really should outlaw them. Yeah, you really should slow. All the bad stuff happens over there, right? Right. So that's the danger that lurks here. That's the. So. So you have the technical path, you have the societal path, and then you end up in the policy path. And that's where we are now.

00:37:55:00 - 00:38:28:02
Unknown
So we're going to see a lot of legislation like this trying to say danger, danger. I'm going to I'm going to save us from this horrible thing, and it's going to set these kind of precedents that I think are really dangerous. Yeah. This bill exempts incidents that would happen during red teaming and testing. Apparently it also, like you said, would only apply to a certain tier companies earning at least $500 million a year from the AI system built, by the way, $100 million worth of computing power, which right now is not very much.

00:38:28:05 - 00:38:47:00
Unknown
Yeah, hugging face would not meet that threshold. So it wouldn't have been a covered company if that were. But OpenAI would have been such that the kill switch would have hugging face would have screamed and yeah, that's true. Yeah, open would have hit the switch. If at what point did hugging Face know that it was open? They know that earlier.

00:38:47:05 - 00:39:07:07
Unknown
That's the other question. What are you turning off. Yeah. And then when you turn it off that something. What else are you turning off. Yeah. Yeah. So the irony of this is it only will increase the demand for open weight locally run models. Right. Because it's kind of like this hardware export to China. What what do you do?

00:39:07:08 - 00:39:38:21
Unknown
You end up giving them the fuel to to become self-sufficient. Look at people who even in a few days started to become dependent upon fable and boom, off it went. Poof. So yeah. Yeah. So this is there's going to be a lot of this. It's coming. And I think it's dangerous. Now, the Wall Street, the I'm sorry, what was the Washington Post editorial, which of course means in this case, I'm sure it's speaking with the voice of Jeff Bezos says the AI kill switch solves for the wrong problem.

00:39:38:23 - 00:40:02:18
Unknown
And I agree with him. I'm agreeing with all the weirdest people on earth, but this is in their self-interest. That's obvious. Yeah, but I also think that their self-interest in this case is not necessarily orthogonal. Orthogonal to ours. Okay, that was the first three blocks, all in one jumbled mess. And I apologize if I jumbled up folks.

00:40:02:19 - 00:40:21:11
Unknown
No, I'm I'm happy that you took the reins on that, to be honest, because coming into this show, I was like, I don't even know how I organize this. So, you know, sometimes you just got to throw the throw the, the playbook out the window and just go where it goes. That's the beauty of podcast. And that's the beauty of podcasting with someone who knows how to podcast.

00:40:21:11 - 00:40:43:08
Unknown
So thank you for taking the reins on that. Yeah. And then I feel like we've kind of talked about to some degree the, you know, the whole fallout around the hugging face thing. Like is there anything else left to kind of yeah, this is a little bit further here in that there's been more of a forensic analysis of it.

00:40:43:11 - 00:41:08:14
Unknown
Yeah. So hugging Face has gone on deeper into the technical kind of like what's happening behind the scenes. And I tried to understand, I tried to read through and understand some of that. And it gets really technical. Said don't entirely understand it. Right. And then there was a second firm that was also yes hit. That's right. Who was that?

00:41:08:14 - 00:41:39:11
Unknown
That was a modal. Modal. Right. So apparently OpenAI's models weren't just hammering at at hugging face. It was also modal, but it was modal via a customer of modal. So it was a customer kind of entry point. It didn't impact the product itself, but that was kind of the entryway that it was able to find. The customer itself left an endpoint that was exposed that didn't have any authentication.

00:41:39:12 - 00:42:05:12
Unknown
The agent somehow found that and used it. And the modal CTO has acknowledged that this happened. So there you go. So at least hugging face and modal at this point, I think we also know that OpenAI has said, I think a total of four different entities. So there's probably more to come here. Yep, yep. Yeah. We're probably going to learn what those are, you know, here in the coming days.

00:42:05:13 - 00:42:28:02
Unknown
And so as I was I was trying to read around this, I went to Politico and their coverage is revealing. I'm just going to read the headline in the lead from from this with With Danger voice here because they used, I think, a lot of emotional language. OpenAI's rogue models roamed the internet for four days and staged a second attack.

00:42:28:05 - 00:43:06:20
Unknown
A new analysis reveals that the artificial intelligence companies most powerful models spent days probing the open internet before breaching AI developer platform hugging face. In other words, it was Google search. It searched the internet. It was searching Intel, the lead, the powerful, the powerful. First artificial intelligence models from OpenAI that went rogue and mounted an unprecedented autonomous cyber attack earlier this month, spent more than four days loose on the internet orchestrating the hack, according to a new analysis from the platform that was breached.

00:43:06:23 - 00:43:26:17
Unknown
So I just I just have this vision of like AI as Bigfoot. Yeah. Lurking around. I just saw the Odyssey. It's the Cyclops. Yeah. Oh, I need takes up, takes up beast and eats the head off of them. Sorry, that was a spoiler, folks. But you know what Cyclops is do. Yeah, that's what Cyclops is. Do. That's what they do anyway.

00:43:26:18 - 00:43:46:07
Unknown
So that's all right. Yeah. By the way, I didn't like the movie, but that's a whole other story. Oh, really? You're like, the only person I've heard. I know I'm weird, but so. So my point here is that this falls out from the first two thirds of the show, is that this is what this is what falls out from media coverage, and then the policy falls out from that.

00:43:46:08 - 00:44:13:15
Unknown
This gives cover for a lot more bills to be made to say, oh my God, we have to stop these rogue attacks. But if you rewrite this in unemotional language, OpenAI's test search the internet and went to more than one company, that's pretty much it, right? That's what we do every day. And it was instructed to find a way to get this information.

00:44:13:15 - 00:44:41:05
Unknown
And it got the information. It didn't stop. So so then in saying that, which maybe this is more of a like a critique of media and like how things like this are written in saying that, is that like, is it are you saying that the choice of that language is done intentionally to or potentially is done intentionally to spur or to kind of motivate things like that?

00:44:41:06 - 00:44:59:23
Unknown
You know what I mean? Great question. Great question. Or is that just some writers write in flowery language versus stating just sheer fact? It's a it's a really insightful question, Jason, because I don't know the motive of the journalist or the editor. Yeah. I mean, yeah, I asked that knowing that neither of us actually knows what, but it's pure sensationalism, right?

00:44:59:24 - 00:45:20:09
Unknown
That's the more people to read your story. So you're going to use that kind of language. More people read a story. This is the attention economy, which was not invented with the internet. It was invented in 1893, says my book magazine, bye bye bye magazine publisher. The attention economy has been around through the entirety of mass media. So that's what that's number one is they want they want more clicks on their stories.

00:45:20:09 - 00:45:41:27
Unknown
So they do that. Then number two, however, is they act as if they're unaware of the impact of that. And the impact of that is it gives cover for politicians to join in there and what they're going to like that then, because they can come back and say, well, our story motivated politicians to take action, right? And so they're taking a stand here.

00:45:42:00 - 00:46:09:03
Unknown
They're saying this is dangerous and something must be done. Is the implied structure of those headlines? Yeah. Interesting. Very interesting. That's why I love doing this show with you, because you bring a knowledge of, of like the background in a, like the workings of, of the media that I just I'm not familiar with. I'm not close. You're the because you're the good replacement for it Jason.

00:46:09:05 - 00:46:34:08
Unknown
And I mean that I do podcasting it's different but all right from one dangerous thing to another. So let's see here. What do we have? We have Claude. We have. Yeah. Switching gears a little bit. Let me see. I've got so many different stories open. Can I even flip to a story that you were there. You were there.

00:46:34:09 - 00:47:01:02
Unknown
There you are. There we go, there we go. Okay, so apparently people have discovered that that some certain Claude conversations. There's a there's a feature inside of Claude. I don't think this is siloed to just Claude. I think you can do this in OpenAI. You can probably do it in all of them to some degree, where you can kind of share a link from a conversation to someone, and there's a kind of a share link feature, essentially, that that appears there.

00:47:01:02 - 00:47:28:27
Unknown
Well, people have discovered that apparently some of those shared links are showing up, indexed inside of Google search, because Google is just looking at the whole of everything that's on a given domain. So okay, now this stuff has a URL so it has a permalink. How so? Would Google know that? Because it's just because somebody shared that link out someplace that was published on the web and it caught that link.

00:47:29:01 - 00:47:58:15
Unknown
I don't know enough these days about how crawling works, but basically when Google looks at a site, it tries to crawl everything that is in that domain and all the deltas in it. So somehow it must have come up in a map a lot. Yeah. And and by the way, anthropic had a robots dot txt that was meant to block crawlers, but that only really block them from refreshing those shared pages.

00:47:58:18 - 00:48:31:06
Unknown
Google had already indexed the Earls from the links that people had shared and basically publicized essentially. So Google can't see an updated no index tag on a page. It's not revisiting because of the robust txt file. So that's sort of how this happened. So as a result, you know, like the like the headline of this, this article, you know, cryptocurrency wallet keys, login credentials, different things that you share into a conversation.

00:48:31:06 - 00:48:56:08
Unknown
And when you share that conversation with the share functionality, you might not think like, oh, by the way, I'm making this entire conversation accessible on a Google search. But apparently that's kind of what was happening. I don't know if it applied to all shared conversations or what, but well, the examples here in the tech story are are all Claude AI slash unique.

00:48:56:08 - 00:49:16:14
Unknown
ID got it. So it had access to that directory and once it was in that directory, guess what it found? Because you created a permalink because you wanted to share it with someone. And Claude nicely hosted it for you. Yeah. And Google. So as a result Google nice directed other people to it. And so this is not Google's fault.

00:49:16:17 - 00:49:45:21
Unknown
Google just searches the web. And yeah this is this is anthropic anthropic fault. Yeah. Thank you. Yeah. They didn't have things set up on their side to protect against this. And so is as ever going all the way back to the beginnings of social media. If you're going to put something on the internet. Be careful. Yeah. And if you're going to tell a hosted model I'm not post terribly paranoid about that, but there's some things I wouldn't share with.

00:49:45:21 - 00:49:51:20
Unknown
I wouldn't put my taxes, I wouldn't I wouldn't give a model my taxes.

00:49:51:23 - 00:50:11:05
Unknown
Right. Whoa, wait a minute, Jeff. Tell me more. I need to know right now. I need to know right now. I'm just kidding. I don't know that I've ever given a model, but I've given. I just spent a good a lot of health stuff, but that's okay. Okay with that? Yeah. Well. And, you know, see earlier conversation about open, open local models.

00:50:11:06 - 00:50:31:01
Unknown
Right. This is this is one of those reasons that like, you know, maybe, maybe you do want to feed your taxes through a model and have it evaluate and analyze and, and help you with that. Or maybe you want a host that's not one of these big companies, but is the the equivalent of Mastodon or. Yeah, some nice guys you trust.

00:50:31:03 - 00:50:52:05
Unknown
Yeah. When I, when I was going through health stuff with my wife last week, which I mentioned in the last show, I didn't share a URL, I cut and paste the text an emailed that to her separately so that it remained just a chat that they say is between us. I was like, I was I was thinking about that as I was reading through this.

00:50:52:07 - 00:51:12:03
Unknown
This story is like, I don't know, like I probably have at some point shared once or twice, but probably not more than I could count on one hand. Like that's just not a function that I use these things for. So much of the work that I do inside of these chats is for me, so just kind of stays with me.

00:51:12:03 - 00:51:26:09
Unknown
I don't do a whole lot of sharing outside of it, and if I do, yeah, it might be grabbing like a screen capture of a particular point, or I might copy and paste it or whatever, but I, I don't share the URL and be like, here's the entire conversation, because I feel like that would be kind of confusing.

00:51:26:09 - 00:51:56:23
Unknown
It's like, hey, this, you know, here's this massive amount of text for the thing that I want you to know from it. You know, I don't know, not super useful to me, but but they surfaced it as an option. Plenty of people used it. Apparently this has been anthropic, has confirmed that that gap has has been kind of undone I think so I think I think things are good now, but something to know if you don't own it.

00:51:56:23 - 00:52:36:23
Unknown
There's always the risk, right? If you're putting it out there, what else do we have. So then we have you had. Yeah. So you included some research, which I think is really interesting here. It's it's all about AI and well, big surprise. But how it's, as it says, reshaping or and transforming the kind of how people work and not necessarily looking at the thing that we often talk about, which is like how, you know, AI is impacting jobs and that people are being let go because the models can do the job for them or something like that.

00:52:36:23 - 00:53:03:02
Unknown
But in this case, how people are using AI to become solo workers, entrepreneurs, self-employed, and I don't know, this piqued my interest because this is here you are use AI. I mean, this is this is it. You know, it allows you it allows me to do so much of my business that it would have taken me a very long time to learn if I didn't have access to these tools, and now I rely on them.

00:53:03:02 - 00:53:09:26
Unknown
So. So the economist Ronald Coase famously wrote, when did he write this?

00:53:09:29 - 00:53:32:27
Unknown
A an argument about about the firm and that there came a point in the economy. He got the Nobel Prize for this in 1991, that when tasks became more effective, efficient to do inside a firm versus outside. And so that's when you saw the growth of corporations, because it was a lot more sensible to do things as employees.

00:53:33:00 - 00:53:52:29
Unknown
Right now. I think what's fascinating is AI finally potentially just turns that around, where rather than saying you had to darn all these socks, you, you, you weave, you wove the socks at home and there were talks of home workers. And then with the loom, it became more effective and efficient to do it in the factory. And you were hired there to do it.

00:53:53:03 - 00:54:26:09
Unknown
And now we're back to darning the socks at home, and we're not fully there. Right. The paper says that this is from the Mercatus Center by the polygon. We screwed that up, and I apologize. The result does not mean that AI is replacing the firm, or the traditional employment is disappearing. Rather, in some forms of knowledge work, AI may be reducing the advantages of being inside a firm and increasing the feasibility of independent work, solo self-employment, freelancing and contract based work.

00:54:26:12 - 00:54:52:00
Unknown
This means that institutions built around a single employer, such as benefits, will become increasingly mismatched to how people actually work. I never asked you what you have to do about health insurance, but I can't imagine it's a pain, right? Oh it's horrible, it's awful. Right? The results point toward portable benefits that travel with workers and modernized income security systems, such as worker owned portable accounts built for workers whose earnings come from clients and contracts rather than a single employer.

00:54:52:02 - 00:55:16:02
Unknown
Know I've always wondered why companies fight so hard. There's a whole other discussion against single payer insurance just because. Why does a company want to be involved in health care right now? Yeah, I was going to be would gladly get out of it. And I think that's where we, you know, we potentially go because because people are are trapped in jobs because of these benefits, because of health and retirement and security.

00:55:16:02 - 00:55:46:08
Unknown
And so, yeah, I don't the fear has been that AI destroys jobs. What this argues is that AI potentially creates many new ways to make a living. Yeah. For for oneself gives people the tools to detach themselves from the idea that the, the way that you work is for someone else and connect to the kind of the solo self self-employed independent option as something that's far more possible.

00:55:46:13 - 00:56:09:09
Unknown
Yeah. Now having said that, and it's easy for me to say that because I've always had a salary. You are living this life. Yeah, and I taught entrepreneurship in a journalism school. I created a program to do that. And it's hard. Don't you know? Right. It's really hard. It's risky. It is very hard. Scary. Yeah. You've got to take on all these skills you never thought you'd have to do, like, sales, right?

00:56:09:12 - 00:56:34:25
Unknown
Yeah. And, you know, and that is something that the tools help me with because I can collaborate in air quotes with the tools when I have questions. It's not just me doing a Google search, but it's like, here's my situation. Oh, I didn't think about doing it that way. Well, what if this and, you know, it's like I'm kind of batting batting ideas back and forth until I find the thing that I think works for me.

00:56:35:02 - 00:56:56:05
Unknown
That's really nice to have that tool, you know, like it's really important here not to get utopian about this and say, oh, people can make their own companies and that's just fine. No, no, you're absolutely right. It's a long because it also carries a lot of risk. Risk exactly. Uncertainty. And but as you just said, the AI can help you with that more than before.

00:56:56:07 - 00:57:17:12
Unknown
I'm I'm incredibly grateful that I have these tools doing what I'm doing. I wonder if I'd still be doing what I'm doing if I didn't have these tools. You know, like I've only been doing this now on an independent basis for about two and a half years, which I'm proud of. I'm like, oh, well, then something is sort of working because I'm like, I'm still doing it.

00:57:17:12 - 00:57:36:23
Unknown
I wasn't pushed back into, you know, working for someone else at the end of the day. But there's still a lot of room for improvement, and it's still very stressful and very difficult. But thank thankfully, tools like these exist because they really are helpful to me. And it's been really interesting for me to watch and learn as you have advanced in this.

00:57:36:26 - 00:58:01:23
Unknown
You used AI tools going back in the day to do rundown or. Yeah, that kind of stuff. It was a fairly basic stuff. You become much more sophisticated about your use of AI at the core of your businesses and businesses, for sure. You know, I would say, what do you think out of the you've been gone for two and a half years from employment, maybe the last year and a half is when it really kind of in my perception, it started coming up or before that or after that.

00:58:01:25 - 00:58:29:01
Unknown
I would say I've been using the tools to some degree or another throughout my entire, you know, to two and a half years of independence. I would say the last year has been a real, real escalation as far as my understanding of, oh, now the now the the pieces are falling into place. Now I'm getting it. Now I understand why this versus that, why or what it can do for me or kind of advanced systems that I can build around it.

00:58:29:02 - 00:58:47:06
Unknown
Oh, if I, if I ask, you know, if I create a folder structure that's like this and I create these types of files and put them over there, then then I don't have to keep repeating myself every time I do this task or whatever, you know, things like that, that a year and a half ago, I didn't know how to do that, and that would have felt confusing to me.

00:58:47:06 - 00:59:07:19
Unknown
So yeah, the more you work with these things, the more those things become possible. And maybe that means the more people you know, like me, can actually run their business on these things. But I also run into situations where I'm like, but these systems could be so much better for this. And, you know, I find myself time and time again going, this is really great.

00:59:07:19 - 00:59:27:25
Unknown
It's like 70%, let's say 65, 70% of what I want. If I could figure out the 30% that's missing and package that up, I'd be a millionaire, you know what I mean? Because I know other people are in a similar situation as me going, yeah, it's great, but it's not perfect. And and that's another opportunity. I mean, that is another opportunity.

00:59:27:26 - 00:59:49:07
Unknown
Absolutely. You started off saying, I'm going to create a podcast company, and you've also become a podcast consultant for other people to help them create their podcast companies. Yeah. And part of that is how they use AI. And yeah, it starts to feed on itself. Totally. Yeah. Yes, indeed. Thank you. Jeff. Let's take a quick break then come back and do a little speed round.

00:59:49:07 - 00:59:54:06
Unknown
We got a few stories on the other side of the jump, so hang tight. We'll get there in a second.

00:59:54:09 - 01:00:29:16
Unknown
All right a few quick stories here Amazon narrowing its own AI model lineup. So basically getting giving the ax to a few of the Nova models. Premier omni canvas real are the the models listed in this article anyways Nova AI models that are that are going out there making room for some new frontier models unnamed. But yeah, so I guess Amazon's just kind of changing the AI that they're offering, changing it up.

01:00:29:17 - 01:00:52:13
Unknown
I don't know how often they do that, but there you go. Yeah. We see these. You know who's going to make the fortune? Not in being AI but in serving AI. We met as leasing out data centers. Musk is doing the same thing. Amazon is obviously at that business already with AWS. I didn't put the story in the rundown, but I think it was Jassy.

01:00:52:15 - 01:01:12:17
Unknown
He also said that making their unique chips is going to be another no. Bezos said this another leg to the Amazon stool is going to be their chips that they make. So that's really interesting is who who decides where their real value is in this market. So I was pulling back in some areas and investing in others.

01:01:12:19 - 01:01:31:16
Unknown
Yeah indeed. Also quick aside, I had not seen the price of bitcoin in a very long time. Wow. It's low. Yeah I don't know how I, I don't really care. I don't really follow crypto. I don't have any crypto holdings or anything like that. But wow it's like half of what it was not got about. I've got about five grand.

01:01:31:17 - 01:01:48:19
Unknown
And through a weird story, it was a company called Civic or Civil. It was supposed to do news and it was all going to be Bitcoin or oh yeah. Yes, I remember you talked about that. Yeah. And so to be a good citizen I had to buy the damn coins and then put it in. Then I got a refund of it.

01:01:48:19 - 01:02:07:28
Unknown
And so I still stuck with some crypto, which I just, I just, it just sits there, it goes up, you just goes up, but it goes down. Yeah. I think it was I think it was $500 worth a few thousand dollars worth now. Fun to watch it. Bibimbap and all that. Yeah, yeah. Once upon a time I had I.

01:02:08:00 - 01:02:32:15
Unknown
This is such an aside, but once upon a time probably like 2017 ish, I was like, I'm just going to put in like a recurring by for like $50 a month for Bitcoin just because I wanted to understand how bitcoin worked. It helps and I wanted to actually care about it, you know, a little bit more. And over the course of like four years, you know, it was $50 a month.

01:02:32:15 - 01:02:50:21
Unknown
At some point I increased it to 100 a month, and I just kind of forgot about it. And at some point it had built up, not to an insane amount, but, you know, something there. And as we know, Bitcoin continued to rise, continued to rise. And so I was like, oh my goodness, this small investment is actually turning into something significant.

01:02:50:21 - 01:03:10:11
Unknown
And then it started to dip. It started to dip. And and it finally got down to the point of, of equilibrium with what I had put in. And I had to ask myself, like, now if I just get out now, then I've not lost, I've not gained. And so I decided to just cash out. And I didn't make anything, I didn't lose anything.

01:03:10:11 - 01:03:30:29
Unknown
And then of course, it went back up and went back up. And I was like, damn it. This is this is why I don't do stocks. This is why I don't do crypto. I will, you know, this is why I don't gamble. I will guarantee you I will lose money. So if I don't do it anyway MCP model context protocol.

01:03:30:29 - 01:04:00:17
Unknown
We talk about this a lot. This is really you know, the has has largely become the architecture and the mechanism that AI agent interoperability is managed through originally created by anthropic but now kind of more of like an open standard sort of thing. Getting a big update. And yeah, it's it's really my understanding of this is that it's directed more at kind of enterprise level accessibility.

01:04:00:18 - 01:04:24:21
Unknown
Let's say it's a, it's a has something to do with this being less tied to the state of something or less tied to maintaining the session with a specific server instance. There you go. There you go in the case. Right. And so now that you know, more enterprises can can trust, I guess working with model context protocol because of that, that change.

01:04:24:22 - 01:04:44:15
Unknown
So also a very nerdy stuff. But it's good. It is flat MPC, MPC, MCP exists and I'm glad it's getting improved under an open structure. Yeah, some of the some of this was a little over my head, but like I wanted to put it in there because like I use I interact with NCP all the time with the tools that I'm using.

01:04:44:15 - 01:05:07:25
Unknown
And so I, I witness as a user the importance of what MCP brings from an interoperability standpoint to all these platforms. And so, you know, if it's one of the biggest updates that's ever had, probably good to mention it. But it's not I don't think the updates are necessarily meant for me, random user of these products. It's just it's a good thing about AI.

01:05:07:28 - 01:05:38:07
Unknown
Yeah, exactly. Meta launched a new Facebook Verified badge. It uses facial recognition. Sit like a selfie check to confirm you are an actual human and you are who you say you are. And it's explicit. Aim here is because there's a lot of AI generated fake profiles that apparently running a bunch of scams, automated scams at a large scale on the network.

01:05:38:07 - 01:06:01:21
Unknown
And so this is one way to get around that and to battle that it's rolling out starting next week. Oh, sorry. Next week because I haven't seen it on my Facebook. The way it works is, is that it's going to look at all the photos you already have of yourself or exist of yourself, and have you do kind of mugshot up and down, left and right, and then verify that you who's doing this now is the same person as in all those photos.

01:06:01:22 - 01:06:25:07
Unknown
Right. Well, yeah. And that reminds me of the news from last week of Google doing the selfie video for account access. Right. Something similar where you could store a video of your selfie kind of looking around, and then if you have it set up in advance and you forget your password or you don't have access to the device with your passkey, you can use your selfie and it'll match that up.

01:06:25:10 - 01:06:52:25
Unknown
You know, I have visions of Weekend at Bernie's moves, right? Oh, yes. And then finally, smart glasses. Not having a great month. Let's just say that a Comic-Con promoter in the UK banned meta glasses R-Truth at basically saying like, no, we don't. You know, it changes the energy of the room. And so we don't want those glasses coming to Comic-Con.

01:06:52:25 - 01:07:28:23
Unknown
So there you go. You can't wear them. If you're wearing them, you can't go in. They'll look for those little cameras on your glasses. There's a Scottish ferry operator, CalMac, that paused public bridge visits entirely because someone was secretly filming covertly filming a restricted, restricted area because and they were wearing a pair of medical asses. Instagram now banning accounts that post, this content around filming people with with them not knowing like secret you know, videos of people while they're wearing the glasses and everything.

01:07:28:23 - 01:07:56:03
Unknown
So covert recording and everything. So a little bit of a backlash coming. And does this remind you of Google Glass? Because I remember this was a Google Glass, too. In a way, it's worse because people just complained about it or said they were afraid of it, but so few people actually had Google Glass. Yeah, yeah. You know, if you were at IO and and Nick Bilton, the former New York Times columnist, now the executive producer of 60 minutes, he was all worried about about Robert Scoble having Google Glass on.

01:07:56:04 - 01:08:17:15
Unknown
He was one of very few people on Earth, whoever ran into anybody who was wearing Google Glass. And now with medical selling, well, it's more prevalent. Yeah, it is, it is indeed. You see a lot more of them. Samsung last week kind of announcing their own AI glasses as we talked about, Google has its AI glasses on the on the horizon.

01:08:17:16 - 01:08:40:05
Unknown
Actually, there was also news that Apple is delaying its smart glasses because of this backlash. That's kind of happening right now. So as a product category, there's some real kind of question marks happening right now. Like does it continue to be a target for the AI backlash as a whole? You know, kind of feels that way right now.

01:08:40:07 - 01:08:59:04
Unknown
Okay. I think that's it. I think we got through it all for the most part. We probably missed a few stories here. Oh, I think we hit them all somehow. That's amazing. Thank you for doing an expert job of walking us through so much of that, because I wasn't sure how to tackle it, so you nailed it.

01:08:59:06 - 01:09:24:19
Unknown
Jeff Jarvis, for everything related to Jeff, you know, it's only a couple of weeks now and hot types finally going to be out. You can still preorder your copy. Just go to Jeff Jarvis. Com and you can find all of his amazing books Gutenberg parenthesis magazine the web, we we've and yes pre-ordering hot type. And then of course wait patiently for AI and humanity, intelligence AI and humanity from Bloomsbury.

01:09:24:22 - 01:09:46:04
Unknown
As for me, as you mentioned earlier, which thank you for that. I do consult on podcasting because I've done it long enough to like, have some opinions and thoughts and stuff, but experience and expertise. There you go. Pod tune up. That's where you go. Pod tune up and hit me up. I'll help you. And then AI inside Show for all of the episodes.

01:09:46:04 - 01:10:11:21
Unknown
Everything that we do in this podcast, including our YouTube worthy thumbnails, what can I say? You do what you got to do to get the people to watch and listen, but everything that you need is there, including a little link to the store if you want to buy some merchandise. And finally Patreon.com Inside show. Hey, if you go there, you can get access to AI inside daily.

01:10:11:25 - 01:10:33:24
Unknown
I'm just realizing a few things. I'll talk about it in a second. AI Inside Daily, which is a daily podcast. You also at a certain tier, we'll get your name thrown out. Doctor do Jeffrey Marikina Radio Asheville 103.7 Dante. Saint. James. Bono. Derek. Jason. Jason. Brady, Anthony. Downs, Mark. Starker, Mark. Starker and Karsten. Usually that just rolls off the tongue.

01:10:33:25 - 01:10:54:09
Unknown
For some reason, I was a little distracted because I also realized that we had a new patron that I didn't call out in the middle of the show. Oh, that's right, Scott Hepburn. Thank you, thank you, thank you Scott. So, so many people helping us do what we do because of the Patreon Patreon.com inside show go their support.

01:10:54:10 - 01:11:16:17
Unknown
We can't thank you enough and I can't thank you enough. Jeff. You're awesome. Thanks for hanging out with me once again. While we were doing this show, I looked down and I realized we're at 140 episodes. We've done 140 of these. That's a that's a good clip. I mean, it's it's no this week in Google intelligent machines. But you know hundred and 40 is not nothing will catch up someday.

01:11:16:19 - 01:11:34:23
Unknown
Someday maybe I think we're always going to be behind but that's okay. That's just like an open weight model versus a frontier, you know? Yes. That's exactly it. That's exactly it. All right. Thank you Jeff. Thank you everybody for watching and listening. We'll see you next time on AI inside. Take care. Oh and thank you Daniel. Thank you Victor for all the behind the scenes work.

01:11:34:23 - 01:11:36:16
Unknown
Okay. Enough. Thank you. Goodbye.