The AI Race Just Moved the Goal Posts

00:00:00:09 - 00:00:27:28
Unknown
Coming up next, Jeff Jarvis and I dig into Zuckerberg's open source AI manifesto, although we decide it's probably not much of a manifesto. Along with Meta's open weight model that you can run on your own laptop, apparently anthropic watermarking everything Claude makes and is watermarking necessary? We talk about that Google's DeepMind shakeup and what it actually means about Google's AI efforts.

00:00:27:28 - 00:01:07:21
Unknown
Are they still a frontier AI lab? We got the pixel launch, the pixel 11 launch smartphone coming out from Google this morning. I was waiting for delivery. It stalls income yet, and a trick that apparently reveals what AI models are actually thinking and whether it's really necessary at this point. Do we care about distillation and those kinds of techniques while we talk about all of that and so much more on this episode of the AI Inside podcast?

00:01:07:24 - 00:01:21:18
Unknown
To AI. Inside the show, we take a look at the AI that is layered throughout the world of technology. I'm one of your host, Jason Howell, and I'm joined by author of the new book Hot Type, Jeff Jarvis.

00:01:21:20 - 00:01:43:29
Unknown
It's back there somewhere. Oh, I got to show you. There we go. There we go. It is a real book. And thank you, Vanna White, for showing that off and got a call out in our live stream. A friend of the show and man behind the scenes who helps us with some of our social video. Daniel Croft says your book is due for delivery during this show.

00:01:44:01 - 00:01:59:29
Unknown
Yay! Thank you, thank you, thank you Daniel. You're so helpful in every way. I really appreciate it. I know that's so. I was just thinking, Daniel, for being here every week to greet us. And then I saw, in addition to that, he bought the book. He bought the book. And he's going to get it during this show. So, you know, he'll he'll probably be doing an unboxing.

00:02:00:02 - 00:02:28:07
Unknown
We'll be talking about it. It'll be very. Netta, a review coming soon. I'm also awaiting a delivery. I should be getting a delivery any minute. Like it said this morning. So literally, it could come through while we're doing this show that I might get some of the new pixel 11 devices. Whoa whoa whoa. So we'll talk about that later in the show because it's definitely in there some some AI related news to the absolute.

00:02:28:11 - 00:02:44:07
Unknown
But I mean, honestly, from the outside looking in, not having the devices yet, it feels pretty iterative as they just do one thing that I came across and said, wow, when I went to put it in the rundown and you already had it there, so we'll talk about that. Okay. All right. Well, we'll talk about it before we talk about the end.

00:02:44:07 - 00:03:01:06
Unknown
And by the way, just a little tease. I have the Pebble index, which we're going to talk about a little bit later too, because I haven't even unboxed that thing. And I'm actually legitimately excited about this thing as well, is that show me the box again. That's the box you're unboxed that it's got a new word for unboxing that it's kind of yeah.

00:03:01:07 - 00:03:21:12
Unknown
You know, maybe what I'll do is I'll go super close up. So it looks really big, you know. But yeah, that's that's the box that it came in. We'll talk about the Pebble index and what it is and why I'm excited about it. But first let's talk about Mark Zuckerberg because isn't every show better when you start with Mark Zuckerberg.

00:03:21:14 - 00:03:41:01
Unknown
Well, we could you know, one thing we didn't include in here, as he was doing martial arts on a barge. Oh, I missed this. Oh, yeah. Full kicking and the whole thing with some big, you know, beefy guy. And there's there's dweeb Mark trying to undress himself. It's just hilarious. Yeah, he's he's been going through the last couple of years especially.

00:03:41:01 - 00:04:00:00
Unknown
He's been going through an image makeover, a lot of testosterone and be there. Yeah, yeah. Because for, for a long time and many would say still very, you know, kind of a stodgy kind of robot. I mean, it almost looks like a robot in a lot of ways. Robot. He's nerdy. And that's okay because nerds controls now, so it's fine.

00:04:00:02 - 00:04:23:08
Unknown
Don't be apologetic about it. There's nothing, nothing wrong with being being nerdy. I've been nerdy all my life. And only only for parts of my life that I feel weird about it, but now I embrace it. Mark Zuckerberg going in the opposite direction. He's like, no more nerd for me. I must be, you know, I to say, the cage fight between Zuckerberg and Bezos, who's trainer is better.

00:04:23:13 - 00:04:42:07
Unknown
You know what, I wonder about that. If that was a pay per view, would like, would would the actual viewership be an impressive number or would people be like, I don't even want to. I don't even care. Yes, it would be interesting. It's like, it's like f Mary kill. You know, like, who you rooting against in this case?

00:04:42:08 - 00:04:59:28
Unknown
Yes, exactly. Not who you rooting for. It's like, who do you want to get more beat up? And it's like both of them. I don't really care who wins ourselves. We're just projecting amount of others. I'm assuming because of the AI backlash, which we will also talk about later in this show. See, it's all it, all all feeds it.

00:05:00:00 - 00:05:28:18
Unknown
It's all it's it's circular financing but in podcast form. Another topic we will talk about later in this show. But let's start with Mark Zuckerberg Publishing. Is this a manifesto? Does this qualify as a manifesto? I don't think so because I don't think it I don't think it said, damn the world we're taking over. It was more a hubristic offering to the world, in his view, that this is this is it was extremely utopian.

00:05:28:21 - 00:05:45:25
Unknown
And the things that AI is going to do for us, and thus everyone should have it right. And so it was more of an offering. I think he thinks that this is what we can do for you, world. And does the world want it? Does the world trust them to give it to them? I don't disagree with much of what he said.

00:05:45:26 - 00:06:14:26
Unknown
I mean, that's the interesting thing here. So it's called the futures for everyone. And you know in in this scribe not manifesto offering, as Jeff puts it, he says that AI should be open. It should be available for everyone, as the title implies, i.e. not locked inside of some frontier companies access, you know, kind of barricades with API access, charge and arm and a leg for.

00:06:14:29 - 00:06:53:23
Unknown
He says that meta is committing to over $1 billion to fund data centers for open source AI research. Well, they're going to stick by that billion dollars. He's also doing a fund to make nice with the communities where they put. Exactly. He wants to build it for the people. And that that also includes like from the you know if data centers are required for this kind of outlay of this technology that is built for the people, then we also have to respect the fact that there are a lot of people that don't like data centers is a two letter word to them right now that I don't know.

00:06:53:24 - 00:07:14:29
Unknown
I think he's got an uphill, uphill battle there to to win over the minds and hearts of people on the data center front. But it's not a bad thing to to aspire to that, you know, to do it. No. And the great news here is that he is still supporting open. Now let's go with what we say on this show all the time.

00:07:14:29 - 00:07:41:07
Unknown
It's open ish, open ish. It's open waits. It's not fully open source. You don't see the training material and so on and so forth. But nonetheless, this is really important because it's exactly. Somebody was moving away or wasn't sure whether he was going to still support it. Yann LeCun has been praising him like crazy, like Yoda to Luke, saying, you know, I can't quote the lines from the movie, but it would sound something like this.

00:07:41:07 - 00:07:59:14
Unknown
But I don't know what the words would or I'll, I'll mix my movies. Very good, grasshopper. Right. So now we're back at Karate Kid. That's okay. Well, I think it's right for martial arts. I think it's probably better. So as I mentioned in a second, they're putting out an open model, and it's an important kind of model to put out.

00:07:59:14 - 00:08:23:23
Unknown
And so holding by that I think is critical. And it's a moment for open weights and a moment for people being able to have II in their own control. And it should make OpenAI and anthropic very nervous, because it's not just the Chinese who are waving the open flag, it's Zuckerberg. I'd be curious to see what Google does going forward.

00:08:23:25 - 00:08:48:01
Unknown
They've had models as well, and videos will put out an open a new open model bigger than this. So I yeah, I think it's important. The essay is 6500 words. Yeah. I mean that's that's why it's easy to jump to manifesto because I think when I think of manifesto, I think of this really, really long like detailed like this is how I feel about the world and everything.

00:08:48:05 - 00:09:13:03
Unknown
I will confess that I took advantage of the listen to article link, and I listened to it on my morning constitutional, where I walked very slowly now. But I'm walking good. There's a fair amount of sophomore. I'm not trying to sound so insulting. I don't mean it to be so insulting, but I mean that literally. As in left school after a sophomore year and still sees that as kind of the essay where there's a lot of word solid in here.

00:09:13:05 - 00:09:34:00
Unknown
Got it, got it. Nonetheless. Do you think he used AI, his AI, to write it? I would have suspect so, though. I just saw somebody. Is it watermarked? TikTok. But I don't know which which one they use. But they put Moby Dick through a detector and found that 12% of it was AI. Written in 20%. Was AI helped?

00:09:34:06 - 00:09:58:01
Unknown
Oh that's perfect. So I don't trust those damn things. But yeah, I would imagine he certainly did. And staff helped to you know. Yeah. Yeah of course. But he believes that you're going to have this extremely capable, exceptionally capable personal agent that understands you, your goals and everything you care about at your side. And that meta is going to give you this, and you're going to be grateful to better for it.

00:09:58:03 - 00:10:19:09
Unknown
You'll have incredible tools for the creation, for creation to express your ideas. I agree with that. And powerful tools to create new businesses. The Mercatus Center just put out a report a week ago saying that the impact of AI so far isn't losing jobs, it's people creating new companies. You'll have a personalized. I would agree with that. Yeah, I think so.

00:10:19:17 - 00:10:38:17
Unknown
You'll have a personalized tutor and coach with a PhD in every subject. And this is as someone who has tried and failed many times to learn languages, I rolled my eyes at this a little bit because it's not so easy as saying, oh, I got my coach, my coach going to beat me up every day to make me remember this stuff.

00:10:38:17 - 00:10:59:27
Unknown
I cannot decline German cases to save my soul, and having the best I in the world isn't going to make me do it. Learning still requires effort on your own part. And so I think a lot of this is reductive to learning and saying, oh yeah, yeah, yeah, you'll have this agent, and this agent will teach you anything you want to know.

00:11:00:00 - 00:11:22:11
Unknown
To an extent that can be true. Yes. But it still requires all the effort of learning, especially languages, and especially along that line, like motivation isn't. I don't get motivated when I work with an AI and it tells me something positive, like, I don't read that and go, oh man, I am pressed up to my yeah, let's go, you know what I mean?

00:11:22:11 - 00:11:50:09
Unknown
Versus working with like a mentor or working with a tutor or an instructor who says the right thing in a very human way, that connects with me on a human level to make me go, you know what? I want to be better at this because that's really interesting. Jason, what if AI when it came out as we when historians look back, what if it had no sycophancy but used that in the rare moment to say, Jason, you really knocked this out of the park.

00:11:50:11 - 00:12:14:05
Unknown
It would have to choose. It would have to choose so wisely it would, otherwise it would be meaningful. Then they completely devalued praise they have. If everything is praised, then nothing is praise. Yes. You know what's interesting? The University of Michigan, I think it is. I think it is, is eliminating grading for the first semester for students. And I know we're going to people say, oh, this is the ruin of nobody's tough anymore.

00:12:14:06 - 00:12:34:26
Unknown
They can't say. Kathy Davidson, who's a wonderful educator I used to work with at Cuny, posted saying, this is great. It's going to mean that students can experiment more. They can try things they wouldn't have tried otherwise. Then they can come to tough. Let me see. Make sure school ending grades first term. Which one is this? University of Michigan.

00:12:34:26 - 00:12:55:14
Unknown
Yes. So it's pass or no credit. So it's not even fail pass or no credit in that first term. Now I work with some really good educators who say that without grades you don't motivate people and they're not going to do stuff. And I think that's that's a problem to get over. I agree, I'm not I'm not saying that it works for everything.

00:12:55:15 - 00:13:17:12
Unknown
I hated giving grades because I figured somebody either is motivated to learn this or they're not. I talk grad school. Yeah. So anyway, at that level, yeah, hopefully you're not. If you don't know why you're here, then don't bother. Yeah. So so I think you raise a really good point, Jason, about that notion of of motivation. An AI is not motivating.

00:13:17:12 - 00:13:42:25
Unknown
I think that's a really important insight. Yeah. Yeah, it's it's kind of a flat like I don't know, people want to ascribe it as being, you know, kind of so similar to being human. But there's, there's a lot of ways in which it does fall flat. And I think that's one big one is, you know, I might be motivated by the idea that I have access to this insane knowledge, you know, tool set and everything like that.

00:13:42:26 - 00:14:00:15
Unknown
But I don't know that I've ever read praise from AI and taken it to heart. It's always with an eye roll, you know what I mean? And so, yeah, there's that. Well, you know the other thing, I have the privilege right now of having access to university libraries because I'm a married to status of Cuny. So I keep that.

00:14:00:15 - 00:14:18:08
Unknown
And plus my Stony Brook and Montclair State give me access to those libraries. And that encourages me to access to New York Public Library. So I have a sense there to a privilege. Yeah, I wish it weren't the case. I wish everybody could get what I can get, but I can get stuff that I feel privileged to have.

00:14:18:11 - 00:14:38:07
Unknown
AI kind of takes that sense away to it. I don't think that's necessarily a bad thing. But in your point about motivation, there's kind of a you don't you want to shame yourself into making good use of of privilege. You have. And so AI makes it because 1.2 is that everyone will have free or affordable access to these tools.

00:14:38:07 - 00:15:02:17
Unknown
And I think that's right. I think it's the way it should be. But in a weird way, it devalues them. Yeah, yeah, yeah. No, I totally agree. That is the way that that it should be. But yeah, you're right. What about the what about the the whiplash that MGC glare on Spyglass was kind of writing about that that, you know, we've we've been here before.

00:15:02:18 - 00:15:28:21
Unknown
Right. Mehta went open with Lama then you know, did a bunch of hiring a bunch of people from scale AI seem to be moving closer back over to the closing of things. Now, swinging back to open again is, you know, is this is this opportunistic? You know. Yeah, I think I think it's worth looking at any of these tech guys and wondering what their strategy is and where they're headed.

00:15:28:29 - 00:15:54:17
Unknown
But I think the important thing money where mouth is in this case is that meta is releasing muse glimmer. Yes. Which is an open a model that is small enough and efficient enough to run on your PC or Mac. And I think that's money. And then they're going to release their their fuller model soon and also open. Wait.

00:15:54:17 - 00:16:11:21
Unknown
So I think they're putting their money where their mouth is there. I think it's legitimate I think is right. They've gone back and forth on strategy. We don't know where they stand and they've all done that. But now I think this is something to pay attention to. And I also think it's another way to put it is it's also Zuckerberg trying to catch up.

00:16:11:24 - 00:16:43:19
Unknown
Yeah. Right. When you think of the big yeah. When you think of the the big labs right now we're very used to saying OpenAI anthropic Google although less Google, which we'll talk about in a little bit, Facebook like meta. I don't necessarily immediately put them in that same category, which is kind of surprising because meta is meta and like they're they're certainly capable, I think from sheer like resource perspectives to be up there in that, in that category.

00:16:43:19 - 00:17:10:13
Unknown
But they're not. Is that because of their dedication to openness, or is that just because they're playing catch up from the metaverse thing and, you know, they're just kind of behind. And this is this is a good strategy, this a this openness strategy is a good strategy to differentiate. Yeah, I think this is where MG has really is a point is that I think it's less important of strategy and more a point of how successful they were or weren't.

00:17:10:15 - 00:17:27:23
Unknown
I mean, lava was really important in its day. Everybody was building on top of the Lamba lava was the thing to do because it was available and it was meta. And that was the philosophy. And I think that deserves some credit, I suspect, for that philosophy. But then they didn't really develop it because they built this whole new arm.

00:17:27:23 - 00:17:38:13
Unknown
He spent billions to bring people in to do that. There was nothing coming out of there. So we'll see how these muse models are.

00:17:38:15 - 00:17:46:08
Unknown
Meta compares it to Gemma Googles and Quin 3.6 which is.

00:17:46:11 - 00:18:10:11
Unknown
When is Baba right. Alibaba. Thank you. Yeah. And it was started with an A. It was not anthropic. So yeah I think they're catching up. But I think the way they see to catch up is to be open. Which when we get to Google you wonder about the Android model of AI, which we've discussed on the show before.

00:18:10:11 - 00:18:34:06
Unknown
And in a sense, this is this is meta stealing that thunder. Yeah. Now you gotta you do have to. Running it on your own machine is is great. That does kind of preclude that the person knows how to do that. Well, that's where I always come back to this. I don't think anyone in AI is at a retail level here.

00:18:34:08 - 00:18:55:15
Unknown
Yes, you can go to a chatbot and go chat with it, but for all the stuff you want to do to create an agent and do all that, it's not really there. So I'm surprised that, glimmer isn't packaged as an app. There are some irony in that, because AI is going to kill apps, but it's the app of apps.

00:18:55:15 - 00:19:17:12
Unknown
Then if I could just if I could, if I could install it as easily as I could an app. Right. And I mean, this is and that's, that's what they got to do and that's what that's what OpenAI has done. That's what anthropic has done. They've they've taken their models and they've said here, just put this install this app on your machine.

00:19:17:12 - 00:19:35:03
Unknown
It has access to all the things, but it's still it's still calling back to Mother Bear. Right. It's still calling back. Yeah exactly. In this case, glimmer would run locally. Oh, boy. Yeah. See, that's the difference. I think that's important. I see what you mean. So if you if you if in retail terms, normie terms you say install.

00:19:35:05 - 00:19:55:25
Unknown
There's still in the app world. I mean I'm still in the app world. Yeah. Yeah yeah. And those terms I think, I think this could be huge and people would start. If people can really understand how to use this to start to create agents for themselves, now they've got to give access to their tools to do that. There's all kinds of stuff that's going to come forward, but I think we'd see a lot happening then.

00:19:55:27 - 00:20:23:11
Unknown
Yeah. Interesting. I mean, the agent kind of baseline, not baseline, but entry point is lowering. It's becoming more and more accessible to more and more people. I don't make predictions, but I think that within the year. So the rest of this year I think we will see two things. One is.

00:20:23:13 - 00:20:50:12
Unknown
Locally run models that are so simple to install. It's like installing an app. And two is WordPress.com like hosts, where somebody makes it easy for you to run one of the more sophisticated models. Right. And you can do that. I think we see the we have the d nerding of for not not for the nerds will still what the nerds do and they'll have more power at it.

00:20:50:14 - 00:21:12:06
Unknown
But I think to to really make this scale, you're going to have to make it as easy as the rest of computing is. Yeah, true. Yeah. I wonder what time, at what point we stop chasing the next big thing. It feels like there's there's never I mean, we're early on AI, but it feels like there's never any equilibrium.

00:21:12:08 - 00:21:31:03
Unknown
It's all it's always about like, yeah, that's that's amazing. But that's amazing right now. And oh, now I was going to say never JSON. But come to think of it and you'll be talking about this in a few minutes. There's not been a next big thing in phones in quite a while. Yeah, that's it's improvements incremental improvements. Yeah.

00:21:31:05 - 00:22:00:13
Unknown
But that doesn't that right now. And I mean granted we're still very early in I think this kind of moment in AI, but that isn't really the case with AI. It is constant changing. Absolutely. No. That's why you know, Google as we keep talking about the things we will talk about why Google was on on top of the world and now is, you know, by some accounts, not even being called a frontier AI lab anymore is an interesting perspective.

00:22:00:16 - 00:22:05:26
Unknown
See, I also think that again, I think that that.

00:22:05:29 - 00:22:31:20
Unknown
We can't treat all the users as if they're the same. Yeah. So if you are a scientist, you have very specific needs around AI. And that's where we had the basically spin off from Google last week is to work on that. And we see efforts from anthropic to serve just scientists. Then you have the whole huge layer of coders, people who used to code software, who now find they can make software in new ways.

00:22:31:24 - 00:22:58:17
Unknown
Well, that's another constituency that's different in how you treat them. And those two layers care somewhat about benchmarks at the high end. But even in the coding piece, 80% don't really shouldn't because everything they want to do, they can do with any of these models. Now then you get to the retail level, which interests me, which we haven't really hit yet except for chats where this frontier stuff doesn't matter in the least.

00:22:58:19 - 00:23:27:08
Unknown
They'll come along with more capabilities. It'll end up on my phone, it'll end up other places fine. So yeah, I think it's a macho size matters thing that matters a lot less to most people. Yeah, yeah, to the majority of people. Yeah. Shifting gears a little bit, this is interesting. I'm super curious to hear what you think about anthropic pledge to watermarking all of its content.

00:23:27:08 - 00:23:55:23
Unknown
This is a transparency move. That is all thanks to the EU's AI act, which has, I think, a deadline or had a deadline of August 2nd for enabling labeling rules. Anthropic basically announced that they're going to be embedding watermarks into everything that cloud producers produces. Cupa signed provenance metadata.

00:23:55:25 - 00:24:20:27
Unknown
They say text watermarks baked into the writing itself that anthropic puts out it will apply to all of their products. Any model released from August 2nd onward would support the watermarking. And actually they're also working on on adding the watermarking to the models before it. Text watermarking is it's not a new thing, but I but I don't know a whole lot about it.

00:24:20:29 - 00:24:39:13
Unknown
Like on its surface, it just sounds hilarious to me. It's like when I think of watermark and hidden watermark, I think of data that's there that you don't see, but in this case, it's patterns that are there that you don't recognize. Is that right? Right. Text. Text watermarking essentially patterns.

00:24:39:15 - 00:25:07:24
Unknown
So last week we talked about about this a bit where in terms in the context of the EU requiring labeling of AI. Yes. And that's one step here. And last week I said that I thought this was just the new cookie click. Everything is going to be marked as AI as a result. So now we have this next requirement which is watermarking.

00:25:07:26 - 00:25:29:19
Unknown
And I see it in that context of some extent I don't care so that it's made with AI. What are you making a big deal about right now when we're going to find that most everything is going to have some AI touching it? And so what does it mean to label it? All right. So then the problem becomes, as I mentioned earlier with Moby Dick, these AI detection tools are worthless.

00:25:29:25 - 00:26:01:26
Unknown
Yep. Not 100% worthless, but with enough doubt that you shouldn't make judgments about them. You certainly should as students. Yeah, as is anything in the world of generative AI, nothing is 100% factual. Nothing is 100% accurate. Neither are these detection detection mechanisms, as you pointed out with Moby Dick, which I think is a perfect example of like if they're saying that about Moby Dick, then why on earth do you put any sort of semblance of trust or, or like respect behind the number that it comes back with?

00:26:01:27 - 00:26:29:18
Unknown
Right. So now we have watermarking and I think it's a technical solution when that's, it's it's techno solution. And I think that's a mistake. The problem is technology. The solution isn't always technology. The problem is it's going to be individuals. It's going to be authors and publishers and professors, and just all of us are going to have to judge whether we trust people, and we're going to have to refute our norms around that.

00:26:29:18 - 00:26:55:07
Unknown
And that's not easy. And it's not clear and it's not automatic. The other thing I don't understand about the text watermarking is that anthropic is saying that it will enable people to then detect the watermarks. I don't know how that's going to work. If they tell you how it's watermarked, then that seems most taken away. So I'm going to guess they're going to provide a tool that will say, oh, I looked at it, I found our watermark or I didn't.

00:26:55:12 - 00:27:19:12
Unknown
Yeah, that's kind of what I assume. But it's not going to be hard to reverse engineer that. Number one. No. Number two. Yeah. Go ahead. No no no no. Keep going. Well, just my number two is simply I wonder if it's embedded in the in the in the meaning of the text itself. It used this word instead of that word in a way that only anthropic knows it would do.

00:27:19:19 - 00:27:45:11
Unknown
That changes the meaning of text. Yes it does. That's troubling to me just in terms of how we look at our own human language. Right. That there become these hidden secret signals of non humanity. And I don't know, that bugs me in a way. Go ahead. Sorry. Yeah. No, I totally agree. And that's that's kind of where I was headed, which is when I kind of saw this news and really thought about this.

00:27:45:11 - 00:28:05:25
Unknown
I was like, we already have AI watermarking. When I read an article, I often will read an AI ism if you want to call it that and go, oh, I mean, it's possible that the person wrote this, but I commonly see that as output from an AI. You know, that's that's a thing that, you know, you're right to push back on that.

00:28:05:26 - 00:28:35:13
Unknown
You know, the, the different things that we jokingly refer to as the tells that something is written by AI. That to me, kind of that strikes me as its own kind of watermarking that doesn't require a technology, nor does it require anthropic to manufacture itself. New ways of importance of producing text that that tells on itself, essentially. Let's go to Daniel in the comments.

00:28:35:14 - 00:28:57:03
Unknown
Raises a really good point. The watermarking can be affected by running the text through apps like Wilbert or other word changing apps. Potentially if they if they if they can figure it out, if they figure out the things that can be undone or whatever, which I mean, you put up a barrier or put up a restriction or a guardrail or whatever.

00:28:57:03 - 00:29:16:17
Unknown
And I'm sure companies like that are going to start putting their research into figuring those things out. So if they can differentiate their product as they see this gets you around the watermark that anthropic, blah, blah, blah. So, you know, it's an arms race in, in itself. So there's interesting. Did you follow the Hank Green controversy about about AI content?

00:29:16:19 - 00:29:40:18
Unknown
Remind me so I have I have in the last few weeks, I have a tremendous admiration for Hank Green. I think he is. He and his brother John have done wonders for the internet, and they show what a good internet is and can be in. And he's beloved on YouTube. So Hank put up a video on July 22nd, a video titled you have Never seen a Normal Star.

00:29:40:20 - 00:30:05:18
Unknown
In one section of the video I'm reading from plagiarism today, I don't think this is plagiarism, by the way. Green included a graphic showing how much spheres increase in volume as compared to the radius. That would be radii, I think. But I'm a human. Though it's unclear where he obtained the graphic, it was clearly AI generated. Not only did the graphic bear resemblance to AI work, but the math was wrong.

00:30:05:20 - 00:30:26:23
Unknown
Hank apologized for this and said he was simply in a rush and found the image online. He also removed the section of the video where the graphic appeared. However, just a week later, on July 29th, Hank posted an Ask Hank Anything video that featured fellow creator Soupy Time. During a cutaway to Future Hank. He says the line, I appreciate the pushback.

00:30:26:25 - 00:30:41:02
Unknown
Seemingly out of nowhere, many recognize the line as something an AI chatbot would say. So now Hank's kind of in his shell because of all this.

00:30:41:05 - 00:31:07:03
Unknown
And there's suspicion even given to Hank Green. Oh, you stepped off once. I think we should presume that everyone is going to use AI at some point when they're rushed. The problem is going to be when they didn't make that check, whether it came from a freelancer or it came from AI, or it came from a Google search or it came from Wikipedia, it's going to be it's not a question of whether it's AI generated or not.

00:31:07:03 - 00:31:39:14
Unknown
It's a question of whether you checked it or not. And Hank, you know, put up his penalty hand and said, whoops, I should have checked it and I didn't. I was rushed, we're all going to do that. And I think this judge mental piece of thinking that we can catch any AI I gotcha is is foolish. Yeah. And especially as I think about this, especially as these tools you know, one of the big benefits that is telegraphed about these tools, parroted all the time, is how much faster and more efficient and quicker it makes you as a human right.

00:31:39:15 - 00:32:10:21
Unknown
I let it do the thing for me so I can move faster, I can get more done, and the more people are getting more done faster, the more pressure is upon you to also meet them at that. Absolutely. And that's that's when you start to make those, those lapses of check because you're merely trying to keep up. I had coffee with a professor at Seton Hall who teaches journalism history yesterday, and I asked him, because every professor talks about AI now, and he had a lot of empathy with the students.

00:32:10:21 - 00:32:34:27
Unknown
They want to do good work, and this is a way to make the work better. And they deal with that. I just put this up at the bottom of the rundown before we got on. I just saw the seconds before wired said, oh Lord, AI reporters are actually breaking the news. They covered the Black Hat conference and runtime wire from video of the presentation about the open AI breach.

00:32:34:29 - 00:32:48:25
Unknown
Call it what you want. Wired had to, with chagrin, say that runtime wire with no one in the room beat wired with the story by three hours.

00:32:48:27 - 00:33:14:02
Unknown
I also talked yesterday to Nicolas de Leon, who happens to be a Consumer Reports writer, but he's also around, and he was on Twitter Sunday, and he's going to be on intelligent machines today. He started a totally AI generated news outlet in Tucson, Arizona, and he mentioned another one in I think it's Connecticut and there's others. And some people were yelling about this stuff and calling it slop.

00:33:14:02 - 00:33:39:23
Unknown
But on the other hand, and I'm trying to work with this in new Jersey, where nobody can afford to cover every school board meeting anymore, and every town meeting, every zoning board meeting, every and hyper local news has been de-emphasized and defunded. He funded. Great, right? So, so with the caveats of AI will make mistakes, etc., etc. there are things that you can do now that you couldn't do before.

00:33:39:26 - 00:34:00:21
Unknown
Yeah. And is it worth doing that for that reason? And I think it is with the caveats presented. Yes, you need to be transparent about but that's that's not a technology problem. That's a human problem. Be transparent about it. Say what you used it for. Give that caveat welcome criticism and correction. And then I think it's going to be helpful to people.

00:34:00:24 - 00:34:26:14
Unknown
So fascinating. But we're still in this mode of AI has cooties. And if it touches you, you got the cooties now, right? Right. And I and I pretty firmly believe that that there are a large number of people that are calling the AI cooties thing. They're still using it behind the scenes. They're just covering their tracks. Yeah. I mean, yeah, I think that there's a lot of that going on, whether people want to admit it or not.

00:34:26:17 - 00:34:44:18
Unknown
Yeah. It's just a it's a weird moment for this stuff. It's like everybody's using it. But because of the factor, not everybody's telling it, like talking about how they use it. You know, if I wasn't doing an AI podcast with you, I'd probably be in a position where I'm like, God, I don't know how much I should be talking about this.

00:34:44:18 - 00:35:04:00
Unknown
Just be just because the environment is such that it makes you question how much you talk about how you use it. But that's the reality. That's the fact. Actually, this this is kind of kind of ties into it is that I've got this AI workshop coming up. I saw, I saw you have a very nice ad for that, by the way, for those.

00:35:04:02 - 00:35:29:02
Unknown
Go ahead, talk about it. Where it when and where is it Jason. Yeah. So it's in Healdsburg, which Healdsburg, California, for the two of you that are geographically anywhere near where this applies. And that's kind of less to the point, but I'm excited for it's in a couple of weeks. But the, the, the, the place that I'm holding it, she has her own kind of designer or design team or whatever that I was kind of working with, with the information.

00:35:29:02 - 00:35:52:02
Unknown
She came back with the with the flier. Right. The flier for this event, which was better than my handmade one. Not a designer, you know, but but I look at it and I'm like, okay, I'm pretty positive this was done with AI. Sure. Does it matter? Like, you know what I mean? At the end of the day, at the end of the day, the images of us like our profile images on it looked weird.

00:35:52:02 - 00:36:07:03
Unknown
And so I was like, this looks great. I can tell, though, that the images were processed with AI. It kind of has this weird kind of like kind of me, but not quite me enough. And so I gave her the real profiles and I was just like, can you put these on instead? And she swapped them in and everything looks great.

00:36:07:03 - 00:36:28:15
Unknown
Everything. But but it's just interesting that like, even a designer. Right. Like it's part of the tool process. How do you feel like you need to even say, I used AI for this, or is it just a tool that everybody is using, whether we talk about it or not? And I think more and more that's becoming the case is that it is a tool that everybody uses.

00:36:28:18 - 00:36:52:24
Unknown
But I just don't think everybody talks about their use of it. And should they maybe they don't need to. Maybe it doesn't matter. I think it's rather like cybersecurity don't extent. We should just presume that your Social Security number is out there. Yeah. The question is what then if you're if you're if you're a company selling something, then you've got to protect people more.

00:36:52:26 - 00:37:08:08
Unknown
You can't make it. You can't make your customer vulnerable because somebody knows their Social Security number. Yeah. Or somebody knows their mother's maiden name. Right. And if that's all you rely upon, then that's your fault, because we should assume that it's all out there. Everybody, everything, everything out there that can be hacked in that way has been hacked.

00:37:08:09 - 00:37:29:23
Unknown
I think the same is true of AI. We should just assume that AI, that there's a good chance a lot 100%, but a good chance someone used AI to create something. The problem comes if I looked at that at that flier and I said, that's not JSON, that that's weird. That kind of creeps me out. Yeah, that's the problem.

00:37:29:24 - 00:37:50:14
Unknown
Right. So you took the extra step and she took the extra step to change it, to fix it. So it's not it's not the AI, it's the human interaction. There's a story that that I'm going to do this again. We didn't put it in because I'm jamming more stories in. But I found it fascinating. There's a new EU law, the EU Artificial Intelligence Act.

00:37:50:14 - 00:38:17:14
Unknown
Sorry Jason. This is down online okay. Obligations of deploys of high risk AI systems. So this says that deploys shall assign human oversight who have the necessary competence, training and authority as well as the necessary support to put up an AI model and make sure it's working. So this is kind of a statute that goes to what we're talking about.

00:38:17:15 - 00:38:42:17
Unknown
It makes it a legal obligation to say, you better fix this stuff, you better put you better supervisor AI and if it does something bad, if it does something wrong, it's your rump on the rack. Okay. We've just spent five minutes saying that's the right way to go. On the other hand, it's the use is going to be so luminous, it's going to be impossible to check everything.

00:38:42:18 - 00:39:16:20
Unknown
Right. How on earth do you actually. So it's not worth checking out of that? Yeah, yeah. Interesting. Well, society is adjusting. Yes. Yeah, it really is. It really, really is. Plenty more along those lines coming up. Thank you everybody for your support. Thank you patrons for supporting us. Patreon.com inside show. By the way, I think you you may remember we are running a sale on the Patreon to celebrate passing 100 paid patrons.

00:39:16:21 - 00:39:45:20
Unknown
We got one more week where you can use the coupon code 100 patrons, this 100 patrons and get the entire year for close to 50% off. And that gets you the daily podcast that gets you all ad free shows, discord, access, discords. Picking up a lot of stuff. I'm holding a bonus. Patrons only ask me anything at the end of the month as a thank you, and I just want to call out a few people because we've had a number of you join since we've been running this.

00:39:45:20 - 00:40:15:26
Unknown
So thank you, Chad, Stacy, coy S, Andrew Wieland, Tom Sanchez, Philip White or sorry, Tim Sanchez. Thank you, Philip White and Bradley Deeley, I am human. I am not perfect as as is the case with every AI model out there. So sorry, Tim, but yes. So hey, thank you so much for signing up, for putting your, your, your faith and your trust that we can deliver solid content for you week in and week out.

00:40:15:26 - 00:40:40:25
Unknown
So we really appreciate all of you. And yeah, that's still going. So go to Patreon.com AI Inside show. You can get that cheaper access for the entire year. Oh, is that you? Islam Islam Islamabad 2956. And YouTube says that's me. His name was called out though. Yay! His or her name was called on. All right. Yay! Thank you so much for your support.

00:40:40:25 - 00:40:57:19
Unknown
It's good to have you here. So anyways, Patreon.com AI insight show if you want to support. We would love that. And you know maybe we'll maybe we'll get it to one 2120 paid patrons. That's kind of my goal and it's running for the next week. So that's a lofty goal. It's a short amount of time. Tell your friends to join us.

00:40:57:19 - 00:41:15:13
Unknown
And if you live in Healdsburg, tell friends to go to Jason's thing. And hey, you get a sticker when you sign up to, you know, we go all out for you guys. All right, I got to take a quick break, and when we come back, we'll talk a little bit about invidious approach to kind of addressing the circular financing thing.

00:41:15:13 - 00:41:21:26
Unknown
Google's leadership shakeup. We got a whole lot more coming up, so stay tuned.

00:41:21:28 - 00:41:59:00
Unknown
All right. So in Vidya signed agreements with six of Wall Street's biggest firms. They're creating a compute financing platform to basically mobilize around $500 billion in outside capital in the coming years. A lot of big names attached to this Apollo Global, Blackrock, Blackstone, Brookfield Asset Management, Goldman Sachs, KKR and this is basically an effort to help its customers pay for AI computing power with no with with only outside money, at least, you know, at least from Nvidia's perspective.

00:41:59:00 - 00:42:17:12
Unknown
And it's aimed squarely at circular financing and the criticism around this like circular financing topic that continues to come up time and time again, usually involving Nvidia, because Nvidia is at the core of so much of what's happening. We certainly talked a lot about this. Jeff helped me understand. So I asked for this to be in the rundown.

00:42:17:12 - 00:42:44:24
Unknown
You can blame me, folks, if you say, oh, this is boring, but $500 billion is never boring. I think it's a huge amount of money, and it's a partial fix to the circular financing. As I understand it. The Wall Street Journal had a explainer. Sorry, talking about this. The problem for Nvidia is that of course we know its chips are in demand, but a lot of the companies that want them to do these neat new things, these startups that we want, they can't afford the chips.

00:42:44:27 - 00:43:11:29
Unknown
They and they can't get hold of the capital because you have the big guys like Google in the marketplace, the bond marketplace, getting that capital. So there's a scarcity of capital available for them at this level of risk. So there's $500 billion now that can, in the words of the journal, theoretically standardized chip financing, creating asset backed pools of capital for AI companies.

00:43:12:02 - 00:43:20:02
Unknown
And part of the part of the deal here is that.

00:43:20:04 - 00:43:46:00
Unknown
Nvidia has some skin in the game, in that if somebody goes bankrupt, it goes into default on this, that in video loses a piece of that as well. Right. But the thing about this demand is if somebody goes into default, there's somebody behind them who wants those chips and will be we'll be going in there. And so it's a the Journal said it's a securitization.

00:43:46:06 - 00:44:10:11
Unknown
Like as with airplanes, credit cards, mortgages, people take these things and repackage them into secure assets. The issue for chips is, and I'm going to forget which what metaphor they used in the in the journal, but they have the half life of lettuce heads. Heads of lettuce. Right. And so unlike an airplane, which lasts many, many years, chips value does not.

00:44:10:11 - 00:44:38:16
Unknown
And so the market was a little bit skeptical of securitized chips. But what this becomes then is that it's an expensive asset people need and United Airlines cannot afford, does not have the cash to buy all its planes. It leases planes that are bought by entities in the middle. And that's the way it looks like it's going to head with chips.

00:44:38:16 - 00:45:01:26
Unknown
And especially, I think, data centers and not just the chips themselves, but the larger assets as we see this market going around now, at the same time, another story that wasn't in the in the rundown that I'll shoehorn in here is anthropic made a deal with a big crypto company. Yes to riot. Just as we saw that meta was selling out its hosting assets.

00:45:01:28 - 00:45:32:28
Unknown
Space has sold out. Its is renting out its hosting assets. This riot platforms is is running out. It's hosting obviously anthropic. Amazon and Alphabet already do that. So compute becomes a fungible asset in this world where you come in and you buy me some some. And rather than saying, I've got to build my whole big data farm because I can't afford to do that.

00:45:32:28 - 00:45:55:29
Unknown
And plus it's getting harder to build them. We'll get to that in a minute. This becomes an interesting way to to try to fund the development of new AI companies. The reason I thought this was so important, besides something new going on here to change the marketplace, is that I think this argues against the bubble. You could argue that it is.

00:45:55:29 - 00:46:17:24
Unknown
The bubble is what happened with mortgage securities. It was that was the definition of a bubble. This could be that bubble. But I think you have a bunch of smart companies that want to shepherd their assets who were saying, let me in on this. Sure. I want, I want to I want to support this future. So I think more about than the maturing of it.

00:46:17:27 - 00:46:41:15
Unknown
Yes, yes. Well said, well said. Right. Yeah. So I didn't do a great job of explaining that because I don't. Well, but I thought this was a really interesting and important story in the journal was very big on it. They saw this as a big deal. And chips are a chips have a shorter half life from what you're describing, mainly because they're powerful when they come out.

00:46:41:16 - 00:46:59:08
Unknown
They're capable when they come out, but very quickly they are they they end up in the kind of not, not nearly as capable as the current cutting edge. Is that right? Yeah. But I think that's I think that's a little bit of macho views from within the AI world. I've got to have the biggest, best chips there are because I'm who I am.

00:46:59:15 - 00:47:15:22
Unknown
Yeah. When I worked at Time Inc. many years ago, we had a debate going on about two imaging systems, things you could do on your Mac and your phone right now cost millions of dollars. And one department was buying the wrong computer. The wrong system was incompatible with other systems. And I said, what the hell are you doing?

00:47:15:26 - 00:47:35:06
Unknown
You know, it's my system and I want to buy it. And if we're timing, it must be the right system. It's just a lot of macho about this stuff. Yeah. The truth is that justice from the consumer end, the models we already have will do 99.5% of what most people want to do. Not scientists, not high end corporates, but most people.

00:47:35:10 - 00:47:56:28
Unknown
I think the same will become true of chips. If you don't have access to the latest generation, you have access to the latest generation minus one. It's probably going to do what you want. Yep, it's probably going to be good enough. And so there's an amortization of that cost, a depreciation of that cost that isn't so rapid as in oh this is now worthless.

00:47:57:00 - 00:48:16:17
Unknown
Right now we build a whole data center around it. So when you listen to Jensen Wong talk about advancements from Nvidia, this is why in one of his recent keynotes that, as you folks know, if you listen to the podcast, I, I watch with the loyalty of a Ted Lasso fan.

00:48:16:20 - 00:48:45:14
Unknown
He was emphasizing basically how kudos improvements of efficiency can improve the effectiveness of the chips you already have, and that that's what they have to do. They have to take the installed base because people have invested in that installed base and make it better. Yes, they get new chips and yes, those chips are better. And yes, for certain people, it's going to be well worth investing in those new chips because it's exponential in its effectiveness.

00:48:45:14 - 00:49:06:18
Unknown
But you don't only have to buy new chips to get more efficiency. No. Totally interesting. So thank you for this. Absolutely. I think it's I think it's something to watch. And it fits in with what we talked about on the show. So that's why I'm glad we. Yeah well indeed does. Yeah. Sometimes have sometimes my eyes you know don't don't see all the, the important pieces.

00:49:06:18 - 00:49:43:18
Unknown
And this was one that I overlooked. And so I'm really happy that you brought it forward because I mean how many, how many episodes have we talked about the circular financing tied into the potential of a bubble and the potential collapse and all these things that people get real, you know, concerned about this? What what you explain to me kind of says to me that this is, like I said, the maturity, you know, it's growing more mature how they're talking about how this stuff is financed and treated and those, you know, those older chips, you know, they they still have use they still have a life like that sounds very mature and and more responsible

00:49:43:18 - 00:50:03:03
Unknown
to me. So and we should mention so it's on the record. So in case all this blows up, you can come back and say that I was completely full of crap. Ben Thompson, who's brilliant and long winded, I will admit that I used Gemini to to summarize his pieces. Yeah, he's writing, but there's a lot of it. Yeah, there is.

00:50:03:04 - 00:50:32:16
Unknown
He's comparing. Nvidia's a risky business. He says here with the railroads of the of the Gilded Age, which led to an investment and a big crash could happen. Yes. Could I? I'm not going to go into detail here because I don't understand it all, but leave that as a caveat. Yep, yep. His work is great. It actually pointed me to one of the things that we're going to talk about next year.

00:50:32:17 - 00:51:09:18
Unknown
One of the articles anyways. Oh, and I don't have the article to show. That's okay. Google's big leadership shakeup in the AI division, which is a story that keeps repeating itself. But this last week was pretty significant. Dennis Hassabis, who has been running Google DeepMind since it was formed, moving into a chairman and chief scientist role. Corey Cavusoglu, who was CTO of DeepMind, now SVP leading DeepMind on the day to day basis, not necessarily CEO him himself, but there you go.

00:51:09:21 - 00:51:36:12
Unknown
Have Jeff Dean, a legend in his own right at Google. He's departed. Sergey Brin, who's Google Google co-founder who we don't really see much of Sergey these days, but behind the scenes apparently has been nerding out a lot. Yeah, yeah, he's he's getting real excited. He's broken down with AI behind the scenes. And apparently he's going to be taking on a more active role in the AI direction of the company.

00:51:36:15 - 00:52:07:17
Unknown
So you know that on top of plenty of other top talents from the department Noam Shazier leaving for OpenAI John Jumper leaving for anthropic. There's a whole list of like these defections that are suddenly happening at Google. Makes you wonder what the heck is going on. And and Ben. Ben Thompson had an. Part of his post. Had a link to an article on semi analysis where they basically said where is that?

00:52:07:17 - 00:52:28:18
Unknown
Do I have that here? Yeah. The title Gemini is cooked but Google Cloud you know is cooking. So Google Cloud is like the beneficiary beneficiary of the failure that is Gemini to to their degree to their analysis. But they're saying that these sorts of defections r not the kind of thing that happened when people are really excited about the next thing happening in that department.

00:52:28:20 - 00:52:48:21
Unknown
Right? Gemini for Pro is would they all be leaving if they were very excited about, you know, ahead of the release of that, they say, quote, for all intents and purposes, we believe DeepMind is no longer a frontier lab, which shots fired as far as I'm concerned on that one. But but I don't know. What do you think this says about Google?

00:52:48:22 - 00:53:09:20
Unknown
I mean, on one hand, I think what it says to me is, again, confirming just how fast things shift in transform here. It was not that long ago that Google was the, you know, oh my goodness, they were the prince. You know, suddenly it was like, finally they're living up to their potential. And and then OpenAI had its code red moment.

00:53:09:21 - 00:53:32:03
Unknown
Oh shoot. Everything you know is is changing. We need to focus now. They're in the news. You know, the balance just keeps shifting away. I mean, what does it even mean to be a frontier lab? If DeepMind is not a frontier lab, I would have just assumed that they are like, what does that even mean? I don't know, so I think first, the folks who left were very important players, but they left for a reason.

00:53:32:03 - 00:53:51:23
Unknown
They want to go to the highest and AI and have it work for science. And I think that's the right way to go rather than going for AGI. And we're going to have one model that can do everything. Yeah, yeah, we're going to develop that high end. And that high end is obviously going to be worth money. I think it's a public benefit corporation, but it's going to be worth a lot to society to build that.

00:53:51:23 - 00:54:13:08
Unknown
So I think I would see that as a spin off. Yeah. Okay. .1.2. So Tim O'Reilly had a very good post. And there's this woman on on Facebook. I don't know how she pronounced Cecily or Cecile. I think it's Cecile Tamura. She's she's just great on Facebook. If you want to follow her, I recommend her highly. Cecile G.

00:54:13:08 - 00:54:41:05
Unknown
Tamura. She summarizes all kinds of news and AI and all kinds of papers, and does a really good job of that. She summarized, Tim's post about AI, and he said and she said that he she has him saying that this is a shift from frontier lab supremacist to AI diffusion, that that they see a strategic shift that says, this goes to your point, Jason, about maturity.

00:54:41:08 - 00:54:55:16
Unknown
Well, AI is out there. It's commodified. It can do all kinds of wonderful things. It keeps going, okay, okay, okay. But there's no differentiation there. So what the speculates is that.

00:54:55:18 - 00:55:17:10
Unknown
This is Cecilia's summary. The article posits that has open source and mid-tier models close the performance gap. The raw capability of top tier foundation models will inevitably become commoditized. Okay, we say that a lot. Then it says that what it's shifting to potentially is diffusion. That is to be the agent that gets AI out there in the economy and society.

00:55:17:12 - 00:55:47:22
Unknown
And the parallel that Tim makes is Westinghouse and electrification. The initial industrial electrification replaced centralized steam engines with massive electric motors that drove central factory shafts. What I've studied for hot type out now, if you went to Harper and Brothers, it had a gigantic steam engine. The first one set the whole building on fire. When they rebuilt, they put it in a courthouse in a courtyard so that it ran a central shaft throughout the entire building.

00:55:47:23 - 00:56:08:08
Unknown
Eight floors or so of every machine was run by these belts. And so when electricity came along, you always think in the analog of the past, they replaced the big steam engine with a big electric engine. Westinghouse came along and said, no, we could distribute this as long as we can get AC to you and smaller electric engines, then you can.

00:56:08:12 - 00:56:36:04
Unknown
You don't have to have a huge engine. You could have lots and lots and lots of little engines. And that's of course, what happened with the electrification. And so I, quoting Cecile O'Reilly, suggests the future of AI lies in distributed, domain specific and edge cloud application infrastructures rather than centralized supermodels. So this, they argue, leads to a realignment of these internal roles.

00:56:36:06 - 00:57:09:22
Unknown
Thus there there's infrastructure and compute. There's an application layer. There's next gen bets. And she says that by positioning itself as the foundational layer cloud infrastructure, developer tools, advertising ecosystem and monetization frameworks, Google can monetize the broader AI ecosystem regardless of which specific model wins end user preference. Like an electric utility or distribution platform, Google takes a cut of compute and usage without burying the sole risk of betting everything on a single frontier.

00:57:09:22 - 00:57:26:16
Unknown
Models dominance. I think it's too soon to say that, Google, but way too soon to say that Google is retiring from the frontier model business. As you just said, Jason, it wasn't long ago at all that they were. They went from bottom of the heap to top of the heap, and it's too soon to say, oh, that's that for Google.

00:57:26:16 - 00:57:47:29
Unknown
No, I think they're still frontier business. But I think if this is true, if this by Tim is right, then I think this looks like a wise strategic move to say we can see where AI is going. Everybody's going to be using it. How do we get a finger in every single pie. Yeah, right. So they're playing 3D chess is what you're saying.

00:57:48:00 - 00:58:16:28
Unknown
They are. And if you also assume open weight models are out there. Yeah. Then you don't want to be your model. No. Totally. No. That's a really good point. That's a really good. Too damn expensive. The payoff isn't there, unless you're doing something specific, specific and high end, like science, which is the spin off. I think there'll be other spin offs like that coming in specific areas, like finance, medicine specifically, and others where you build models that are specific to that.

00:58:16:28 - 00:58:41:24
Unknown
And yes, you can charge a premium for that and a lot for that because it's not mass, it's not B2C, it's B2B. Right? Right. If you're going to be B2C, you're going to count on there going to be a lot of open weight models, a lot of competitors, a lot of commodified abilities. So you might as well find a way to hold on to things like distribution and service.

00:58:41:24 - 00:59:19:02
Unknown
That's what Google what did the Google do with the internet? Just that. Yeah, right. Yeah. No, that what you're explaining actually seems a lot like what Google has done in the past and how they operate. Well, they know how to do what they know how to do. So that makes a lot of sense. And does it also kind of remove a bit I mean, not completely de-risk, but does it remove some of the risk of, of, you know, being the the frontier model company that leverages all of their resources into that single thing just strikes me as like, all right, well, you're putting your eggs a lot of your eggs in the same basket, and

00:59:19:02 - 00:59:43:14
Unknown
that's risky compared to what you're talking about here. With Google being a lot more diversified and spread out and not quite so much about this one thing working. But I think there are different business opportunities here. Frontier models will still be an opportunity. But speaking of having the half life of a head of lettuce, that's where your advantage disappears.

00:59:43:15 - 01:00:10:09
Unknown
Just witness the story very quickly. Totally. Yeah, right. Specialized models, I think, will have high value, but a whole different sales calendar and value proposition. But going into those specialized models will be important. And then there's the retail end. And I think that's where Google can shine. Yeah. Because as you're about to mention, they have AI embedded in things like your phone.

01:00:10:12 - 01:00:46:26
Unknown
Oh nice segue. Yes, indeed. This morning the wraps were taken off of the latest lineup of pixel devices, the Google Pixel 11 series of phones. And we don't need to talk about this very much because there is an insane amount to talk about here. It's just a new, you know, their new current flagship crop of pixel devices. But, you know, it does have a chip that they that Google is pretty much in command of with the tensor G6 that's handling a lot of the AI stuff on device, and it has a bunch of features that are AI related.

01:00:46:26 - 01:01:12:00
Unknown
I'm literally I'm waiting for like, I keep checking my ring because I would love for it to show up while we're talking on the podcast, so I could like, show it off. But I will be so jealous. I know, yeah, I just want to rub your face in it. The pixel 11. But you know, some of the AI features, one that I'm actually really looking forward to is Rambler, which we had known about because I, because I realized that's that's how I am with AI.

01:01:12:00 - 01:01:28:15
Unknown
I ramble a lot. I, you know, I now use Whisper Flow. I've been using that for months where I just hold down a key on my keyboard and it opens up the mic and whatever field my cursor is parked into, it's just immediate voice dictation. I'm doing that all the time now and just rambling into AI for the things I want.

01:01:28:16 - 01:01:57:22
Unknown
And you know, I can I can think doctor do in the in the AI inside discord for for opening my eyes to that because Doctor Do is very prolific with using voice to text. Anyone who's in our discord would would see that that he uses it a lot for all of his AI work, which is really impressive. But Rambler is essentially just like baked into the OS so that when you are doing voice dictation into like a text messaging section or whatever, it cleans things up.

01:01:57:22 - 01:02:17:18
Unknown
So if you, or, or if you say something and then change your mind and say the other thing, it cleans it up in a way to where it still presents it in your voice. Or at least they say it does. Time will tell to figure that out, but it cleans it up so it isn't all kind of broken up and kind of hard to read, even if it sounded fine, you know, that sort of thing.

01:02:17:21 - 01:02:47:18
Unknown
This is really cool, though. Sign to text in Gboard. And actually, yeah, this is really neat. Google DeepMind's SL2 model, which essentially enables and I think this I won't get the fold, the pixel fold, but apparently the pixel fold is where this really shines, because you can kind of prop it up and using both hands, you don't have to just do, you know, it doesn't just lock into one hand, but signers can naturally sign their messages into the phone.

01:02:47:18 - 01:03:06:06
Unknown
It's a really neat accessibility feature that uses or if you're if you're in a conversation with someone who's using sign language, that's how you can interpret it, right? You can interpret it. Yeah. That's so cool. Yeah, it really super cool. I mean, that that is an accessibility. Some of some of the neatest things happen in the accessibility realm.

01:03:06:06 - 01:03:23:15
Unknown
And, you know, the vast majority of us never even consider them because we don't fall into that smaller group of people. But I'm so happy that companies like Google and Apple does this a lot are creating those features for people who would really need this. Like, I'm sure there are plenty of people who see this and they're like, oh my God.

01:03:23:16 - 01:03:45:03
Unknown
Finally, you know, one thing I didn't know until I started researching the history of the typewriter and keyboard for my book Hot Type out now. Sorry, is have it all ready to go. You just mentioned it. I just popped it right on the screen, buddy. The drinking game. The very, very first typewriters were written to benefit blind people, right?

01:03:45:05 - 01:04:27:21
Unknown
Yes. And also people who were hard of hearing as well, because people who could sign faster than you could. Right. And so these efforts accessibility often work around innovation. Question for you for the Google stuff because you had a list of, of of various software things that they're announcing. How much of that will come back to our pixel tens that I don't really know the answer to yet on a, on a piece by piece basis, I don't know, I would imagine a lot of this will, but I couldn't tell you which ones of these, you know, like the live translate that auto dubs video and audio into your language while you're watching it.

01:04:27:21 - 01:05:06:17
Unknown
That's powered specifically by the tensor G6. So that's probably not going to back port. So some of these will be tied to the current hardware. But I'm sure Rambler I'm guessing Rambler will will, you know, scale back. That's going to be embedded in Gboard. There's a magic capture where you in camera mode if I'm if I'm taking a picture of something and I have that active, it basically takes around 400 frames on both sides of my taking the picture so that it can capture the before and the after in all of those frames, and you can get that perfectly timed picture.

01:05:06:20 - 01:05:23:04
Unknown
I feel like they've introduced features similar to that before, so I'm curious to see how that actually works. But you know, you're a parent when you're, you know, when your kid is younger and you're trying to capture that perfect moment and you're always early or late for the perfect moment, this would be the kind of thing that would help you through that.

01:05:23:04 - 01:05:40:26
Unknown
I don't know how much these things will scale to earlier devices. Some of them will, though. Inevitably. It'd be interesting to hear your reports once you get the phones. Yeah, I skipped from the pixel six to the pixel ten, so big jump. It was a big jump, but I'm going to doubt that I'm going to have much motivation to go to the 11.

01:05:40:29 - 01:06:07:11
Unknown
Yeah, yeah, you'll get some of these. Almost certainly Gemini Gemini multi-step task automation working across more apps. That's something that you probably already have access to. Yeah. Interesting stuff there. Looking forward to testing that. And let me just check. And I don't think that anyone's at the front door, so no, maybe if it comes before the end of the show, I could show it off real quick.

01:06:07:14 - 01:06:38:18
Unknown
This was catching some steam on Tech Meme, this wired article that is about a new technique for seeing what AI models are thinking, their reasoning traces, and researchers apparently have found a way to extract those traces from AI models, which is basically seeing the internal steps that that model is taking to reach an answer. Some of these models that we work with, and you pointed this out many times on the show that you that you think that it's kind of performative more than anything.

01:06:38:18 - 01:06:59:06
Unknown
But some of these models present their ways of thinking, you know, in a kind of an endless stream. And a lot of them, like hide it behind a little pull down or whatever, so that you don't see it while it's kind of working through things. That leaves a trace, but that's a visible trace that you can identify. And a lot of these models do those traces behind the scenes, not surfacing them for you.

01:06:59:06 - 01:07:25:16
Unknown
And that's apparently how they kind of work through their own specific flavor of reasoning, let's say. Well, these researchers have said that you can feed they've discovered that you can feed that proprietary models, encrypted reasoning trace to a smaller, weaker version of the same model. And doing that can reveal that hidden chain of thought. So you can you know, it can expose things like embedded passwords, API keys.

01:07:25:16 - 01:07:48:06
Unknown
It can also, you know, it's kind of sounds like distillation, but it's a different mindset. They're trying to argue deflation though. Yeah there because they see similarities model to model. Yeah I think we've got to get past this notion that distillation is a bad yeah, life is distillation folks. That's true. Right? I learned I distill you every week, Jason.

01:07:48:11 - 01:08:13:14
Unknown
Yeah, I learn from you. I take away something you said. I quote you later. I distilled I think about as well Jeff Jarvis distill each other. We are we are a distillery, the two of us. And so that's how life works. Yeah. And I remember back in the day when Larry Lessig of Harvard was fighting copyright battles all the way to the Supreme Court, where he lost, the argument that all creativity is dependent upon earlier creativity.

01:08:13:14 - 01:08:33:29
Unknown
And I doubted that at the time, but I come more and more to see that. And so if it's true of art, if it's true of music, we all learn from prior art. We stand on on other shoulders, then the same thing can be true in AI. So get over it. Especially because the models distill themselves there in a hell of a bad position to argue that you shouldn't distill.

01:08:34:02 - 01:08:59:24
Unknown
So I didn't I didn't see any scandal to this. Yeah, yeah, it was at the top of tech meme. Everybody on the acting is distillation is oh, but I think it's no big. Well, you know, there's the whole like Kimi the moonshot ai Kimi, k3 producing some output that seems like continuations from like opus 4.8 and GPT 5.6.

01:08:59:24 - 01:09:26:14
Unknown
Soul's own reasoning, things like that, that go see CC. But you're right, we're all yeah, I mean, I would totally agree with your comment there about creativity being kind of based on previous creativity. Like there's nothing there's almost nothing in this world at this point that is truly unique when it comes to human creativity. And it's not to say that it's impossible, but it's all influenced by something, well, something that may come off as truly unique, maybe stupid.

01:09:26:14 - 01:09:48:05
Unknown
That's why you haven't heard it before. Yeah. And obviously if you look at academics, the first role of I'm not a real academic, I only play one on TV. Is that the first thing you do is literature reads. You make sure you read what's out there so you know what you're saying. You don't act as if you had a new thought when somebody else had it.

01:09:48:06 - 01:10:17:00
Unknown
Right? Right. So yeah. Interesting. Okay, so I didn't get the phone. I didn't get the pixel 11 in time, but I do have the Pebble Index oh one, this tiny, tiny little box. This as a callback, I don't know. Eight, nine months ago I, I put my order in, paid for one of these, which did not cost that much.

01:10:17:00 - 01:10:40:22
Unknown
As you can imagine, it was $75 for preorder. I think the normal price is $99, but I just got it and so I thought I'd unbox it. So what? So what this is, is it is an AI smart ring essentially. And I got mine. Okay, so I wonder I can't even remember what finger I fitted it to. What finger did I fit it to?

01:10:40:26 - 01:10:58:29
Unknown
Oh, it doesn't really fit on my middle finger. It's a little bit my ring finger. It does. Yeah. It probably won't fall off. So the idea is that when you're wearing this and I still need to set it up with software and everything, and actually, I probably want it to be like that because that is the microphone right there.

01:10:59:00 - 01:11:12:22
Unknown
I don't know if you can see that, but when I'm wearing this, God, I almost wish that I had done it on my pointer finger so that I could just go like that. Oh yeah, that would have been the better one, I wonder. Oh, maybe I did it on this side. Maybe I did it on my other finger.

01:11:12:24 - 01:11:35:09
Unknown
Yeah. Because you're using right figure, right? Oh, you know your left handed out you. Oh. Oh yeah. Anyways, I've got options, but not the options. Like, I kind of wish it was pointer finger on my right side. That would have been my preferred my preference. So anyways, so the idea here is there's it records little short bursts of audio.

01:11:35:10 - 01:12:00:08
Unknown
So it's I tap and hold the button, it starts recording. It's capable of holding up to between 12 to 15 hours of total record time for wow hours before it's completely done. You're not meant to just record hours upon hours at a time. It's meant for very short bursts. So like 10 to 20, you know, short 3 to 6 second voice notes per day, let's say.

01:12:00:09 - 01:12:24:03
Unknown
And the idea is that, you know, when you have a thought, Boop. You speak the thought, let go. It sinks to your phone. I think it actually, I think it actually stores on the device. I know that the battery is not changeable. So there's a little bit of an e-waste kind of thing to this. Although when it's done, when the battery is dead, I send it to them and I can get it replaced, probably for a fee.

01:12:24:05 - 01:12:48:14
Unknown
So hopefully they do something responsible with it from there. But yeah. And it has, it has a companion app. It connects to integrations with Apple reminders, Google Tasks, webhooks, that sort of stuff. But it's kind of billing itself is kind of like a privacy focused, you know, sort of thing. It's local transcription, offline operation, open source app, all that kind of stuff.

01:12:48:18 - 01:13:05:10
Unknown
So I'm curious to see, because so often I'm out in the world and I'm like, oh, I have an idea. You know, it would just be so much easier to tap the button and hold it down and speak the idea and move on versus get out my phone because, like, I'm driving, I can't reach out for my phone and launch that app and hit the record or whatever.

01:13:05:10 - 01:13:23:12
Unknown
But I have those momentary ideas. That's why I wanted this. I was like, that happens to me all the time, so I'm curious to see if that's useful. I wonder whether it will integrate with agents you create to say, when you say, add this to my calendar, it's not just a reminder to you to do that. It can go ahead and do it right.

01:13:23:12 - 01:13:41:24
Unknown
Yeah, that'd be cool. Yeah, I wonder. Dang, I wish it fit this finger. You got to go on a finger diet, man. Yeah. Apparently so. And then it'll get stuck on my finger forever because I'm not wearing the rings I normally wear. You know, I have my wedding ring, and then I have a ring that I normally wear on my left hand middle finger.

01:13:41:28 - 01:14:06:26
Unknown
So wearing this over here, that would, would really way down this hand with rings. And actually that's not even comfortable like that feels too tight when I do that. So I don't like that. So anyways, I probably got the wrong size. Oh well I'll figure it out. That's that's Pebble index oh one on my fingers now and hopefully not falling off.

01:14:06:26 - 01:14:30:26
Unknown
Although God I think I got the wrong size. That sucks. Oh well it's okay. Let's take a break while I figure this out on my hands. Definitely. Real quick reminder if you like what we do, give us a review on Apple Podcasts. It ends up on our website. It convinces more people to listen. We appreciate you going to take a break, come back and do a speed round and get you out of here because we are going long.

01:14:30:27 - 01:14:35:05
Unknown
It's one of those days. Back in a moment.

01:14:35:07 - 01:15:13:05
Unknown
All right. The AI backlash roundup got some UK venues banning meta smart glasses. More and more of this is happening. Wetherspoons, which operates over 800 pubs across the UK, has banned them from all occasions. They're joined by Jeremy King Restaurants, Soho House, Monopoly Events, which runs Comic-Con UK. I think I talked about that possibly last week. ATG Theaters and Wetherspoons CEO Tim Martin is calling it surreptitious surveillance, so.

01:15:13:07 - 01:15:36:04
Unknown
Yeah, but you know, a University of Sydney study found that 60% of analyzed, covered point of view videos, if you want to call them that showed harassment or potential harassment. There's like this whole video movement happening with people wearing these glasses and like harassing women. And basically, you know, there's a there's terminology for this. They're called pervert glasses by a lot of people.

01:15:36:04 - 01:16:02:25
Unknown
And, you know, that's that's troubling for me because like a lot of what I review online is yeah these glasses. Well but it's weird. I mean is is the capability of the latest metaclass that different from the prior generation? They all had cameras, right? No, they're they're the same. So. So why now? Why suddenly? Because I think because they're becoming more popular and you're seeing A and B the companies are becoming more hated.

01:16:02:28 - 01:16:32:05
Unknown
Yeah. That's true, that's true. It's the same as we're going to get into next. The same as the data centers I think it's it's these are these are MacGuffin people want to go after because they have a bigger opinion about all this. It's a AI backlash icon that's easy to spot, easy to see. You know, I yeah, I'm reviewing a few right now, and I will admit, when I wear them out of the house, I still have not gotten used to the fact that I'm wearing them when someone walks up to me because I almost just want to take them off and be like, hi, how are you doing?

01:16:32:10 - 01:16:53:16
Unknown
Because because it you know, because I have had a conversations with people where they're like, oh, that's one of those blah, blah, blah. And I'm like, God, what are they thinking like that? I'm recording them right now. You know, it, it it it has a growing stigma to it. And things like this are happening more and more. So yeah, I mean, the problem is if someone is doing that awful a thing.

01:16:53:17 - 01:17:11:21
Unknown
The problem is the human, not the technology a b I still don't think there's a lot of use case for these things. And the use case that is going to come is when they become smart enough to see where you are and direct you and do other things that were the information is for you. It's not so much recording in the world, it's finding your context.

01:17:11:22 - 01:17:36:17
Unknown
Yes. And if we if we lose that capability, see our earlier discussion about accessibility. You know, if, if a person with poor sight walks down the street and uses these and they get attacked because they're using it to navigate the world, that will really suck. Yeah, absolutely. These can be really great. Accessibility devices, you know, see earlier conversation about accessibility and tech and AI.

01:17:36:18 - 01:18:03:27
Unknown
So and yeah, I think you're spot on it. These are, these are these are very easy to identify targets for AI backlash. You know, people who feel strongly against AI. Another target, of course, data centers and Annenberg survey found that 61% of Americans oppose having data centers built in their communities. Not in my backyard. It's up 12 points from 49% in the spring.

01:18:03:28 - 01:18:32:00
Unknown
14% actively support local data centers. So very low number under 30 oppose data centers at 70%. There's there's a lot of like younger strong pushback on this. That's one thing that I've been kind of noticing lately is like, oh, so it's like like I'm generation X, I'm more pro AI, but I'm not like AI, bro. I would say I would describe myself as not AI bro.

01:18:32:00 - 01:18:50:17
Unknown
Anyways, and so am I in that older category of like, oh, you're just a boomer. You think that there's nothing wrong with these things, but hey, what are the ones that have to live with the impacts? Because we're younger, we're growing into it and blah, blah, blah. You know, there's a little bit of that going on right now.

01:18:50:19 - 01:19:15:24
Unknown
Yeah. Interesting. Watching the coverage this morning, I missed now of the various primaries. Data centers came up a lot. It's a it's football. Absolutely. Yeah. So it's going to be interesting to see as we get closer and closer to the midterms here in the well from footballs to pucks. Yeah it's our sports moment. Open AI working on a hardware device.

01:19:15:24 - 01:19:41:10
Unknown
As we know Bloomberg is saying it's a hockey puck sized device that costs will cost more than $300. Donut shaped smart speaker type thing. Premium metal finish. Battery powered, designed to be carried one handed between rooms. So this is like a little a hockey puck companion with AI, I guess. No display though. Voice based ChatGPT conversations with a little speaker grille.

01:19:41:10 - 01:20:04:29
Unknown
So basically this is like a Google Home. Like those little Google Home pucks that you just put. Portable but portable. I think I've just used the ones that connection wired in. Yeah, yeah. Nor do I need feel the need to like, bring it from room to room. But yeah, I don't know. More and more. I'm really. If you're paying $300, you're not going to buy five of them.

01:20:04:29 - 01:20:42:06
Unknown
That's for one thing. That's for sure. That's for sure. OpenAI also launched Daybreak Cyber Partner program. These are two different models for different types of cyber analysis and using AI for cyber work Daybreak blue for defensive cybersecurity. Daybreak daybreak Red for red teaming kind of uses. And yeah, basically OpenAI building a cybersecurity specific channel for its models within enterprise guardrails and that sort of stuff.

01:20:42:13 - 01:21:03:22
Unknown
So yeah, they're doing it to see what is it? It's anthropic project Star Light, the name of the. Yeah, right. No, it's another name. But yeah. Which is fine. I think they should put them in the hands of companies. The more that we test companies software before you put these models out, the better off were going to be.

01:21:03:22 - 01:21:28:24
Unknown
It's fine. Yep. It's smart. Got it. Just realizing there's a bunch of ants on my desk. Did they come out of the ring I could you imagine? Oh no. It's the it's the ant version of Odyssey. Oh that's frightening. All I got were these was this pile of ants when I opened the box. Now, I don't know our house.

01:21:28:24 - 01:21:51:29
Unknown
I'm pretty sure our house is on a massive ant hill. We are constantly battling ants here. It sucks. I hate ants, I'm sure they're great for the environment or something. Something. But I hate it, I hate him. Cloudflare has open source what they're calling Cloudflare OS. It's not actually an operating system. It's a browser based workspace for AI agents built on Cloudflare workers.

01:21:51:29 - 01:22:14:03
Unknown
So it's like thousands of Cloudflare agents, basically creating documents for you. You know, the typical stuff at this point that we're seeing with. Yeah. The weird thing about this cloud, I wouldn't use this just because I'm not I'm not a company and it's meant for companies and all Cloudflare are saying, hey, we built this for ourselves. We think it's so good we're going to share it because we're wonderful.

01:22:14:05 - 01:22:43:12
Unknown
Cloud for Cloudflare is getting more and more to be a meddler. Right. And their whole efforts around trying to say, oh, Google's bad for for for scraping you when you're going to end up in the search where you want to be in an AI, where you should want to be with the right arrangements. Cloudflare is putting itself in the position of being the cop of the world, and then here they're putting them in the position of perhaps even competing with the companies that they serve, which I don't think is going to be very smart in the long run.

01:22:43:15 - 01:23:13:13
Unknown
So I don't know what's happening in columnar. I think they're getting weirder. They're they're spreading their wings a little too much. I think they should stay in their lane, but they're not. There you go open source and GitHub so you can see it for yourself. But yeah, interesting. I know we don't normally talk about AI and grok a whole lot, but actually I launched Grok Bot its own kind of like AI teammate thing.

01:23:13:16 - 01:23:46:10
Unknown
You assign work to it, the agent runs on your desktop and, you know, authenticate third party tools. It can execute all of these workflows, multi-step and all that kind of stuff. But it's but it's very expensive. It's beta access for super heavy subscribers, $300 a month or cursor ultra at $200 a month. But one thing about this that I'll say just real quick is that they they did this whole animation system with the bot to give it, like, I don't know, emotions and stuff.

01:23:46:10 - 01:24:13:07
Unknown
And I thought that was interesting from the perspective of how we interact with these agents and these bots and everything. And I wonder what I wonder if this just becomes visual noise or if it's actually like, oh, I, I feel something about working with this bot because it has some sort of visual, perceptive emotion in the animation. Once I see the word grok, I know everything I need to feel.

01:24:13:12 - 01:24:40:21
Unknown
Yeah, yeah, fair, fair enough. Rock might use Nvidia's new open model Nemo Tron 3.5 lightning. New open weight model that you were alluding to earlier. 30 billion total parameters, 3 billion active per token mixture of experts, basically for up to four times the output speed of similar sized models. According to vedere runs on a single 800 or a DJ's X spark.

01:24:40:21 - 01:25:07:29
Unknown
Maybe Leo will run this on. He got a spark right? He's hoping that it arrives today as well. So maybe on the same truck as your phones? Yeah. Could possibly be, open for commercial use under a license. And yeah, there you go. And then finally, we've got after I put away this little scare overlay thing, Spotify announcing it's going to start labeling AI artists on the platform.

01:25:07:29 - 01:25:35:18
Unknown
This isn't about labeling music. This isn't oh, this is a this music was created with AI. We're going to label it. I think they're already kind of doing that. This is about labeling for for AI. Well persona. Right. So an artist's public identity. So if that is a is an AI artist from a public identity perspective, not an actual person or whatever, then that's what this is all about.

01:25:35:18 - 01:26:01:15
Unknown
So I'm curious to see where that kind of appears. I mean, it's kind of the same thing, right? Like if it's an AI persona, it's probably AI generated music, I'm guessing. Yeah, but yeah, this KR prior discussion about about anthropic watermarking things. I think that transparency about AI use when it is relevant and meaningful is an ethic that we humans should engage in.

01:26:01:17 - 01:26:17:13
Unknown
And I think when it comes to seeing the music, people want to know whether this came from someone's heart or whether it was intended to speak for me. And I don't think there's anything wrong with the latter. I think that that that I've long thought that these tools can be used at people to express themselves better, and they got help from the AI.

01:26:17:13 - 01:26:37:23
Unknown
I don't think there's anything wrong with that. I think transparency is probably not a bad thing because people feel see the green discussion. They feel like they're taken, even though he did what everybody does when they make PowerPoints now. Yeah. So fine. The safest thing to do is to say, well, the way I used AI for this, I think it's a new ethic.

01:26:37:25 - 01:27:00:27
Unknown
Yeah. Totally. Totally. All right, well that's it. We've given you a little bit of a super sized episode today, and I love that hot type has come up multiple times because. So much. Yes. August 20th is the arrival date. Although, Daniel, did you get your book? Let us know. Because he's waiting for it to be delivered, so you might even get it earlier in advance.

01:27:00:27 - 01:27:34:04
Unknown
But Hot Type can be found at Jeff Jarvis com, along with all of Jeff's amazing books. Just go to Jeff Jarvis com and you can order any and all of them. We've got upcoming series from Bloomsbury intelligence, AI and humanity. So be excited about that as well. Coming from Jeff Jarvis for me pod tuneup. Com. If you have a podcast or have ambitions with a podcast and you want to, you know, have a have a mind in podcast to kind of share the thought and vision and all the things with to help you do better with it.

01:27:34:05 - 01:27:53:07
Unknown
Hit me up pod tune up. Com AI inside Dot show is where you can go to find all the information you need to know about this show. Everything. All of our episodes. Are there information for the store, links to the store to by you know, AI inside merchandise and all the things that you might want attached to the show.

01:27:53:07 - 01:28:27:03
Unknown
And then finally Patreon.com inside show again code. What is it? The 100 patrons and you get almost 50% off for the entire year. And we really appreciate that. Also want to give a huge thank you to our executive producers because they help the world go around on this show. It is Doctor Do Jeffrey Martini, radio Asheville 102.7 Dante, Saint James, Bono, Derek, Jason Nefer, Jason Brady, Anthony Downs, Marc Starker and Karsten.

01:28:27:05 - 01:28:51:28
Unknown
Appreciate you all for continuing to support at this level. It really, really does help. And I think my last thank you is to Victor Bogart and Daniel Croft behind the scenes. Oh, and hey Daniel, speaking of Daniel, he says it right? Daniel can't wait to hear what you think. So we're gonna end the show so that Daniel can read a favorite post, a picture of yourself with it, and I'll share it right about that.

01:28:52:00 - 01:29:09:04
Unknown
Love that, love that. Thank you. I'm so happy it arrived during the show. Daniel. That's awesome. And congratulations, Jeff. All right. Thank you everybody for being here. We do the show every week and we'll see you next week. Probably same bat time, same bat channel on AI inside. Take care everybody. See you soon.