Google Changes Search for the First Time in 25 Years // AI Inside #128

Ajouté : 2026-05-21

[00:00:13]Hello everybody and 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.

[00:00:20]uh from coast to coast took me to Mountain View, which is not actually very far from from me and definitely not coast to coast for me, but would be coast to coast for you, Jeff Jarvis. Uh it's good to see you.

[00:00:32]>> I was there virtually. It's also good to see uh you and our new look. I like it.

[00:00:38]>> Yeah. Yeah. Kind of I don't I'm sure people who you know, if you've if you uh found your podcast this week and you were like, "Wow, things look a little different." I just I you know, I got wild hair. I was like, you know what?

[00:00:50]>> This has been watching too many home re redecoration shows. You know, >> actually, it totally ties in with this show. I mean, aside from the fact that like, yes, I I redid the the podcast art and I'm really proud with the way um happy with where it ended up, but I used Claude design not to create it, but to help me analyze it. And so, we can talk about that a little bit later. I'll show you >> the hilarious results because it's hilarious, but it still got me to a good place. So, it's interesting stuff. But, um, but yeah, uh, there's that. There's the daily that I've been doing and, uh, I've been, >> man, I've been having a great time with it. It's, it's great. It keeps me really kind of like on top of things.

[00:01:32]>> So, you didn't have that second week regret. What did I do? What did I >> commit myself to? It's well often with this show like I'm just usually so often so busy throughout the week that I don't have time to like sit down and read the news >> until it's like the day before the show and like got to cram for a test. You know what I mean?

[00:01:51]>> Yeah. And this way it just kind of it's like slow trickle sort of thing. And then um we're as we will you know I will thank a bunch of new patrons later because a bunch of people are like okay yes sign me up.

[00:02:02]>> Wow. Yay. So yeah, but what I've and and I won't do too much house cleaning here at the top, but um the daily is still free through the Patreon through the end of this week. patreon.com/aii show. If you want to take a listen to these episodes, all you got to do is become a free member of the Patreon. Uh you just sign up and you will get notified when the new episodes hit.

[00:02:25]Starting next Monday, it switches to a paid uh tier um perk. So, $5 and up.

[00:02:31]Basically, $5 a month and you get what what is it? 20 some odd episodes of a podcast uh that didn't exist more than a couple of weeks ago uh in your Patreon feed and actually in your page in your uh podcaster because you actually get a podcast feed that you can load into whatever you use to listen to podcast.

[00:02:49]But I in talking with you Jeff and just kind of thinking about it, I've decided that one episode per week will always be free. And I think that way if you become a free member of the Patreon, I you know, I don't know what day that is.

[00:03:03]Maybe it's every Friday or you know, because then you get the Wednesday normal episode and then a Friday news update or whatever. Um that way you can continue to get some value out of it and uh maybe someday you'll be like, you know what, >> the spider to the fly once you see one once a week, you're going to want it every day. So be like, how can you not have a Jason a day, >> right? So anyways, uh, wow. The last few weeks for me have been full of impulse, but I'm happy at the result of the impulse. It's like, I should do that.

[00:03:33]Okay, I'm going to do that. Oh my god, what did I do? I did that. And then >> I'm happy that I did that. So anyways, so that's that. But you know what else I did? I went to Google IO yesterday and uh, boy, that was a day. So you were watching the live stream.

[00:03:48]>> I watched the live stream u Yeah. uh with uh with Micah and Leo over the other place.

[00:03:55]>> Okay.

[00:03:55]>> That's that's a club only event, so not as many people see it. Um but it's a way to have company while I'm doing it.

[00:04:00]Otherwise, when I've been at IO, I sat next to Jason.

[00:04:04]>> Yeah.

[00:04:05]>> And when you're kind of like, oh, is that a big deal? Oh. Um >> we need to get you back at IO is what needs to happen.

[00:04:10]>> Well, they invited me this year, but it's just not worth the flight. Um especially in my infirm condition.

[00:04:16]>> Um United Airlines six hours with my bad back right now. No. Yeah.

[00:04:22]>> But uh but it's because it I because I don't only want to see the keynote. I want to see what they're showing off around the kind of show they put on.

[00:04:30]>> Um and before we even get to the keynote, let me just ask real quickly.

[00:04:33]>> Uh because they the keynote didn't touch on Android. They was without a week before. They didn't touch on, >> you know, >> two or three mentions of Android maybe through the entire keynote.

[00:04:42]>> They didn't mention the Google book, which I thought they would even >> Yeah, they didn't even mention said there'd be more. Um uh they mentioned I can't think of what else they left out but there was stuff in past years hardware obviously in the past they have the glasses but that's it.

[00:04:58]>> Yeah.

[00:04:59]>> So what about the the the tents that they have around did they show those other things there or was absolutely everything just AI AI AI?

[00:05:07]>> Uh well I mean yeah it's mostly AI. I mean pretty much what you saw on the keynote was the real priority for Google continues to be the priority. It's part of the reason why they do the Android show a week in advance is because it's kind of like clearing the deck for the thing that's their real, you know, concrete uh large scale uh effort and commitment. Um I wasn't surprised at all that there was no Google book there.

[00:05:30]They were really tight lipped on on Google Book the week before and I I would have been really surprised that they'd be that that tight-lipped and then only a week later show you the the full deal.

[00:05:39]>> Um so they're going to hold on to that and build some excitement around that.

[00:05:43]they did have glasses and so you know that was one thing that they talked about on stage was the Android XR with Sheram Mazati um who leads that team and you know and so there were ways while you were there to try out their AI glasses. The thing that's interesting about that though is like we've now seen at two, this is the third consecutive Google IO where we've seen the glasses, >> right? The first one it was >> it was a super tease, right? It was that one video of of uh Google contextualizing with with what eventually became um Gemini Live, but at the time it wasn't called Gemini Live.

[00:06:22]And uh and then it, you know, the the woman's identifying a bunch of things and then she picks up glasses and that was the big wow moment in the audience was like, "Oh my goodness, the glasses can see things and, you know, contextualize and you could talk to them. That's amazing." And then last year we saw them and you could demo them if you were there. You could test them out. Here we are a year later and it's still like, "Yep, coming soon." And they gave you more information about the Warby Parker and Gentle Monster um collaboration for frames and they showed off what the you know a couple of those frames look like.

[00:06:55]>> Did they give you a price?

[00:06:57]>> No, no, no, no. We have no details.

[00:06:59]>> That doesn't make it real enough yet.

[00:07:00]Yeah.

[00:07:01]>> No, it's still kind of like coming in the fall.

[00:07:03]>> I was surprised that a lot of the consumer press coverage led with the glasses.

[00:07:08]>> Yeah.

[00:07:08]>> I'm gonna ask you in a second what your main headline was from from the event. U compare headlines. Uh, but the glasses is catching up to meta. It's fine. I'm glad it's there.

[00:07:18]>> It is catching up to meta. That That's kind of my point. Exactly. It's kind of like two years later.

[00:07:22]>> So fine. You should catch up. That's okay.

[00:07:25]>> But uh >> which and and also which is not to discount my excitement about the glasses because >> Agreed. Agreed. I'm glad they have they should.

[00:07:33]>> Yeah. I mean, you know, I have the the Meta Oakleys and they're fine. Like for what they do, they're fine. I But I don't care for meta AI. Like I just that's not >> and it's not tied into the rest of your ecosystem of information. This thing will be tied into your email and your maps and your preferences and all the things they showed about Gemini. So okay, >> it's tied tied into that. It's also a familiar um chatbot experience for me.

[00:07:57]Like I'm very familiar with with Gemini Live. I'm very familiar with Gemini and kind of the >> the ways that it you know the strengths that it has and how to interact with it and everything. And I did test them out again this time. the the um moninocular version, which is the one display version. And yeah, and it was still the prototype hardware. It wasn't the Warby Parker, the gentle monster frames >> because we don't know when the when the um binocular version is coming next year, we think, right?

[00:08:25]>> Right. Right. They're saying the audio glasses. Um that's what they were calling them, right? Audio glasses.

[00:08:32]Yeah, I think so. Something like that.

[00:08:33]>> Later this year.

[00:08:33]>> Um >> it was a term that I hadn't heard before. I was like, "Oh, is that what you're calling it?" Um, yeah, that's that's later this year. And then the display, like I was talking to someone, um, Max Spear, who was showing me the glasses, and I want to say that he's he's mentioned something about by the end of the year, the display ones might come, but still, >> I don't believe it. We'll go to the We'll be at the fourth IO to see those glasses.

[00:08:56]>> Yeah. But I will say, having experienced them multiple times now, they're pretty darn cool. Good.

[00:09:01]>> Um, and I look forward to that when it happens.

[00:09:04]>> But there's bigger news. So, I'm curious, what was your headline from the whole event?

[00:09:08]>> I think my headline, what I kept coming back to throughout the day is Google's strategy with AI is bringing everything to everything. It's like it's like this two-tiered system of all right, this year we've got all this new stuff that's coming to Gemini, and last year's new stuff that came to Gemini now goes to everything else. And so it's constantly like every which which is a little confusing when you were talking and we had this conversation last year about AI mode and like why would I use the AI in AI mode when it does all the same things that the AI in Gemini does mostly but it's inside of the search product not Gemini. And I mean I mean I do use AI mode. It's I I think my point is that so many of these products do a lot of the same things. So the question that I always or the confusion that I often have is like well why do I choose to use it over here instead of over there right >> or over there instead of over here and I think Google's just philosophy or approach is just like Gemini is the effort you know 3.5 flash was the big news that they announced um their new model you know better and uh they say frontier intelligence with action so agentic workflows not just answering questions that sort of stuff um that powering everything and a lot of those capabilities that it's capable of just filtering into everything. Um, so everything being able to use everything eventually, you know, that's that's always the the goal of Google with its AI right now.

[00:10:43]>> Yep. So my headline >> is that I think Google bec is is positioning itself as the retail AI company.

[00:10:51]>> The retail AI >> the B2B the B TOC retail by by B toC.

[00:10:56]Right. Sure. So at the beginning I was I was interested that at the beginning they started heavy on the tech. They started with their new tensor chip. They started with command line stuff things that vibe coders would do to compete with where anthropic is and where openi wants to be.

[00:11:14]>> Yeah. Maybe they were making they were uh start giving a soft landing in the beginning to the developer community that the conference is actually geared towards.

[00:11:22]>> Exactly. Even though the keynote often isn't geared towards the developers, they're throwing them a bone at the very top.

[00:11:27]>> Yep. Yep. Exactly. Um but then it turned around and I think everything that they do is about bringing AI to wherever you are.

[00:11:36]>> Yeah.

[00:11:37]>> And that's kind of my point. Yeah.

[00:11:39]Exactly. That >> you don't need to use command line.

[00:11:41]You're going to use their agent stuff to have it do things and basically write code and do what you want to do and not u think of it as vibe coding.

[00:11:51]>> Yep.

[00:11:52]um you're going to uh they're and and and you're right too that they put it everywhere so that they're going to transform search.

[00:12:01]>> You go back to the day when when the book um >> uh filter bubbles came out the contention was that everybody got a personalized different Google and it separates us all. And the research says no actually that wasn't the case at all that we all pretty much got the same Google responses. Well, no more. Now, Google's going to be highly personalized to your information. Um, you're going to be able to create ongoing agents to do things for you and get alerted to things. Um, and >> yeah, agents all over the place, right?

[00:12:32]Like that's a that's a big headline for sure.

[00:12:34]>> And that's vibe coding with the vibe and not the coding, >> right? You don't have to worry about I'm creating an app and I have to run it or any of that. Um, uh, it'll be up in the cloud as I think you mentioned in one of your your pieces about you walking around with your laptop open.

[00:12:48]>> Spark. You're talking about Spark. Yeah, it's like the 247 agent, uh, running in the cloud, not just on your device.

[00:12:54]>> So, we'll go through all these details, but um, I don't want to I don't want to steamroll over the details because I think they're fascinating, but yeah, I think I think it just you're right that it's everywhere and I contend that it's everywhere for everybody, >> right? And um I haven't wanted to get into vibe. I mean I haven't had that much use for it. I use it for things like my book uh to put research together and other things. I use it to summarize stuff. I haven't used it yet to make actions which is what vibe coding does.

[00:13:26]>> Really good. Uh theic internet. Yes. I think so.

[00:13:29]>> That's the that's a big Yeah. Another big takeaway, right? is when you're talking about agents and you're talking about the capabilities of these models and then we're talking about AI mode and search and kind of the blurring between uh what was once AI mode now coming into the new search box and and all these things and then you bring agents into search. I mean search as a product is changing rapidly. Not not that we didn't know this already, but yesterday made that really clear like and and this is this is one of the points that I definitely walked away with is often on this show we have talked about the um the impact the potential impact of agents on let's say a smartphone and the app ecosystem as we know it right now.

[00:14:15]Users are very used to opening their phone. They have it they have a task in mind. They open the app to do the task.

[00:14:22]they go and they do it and the app is the conduit through which that happens.

[00:14:26]And then there is this, you know, future that keeps being um presented to us and and kind of seems like a direction that things are heading where the agents are working underneath to make it so that you don't have to go to those apps that the agents just kind of go to the places and do the things. And now we've got agents coming into the search experience. And you know for the last however many years people have been concerned about the impact of AI on search and does it reduce page views and reduce users going actually to the pages where the ads are and everything. Now you got agents working inside of search basically doing the same thing that we're talking about with apps on a phone. More and more that search becomes the conduit for the action to be done.

[00:15:10]If I open a browser because I want to carry out an action which is to go to this website and do this thing, I maybe no longer have to do that because the agent does that for me. What does that do to the internet? Like what does that do to the internet the way we know it?

[00:15:24]I'm not saying it's good or bad or anything. I'm just saying that's the change that we are facing right now with as these features become more prevalent.

[00:15:33]>> Yeah. Um and in Google has always been a search company >> and so that heritage continues strongly.

[00:15:43]It is the it is the lens through which you get to AI. I think most often not all the time you're going to you're going to use it in your email. You're going to use it in your sheets and wherever else.

[00:15:51]>> Yeah.

[00:15:51]>> But but search stays centered and I think that's really important.

[00:15:55]>> Um >> and Google's so in a few ways Google's in the absolute best position here.

[00:16:04]uh right they have search they have you an ecosystem that you're part of they um the the nervousness about all the money that open AAI and Anthropic have to pay for hosting well Google is the host they're earning money from those guys too at the same time >> yeah yeah >> right and and more of this is going to shift over to their tensor chips from from Nvidia alone >> um they're I think they're extremely well positioned much better than beta >> um and much better than Apple. It'll be really interesting to see what happens to WWDC where Siri is based on Google and will the power that was demonstrated yesterday for Google users come through Gemini into Siri at Apple >> will be interesting. Or will Apple try to tamp it down um because it's building its own thing or they've given up and said no we're going to get we're going to be equal to that but it's still Google underneath >> or does Google want that on its own phones? you know, the the the absolute maximum, you know, >> Yeah. That's right. Just Google tamp it down. That's right. That's right.

[00:17:05]>> Yeah.

[00:17:06]>> So, um Google stock went down 2% yesterday. It's down 3/4 of a percent right now, which this is the dumb market.

[00:17:14]>> Um I think >> that's what do you think drives that?

[00:17:17]>> Who knows? Who knows? Um you know, as the old saw goes, you you buy on the news. You you buy on >> what is it? You buy on the anticipation of sell on the news, >> right? Um.

[00:17:29]>> Mhm.

[00:17:30]>> Nvidia, I'm just looking at the map right now. Nvidia is up 2%, Amazon is up a percent and a half. So why is Google down? Um >> I I they're going to be investing a lot.

[00:17:38]Maybe that's it. But I think um I think it was a damn good day for Google.

[00:17:45]>> I thought so. I thought I thought that was a lot of really uh interesting um compelling news. And like you said, the products that you're already using getting very notable and and meaningful uh new features. Uh had someone show off kind of image editing. What is it called? Google Pix or something.

[00:18:07]>> Pick. Yeah, >> pick inside of uh inside of the kind of like Drive ecosystem. Um creating images, you know, creating a slideshow just directly inside of Drive. uh inside of the drive uh experience and then having certain elements of the picture be not layered but be detected by the image generation model so that you can be like all of this is great but I want this word to actually say this instead of that. And now they've gotten it to the point to where you can just kind of click that area and tell it what you want it to say there and it will keep everything else the same but change that like the iterative process is is a lot more dialed in there. that's a user like that's a really solid user choice. You don't have to know how to learn the language at this point. I guess that's kind of my point is that these models are integrating in a way to which hopefully for the general consumer the norms the normies that aren't you know following this stuff deeply on a technical level still understand how to get the most benefit out of them >> right >> based on how they're building them. And I think they're getting there with these tools. um they're not there entirely, but they're certainly getting there. But that does run me up against another kind of question that I have is, you know, also in search, it's not just agents, it's also coding, like right and and they're they're talking about vibe coding inside of search, which you're like, why the hell would I want a vibe code in search? But the idea is that like if you're looking for an answer or you're looking for a synthesis of information or whatever, instead of having to poke through and and you know have it collect into a doc or whatever, what if it just presents to you an on the-fly created interactive app essentially that allows you to then go in and play with it immediately with controls that it just built. And um I wonder how they're going to present that to the everyday user. Are they going to go the route of code your own app or are they going to go the route of, you know, an interactive multidia, what a throwback term, uh, presentation that helps you understand.

[00:20:15]>> Yeah. Multi. But I mean, even that might lose the norms, normies out there.

[00:20:19]>> Yep. That's right.

[00:20:20]>> My my point is like these capabilities are increasing. How do you make that accessible to even the people who might be scared from the word coding? You know what I mean? like, oh, >> vibe code, I can vibe code in search is going to mean something to somebody, but it's going to scare other people away, but yet there's usefulness there. So, how do you frame that in a way that makes them curious instead of afraid of it?

[00:20:45]>> Yeah, but that's where I come back. I think I think it's the BTOC AI company. It's the retail AI company.

[00:20:52]Um, you won't think you never have to do command line. You won't think you're making an app.

[00:20:57]>> You're asking it to do something and it's doing it.

[00:21:00]Um, and and it's not just a one-time question.

[00:21:04]I was thinking yesterday, I mean, I've changed my search behavior already. I used to be put in a keyword. Now, even in the search box they've had for some time, it's more of of a prompt.

[00:21:15]>> It's a fuller question or it's a command or it's whatever. Right? So, we've already kind of gone that next step, >> right?

[00:21:22]>> Um, but that was still to return some information to me. Mhm.

[00:21:26]>> Now I'm going to be able to return action to me.

[00:21:30]>> Yeah.

[00:21:31]>> And and I and I and I don't have to think about an app coding, vibe coding, command line, any of that. I think that's powerful.

[00:21:40]>> Yeah, I think it is. It's uh real compelling uh presentation of these tools that we're so familiar with. I'm trying to see if maybe some of this uh if you're watching the video version, this video might kind of show you some of the directions of uh what happens is what is the risk of Google, you know, because Google pointed out this is the first major change to the to Google's search box in 25 years. They said is what is the what is the risk there?

[00:22:08]Because I mean that's you know that's a quarter of a century. People have been using Google search very one very particular way. Are people Yeah. Do people want this? Are people ready for this?

[00:22:19]>> I guess is a big question I have.

[00:22:21]>> Well, you know, in a sense, it'll for some it'll operate the way it's always operated.

[00:22:24]>> Yeah. Yeah. I mean, it's not like it doesn't operate that way anymore.

[00:22:28]>> So, the hard part for Google is to is to tell you, hey, hey, you could also do this. People won't know it. They won't think of it that way.

[00:22:35]>> Um, now, in a sense, Open AI >> and and Claude have taught people new behaviors, but that's only a portion of the population.

[00:22:43]>> Yeah. And that's still a chat uh interface chat to chat rather than chat to action.

[00:22:53]>> The to get action out of claude or open AI you had to vibe code.

[00:22:58]>> Yeah.

[00:22:58]>> So this is it'll be really interesting to see how they um try to train people here.

[00:23:03]>> Yeah, for sure.

[00:23:05]>> By the way, >> uh yeah, >> just just to have my because it's a it's a standing feature on this show. Mhm.

[00:23:12]>> I'm about to have a raging sit.

[00:23:15]>> Oh, great. Well, maybe I'll go to uh the big shot here.

[00:23:19]>> I don't have 35 in my Chromebook work place. I went to a a uh to my regular old Gmail account. And there.

[00:23:28]Yep, there it is. 35 right there.

[00:23:31]Wonderful. It's not in this. I can't see where I'm supposed to do anything. It's driving me bananas.

[00:23:37]>> Oh. So anyway, I just want to do that because go >> come on. Give Jeff all the things in all the places.

[00:23:46]>> Yeah, >> seriously.

[00:23:47]>> So I think I now we go back to our regular programming. Um >> got to get that in there as well.

[00:23:51]>> So we should probably go through, you know, make sure we've covered everything they did. So So um we talked about search >> a lot. It's a lot.

[00:23:57]>> We talked about um >> yeah, >> go ahead. flash mostly, but I mean Gemini Omni Flash was another aspect of this, which is kind of a image and and video uh generation.

[00:24:09]>> Although, it was interesting. I went to a kind of a gathering, an on there gathering last night for um for early access Gemini uh folk like myself. I have early access to some of these tools, which of course I can't talk about until the embargo has lifted.

[00:24:26]That's why I don't share things with you that you've never heard before because I'm not allowed to. But um but I do belong to >> folks Google. He actually doesn't even tell me. I'm pissed as hell, but he doesn't even tell me. Here I am as partner on an AI podcast and I don't know.

[00:24:41]>> I try to. I just don't.

[00:24:42]>> So he's he's he's being he's being legit. He's being good. I'm telling you the truth.

[00:24:47]>> Uh I only had access to Omni for like a day beforehand. So So there's that. And and I wouldn't say that like actually do I even have it up here? I could I could show you the uh the outcome of of Omni.

[00:25:01]Let me see here. Give me one second. I'm vamping while I pull this up. And I you know, maybe now everybody has access to Omni, but I I put in two images of me and you and uh I said Jason and Jeff find themselves standing next to a robotic embodiment of either models >> at at Google. I I know. Well, I did only use one image of each of us.

[00:25:25]>> Uhhuh.

[00:25:25]>> And really what Google is saying is that Omni is like nano banana for video, which I don't know what the difference is between that and VO. VO not mentioned at all. I kind of think Omni is is doing to VO. What Nano Banana did to God, what was before it? It was um Imagine, wasn't it? Imagine before Nano Banana Imagin or whatever >> and then Nano Banana came out and it was like okay, we don't need that other thing anymore. I think that's what's happening here. But if I had fed it more context, maybe they it would actually look more like us. And you can't hear it, but I'm just I don't want to slow down the show. But yeah, it only sort of looks like us.

[00:26:04]>> No, not at all. It looks like Brian Serest or something.

[00:26:09]>> Uh anyways, so yeah, so this is their video generation. Um they say improve physics, consistent character rendering, although you know, remains to be seen. I think, like I said, if I had loaded this with a bunch of examples, it might have gotten it closer. Um, and yeah, synth ID watermarking in all the videos created here. They also have a little bit of a coalition uh forming around synth ID now that 11 Labs, Nvidia, uh, Cacao, and OpenAI is now uh, part of the Synth ID crew. So, you know, there's that.

[00:26:44]Um, we did the search box.

[00:26:49]We did search agents. Agentic coding in search. I'm just kind of like scanning through here to see Gemini Spark. We mentioned Spark, but we didn't really dive into this.

[00:27:00]>> Yeah, talk about tell people about that.

[00:27:02]>> This is one that I'm actually pretty excited about primarily because I've been using, as I've talked about, Claude Co-work a lot. And Cloud Co is this like aentic AI platform that is on my device. You know, it's it's on my laptop. It's tied to my laptop or it's tied to my Mac Studio. Um if I create a a Cloud Co-work on one device, it doesn't necessarily talk between devices. So, it's very device specific, not as portable. Um but I can assign it schedule task and have it go out there at any time. Given that the computer is powered on, given that my laptop is open, it will do those things when I set the time >> because you're running it locally.

[00:27:44]>> Yes. And it's running locally. I mean, it's still using cloud in the cloud, but it's driving a lot of that agentic stuff locally. Uh Gemini Spark is Google's 247 AI agent happens in the cloud uh on their uh Google Cloud servers um tailored specifically for this use case.

[00:28:03]So, in other words, it's basically AI agents that operate in the cloud 24/7 without you, you know, being tied to a device. So, I could open it on multiple devices. I could have it run even though my computer is off. Um, it will not, it will never not run. In other words, I think that's pretty uh pretty interesting. Although there was a part of me that really liked the ondevice thing because we often talk about like >> pulling away our reliance on the cloud to to go local.

[00:28:33]>> But that's a very geek thing. That's a that's a that's a B2B thing.

[00:28:37]>> Yeah.

[00:28:38]>> That's a geek thing. That's a command line think kind of thing. That's not a consumer thing.

[00:28:42]>> This is very Google of It's very Google consumer oriented way to look at AI.

[00:28:47]>> Yeah.

[00:28:47]>> Once again, >> I think it's important >> to your point.

[00:28:50]>> By the way, correction here, not correction update. I restarted the machine. Now I have it.

[00:28:55]>> Oh, hey, there you go.

[00:28:56]>> All is forgiven for now. Until the next time.

[00:28:59]>> Until the next time.

[00:29:01]>> Good. I can't wait to play with it more.

[00:29:02]I've been all night. I've been thinking, "Oh, when is it going to come? When is it going to come? I want to play with it for the show. I couldn't play with it."

[00:29:07]Now I will.

[00:29:08]>> Well, all right. So, so, uh, Spark, what else do we have? Oh, >> integrates with MCP, by the way. So, you're going to get your, uh, context protocol connections there. That's great. Coming into the Chrome, uh, browser directly >> soon. Uh, and then also coming to the phone via Android Halo, which was one of the few moments that they mentioned Android on stage. And they really didn't even talk about that other than to say that. And eventually coming to the phone via Android Halo. Um, so that's coming this summer. So I'm excited about that.

[00:29:38]I'm definitely interested in playing around with that. Uh, universal cart, which is >> Yeah, I think this is important. There there there are three um protocols on commerce really add up. Okay. Yeah. So, tell me tell me tell me about your your thinking it's its importance here. It's um I mean and I'm not saying that to say I don't think it's important. I'm just curious to know what you think. Um >> it's a universal it's kind of an MCP for um commerce. So, universal carp then the agent payment protocol so your agent can pay for you and then the universal cart.

[00:30:13]Now the agent payment protocol is that meant to be >> it pulls the trigger, it actually pays.

[00:30:20]Yeah. Okay. That's my understanding.

[00:30:21]>> You give it the authority.

[00:30:23]>> Um I'll I'll credit this to to Leo Leaport as we were on the show. He said he kept coming back to >> and this is so different from the early early early days of Google. They don't want you to leave.

[00:30:33]>> Yeah.

[00:30:34]>> So what they demonstrated on the stage is you bought something from Best Buy and you never went to Best Buy.

[00:30:38]>> Yeah. Yeah.

[00:30:39]>> Right. your agent found it for you, did whatever you wanted in terms of comparisons or whatever else and had the authority to pay and paid and bought it.

[00:30:47]>> And I think that's that's if I were I mean Amazon's sitting there saying, "How can we cut this off?" But I think I think consumers will say, "Well, then I'll screw you Amazon."

[00:30:58]>> Yeah, that's interesting. If consumers trust it >> if they trust it. But but you know, I I'm old enough to remember when people thought nobody was going to buy anything online.

[00:31:06]>> Oh, yeah. I remember. Then nobody was going to buy anything expensive online.

[00:31:10]Uh nobody was going to buy clothes online and all eventually they all they do.

[00:31:15]>> Eventually they do. I'm And I'm not saying that consumers never will. But I am saying that it might take a while cuz that took a while. I remember you know like I I witnessed it with my parents like it took them a while for them to get to it took them years before they felt comfortable buying anything online.

[00:31:31]>> Yep. Yep.

[00:31:32]>> And uh you know it's it's kind of like you don't do it until you do. and how you get to the do part, you know, really depends. Everybody's different and, you know, it always comes back down to trust, too. Uh, agents, AI, and trust have bad reputation among a lot of people. So, the idea of giving giving an agent full um, you know, cart launch to pull the trigger for you, >> that's going to take a lot that's going to take a good number of people a considerable amount of time before they feel comfortable with that. That's probably only going to be proven by the fact that many people they know do it all the time and continue to say, "Oh, yeah, there's not you have nothing to worry about. I've been doing it for years or whatever, you know, before they're like, >> well, I already have my credit card in Google Wallet, right?"

[00:32:16]>> Yeah.

[00:32:16]>> And I use that to pay now. And uh I can see having it check with me before it does it if there's something like every month I want to buy.

[00:32:25]>> Uh and the first time it it all happens, okay, and the second time it happens, okay, and the third time I say, "Okay, go ahead and do it on your own."

[00:32:32]>> Right. M you know and and Amazon of course was there with with the little buttons they discontinued you put this button the tide button next to your washing machine.

[00:32:40]>> Oh >> and hit it.

[00:32:41]>> May it rest in peace.

[00:32:42]>> Yes. Exactly. Um but that was that was an agent for you in essence.

[00:32:47]>> Yeah.

[00:32:47]>> Right. So >> well you had to trigger the agent but uh you were the >> you were the this in the if this then that equation, >> right?

[00:32:56]>> Right. This >> then that. Yeah. Um, I think I think the other thing that I I didn't get enough about yesterday, Jason, was was um because search becomes personal.

[00:33:08]>> Um, I'm not sure how you give it your standing context.

[00:33:15]>> Um, >> okay. Explain. I'm I'm trying to follow.

[00:33:17]>> Uh, I'm a I'm a teacher and whenever I ask you about uh how to explain things, you should give it to me in the context as if I'm going to teach a class on it.

[00:33:26]I'm making this up. Right.

[00:33:27]>> Okay. And that's the kind of thing that you give as an instruction >> into claude or open AAI, >> right? You can put that into the system instruction.

[00:33:34]>> You can put that here. But but but because it's personal search, how much can you have impact on what it remembers about you and how it acts on that?

[00:33:44]>> H it's a good question. I mean I do know >> hear an answer to that yesterday.

[00:33:48]>> Yeah, that's a good question. I because we have known for a while because Google ruled this out I don't know how many months ago the personal intelligence kind of uh layer into Gemini, which is essentially if you opt into it and you tell it like yes, you can you can have access to my Gmail and my drive and blah blah blah and you know all those sources help inform you as far as who I am, what's important to me and and become data points, you know, my photos, all that kind of stuff. And I do know that parts of that personal intelligence do loop into the search experience. um you know in in this kind of new new search experience >> how that transforms that experience to what you're talking about I that I don't know yeah does it actually change how things are presented to you uh at a foundational level >> you question you know do you have the chance to influence it and tell it from now on do this >> yeah that's a good question right you're in the search box and you're like you know what going forward when I do this search never show this particular outlet or ever show this website or >> right never tell me I'm smart for asking this.

[00:34:59]>> Yeah, that would be kind of interesting though. Could you like maybe that's already possible but >> that's what I don't know. Um I mean it's I've long dur in the world of personalization and fears about it with privacy. I've long argued that companies should give you the opportunity should tell you what they know about you how they use it and give you the opportunity to change it. They would only learn more about you. They would only learn more, you know. Stop telling me about this. I bought that for my daughter. Stop. Stop.

[00:35:27]>> This is how we use AI. This is how we use LLMs now. This is how I build agents now is >> I I you know, tell it what I need. It spits out something at me and I go, "That's close, but this is what I want you to do differently." And all of that preference.

[00:35:43]>> Can you have Can you have an input on that? Can you make it persistent?

[00:35:47]>> Can you perhaps have more than one persona? Okay, this is work Jeff talking to you. This is fun Jeff talking to you.

[00:35:53]>> Right.

[00:35:53]>> Right. Right.

[00:35:54]>> Um that's where there's a big opportunity for Google I think >> to make they talk about personal search.

[00:35:59]It's no it is a filter bubble now. Okay.

[00:36:01]Let me let me go full boore with that and see what that looks like.

[00:36:04]>> Um >> that's really interesting idea. I hadn't hadn't considered that. Yeah.

[00:36:08]>> Two other things. One is they announced uh the Google for science research which they didn't spend a lot of time on but I think that's that's that's a a salute >> to how this is being used. I think it's important. And finally, I found it interesting that um Devisabas came out twice and closed the event.

[00:36:28]>> I thought the close was was kind of everybody in the audience, everybody was like, "Oh, it's done."

[00:36:35]>> Yeah.

[00:36:36]>> Oh, like it was early.

[00:36:38]>> It didn't Well, it wasn't that it was early. It was however he chose to close it didn't feel like a closer.

[00:36:43]>> No.

[00:36:44]>> To anyone there. I don't know how it came across in the live stream, but suddenly he was walking off the stage and it was like, >> "Oh, wait a minute. That's how he closed it."

[00:36:52]>> Yeah.

[00:36:53]>> Okay. I guess we're done. I guess. Okay.

[00:36:55]Let's file out now. It was It was I can't remember. I kind of missed a few lines that he came up with. Nle Patel put up a video uh quoting him saying, you know, the most the craziest thing I heard at OpenAI and it was three quotes from Demesis Avis saying that we're nearly at AGI and we're at the singularity.

[00:37:13]>> Yes. I kind of missed him saying that which like >> but yesterday I speculated uh when watching it that uh are they setting up Deis Sabus to be uh Sundar's successor?

[00:37:28]>> Speculation but that was a pretty prominent position.

[00:37:32]>> Pretty compelling.

[00:37:33]>> I don't even do you know whether it was was Demis there last year or the year before?

[00:37:37]>> Yeah, he he's okay. All right. All right. I didn't know that. Uh but I thought it was a very prominent position to be the closer >> in essence. Yeah. Uh if it's an AI company and he's your AI guy.

[00:37:48]>> Yeah.

[00:37:49]>> Um >> yeah, makes a lot of sense, doesn't it?

[00:37:52]That's a really interesting idea. I wouldn't be surprised about that at all.

[00:37:56]>> I mean, I'm not to say Sundar, you know, can be president of Alphabet and they could they could split him up again in essence. Um uh it's just speculation.

[00:38:04]Just wondering but I think I think the Demis is clearly a star there and they were they must be very happy they bought DeepMind back in the day.

[00:38:10]>> Oh heck yeah.

[00:38:11]>> Buying Deep Mind buying YouTube and these were incred buying double click these were u changing transactions for the company.

[00:38:20]>> Yes matter. So uh I thought it was I thought it was a good day.

[00:38:24]>> Yep. Lots of stuff. Uh before we end things I do want to show you that something that I think you'll appreciate. Jeff, when I was in the AI sandbox, the center of the AI sandbox was Genie, which is Google's uh world model demonstration.

[00:38:42]>> Ah, >> and so you could pick your character or you could choose to create your own and then you pick kind of a scenario environment, I guess, or choose to do your own. I think I picked what did I pick? Uh that's that's Ron Richards. I chose the like a spinning top and the create my own uh environment which ended up being a podcast studio. And basically for those who don't remember and it's just loading if you're watching the video version it'll get there. Um this is the setup the system that generates an interactive world in real time. And so you have a couple of joysticks and like a jump button. And you could kind of consider this as like it creates a 60-second game for you on the fly based on the inputs that you give it in a generative sense. And so I basically said take this thimble and uh put it into a podcast studio. So it's loading.

[00:39:36]I'm trying to get to the actual demo.

[00:39:40]Um come on. Apparently, it comes very late in the video and just vamping >> YouTube. What did I just say? Nice things about YouTube. Come on.

[00:39:52]>> Yeah, come on. Come on. But anyways, it creates it all in real time. I would say my my playing around with it. You know, you can see it's a little a little choppy, but it is creating this, you know, live as I'm moving around all it's not like it this is a pre-programmed world. It just kind of took the the spinning top and created this podcast studio environment. And now I'm >> But the spinning top is is not operating like a spinning top does.

[00:40:21]>> Has to hold to a surface and it doesn't.

[00:40:24]>> Yeah. It's >> so ignoring the gravity.

[00:40:27]>> It is, but it's it's capturing a lot of the like reflections of the room and everything which was kind of kind of neat.

[00:40:33]>> So this is the difference I think yesterday. This is this is my Yan Lun.

[00:40:36]Um, and you just bumped off something off the table and it didn't fall the way gravity would have it fall.

[00:40:41]>> It flew.

[00:40:43]>> Um, so what what strikes me here is that what Google was showing off yesterday, and there's Jason's hand.

[00:40:49]>> Um, is uh that they're showing impressive madeup worlds.

[00:40:56]>> Yeah.

[00:40:57]>> But the key to a world model is to deal with the real world.

[00:41:01]>> The real world >> under the rules of the real world. and that a spinning top has to be on a surface to spin and when a ball falls off a table it falls in a certain pattern because of gravity. Right.

[00:41:12]>> Yeah.

[00:41:12]>> And so I think it was this is speculative or fictional or I don't know what to say ren rendering of imaginary worlds. I wouldn't call it real.

[00:41:23]>> No, >> it's a world model but not a real world model. And so I think it'll be interesting to see, you know, what Lun does and Fe Lee uh in the future. Um because that what you just showed is fun, but that wouldn't help you if you're a self-driving car because it thinks the car can fly, >> right? Yeah. What are the what are the confines in which it actually operates?

[00:41:45]Yeah.

[00:41:46]>> Um hey, everything starts somewhere, I suppose, >> right? It's a constrained how does it work in a constrained world?

[00:41:53]>> Yeah.

[00:41:53]>> Is the key to role models. and Yan Luna said that.

[00:41:56]>> Mhm. Yeah. But I was excited to see that. I was like, of all the things, I was like, "Okay, I got to check that out because there's no other way to test this." And uh so there was a little bit of a long line waiting for that.

[00:42:09]>> All in all though, it was a great uh Google IO. It's still going on today. I just usually don't go day two because I get everything I need in day one.

[00:42:16]>> You miss me. I know you miss me. You want to be here. Yeah, >> that's right. I miss you, Jeff. And I don't want to miss the podcast as well.

[00:42:22]So that is Google >> folks >> and all of you which speaking of all of you patreon.comiincid show you know we got things happening at the Patreon even for free subscribers so you should go and and join uh for free and you'll get some extra bonus stuff.

[00:42:39]Uh and I do want to throw out some thank yous because we have had an influx of new patrons um which I'm happy to see.

[00:42:47]Captain Caveman, Counterpoint, Richard Riker, Lisa Edwards, Ryan Lounsberg.

[00:42:53]>> What's that?

[00:42:53]>> Michael Riker.

[00:42:54]>> Michael Richard.

[00:42:55]>> What did I say?

[00:42:56]>> Richard.

[00:42:56]>> Did I say Michelle?

[00:42:57]>> No, you said Richard. It's okay.

[00:42:59]>> Oh. Oh, okay.

[00:43:01]>> Cuz Riker looks like Richard. That's okay.

[00:43:03]>> Got it. Okay. Michael Riker.

[00:43:05]>> Lisa Edwards.

[00:43:06]>> Lisa Edwards. Ryan Lounsbury. There you go. patreon.com/aiidshow show. Great to have you all on board.

[00:43:13]Hopefully we can convince more of you to get on board. Just saying that daily podcast it exists.

[00:43:19]>> Jeff's gonna answer his >> his old school phone. Is that an actual ring or is that a digital representation of an actual ring? Okay. All right.

[00:43:29]Right.

[00:43:29]>> But yes, there's a certain age at which it's comforting for us old folks.

[00:43:33]>> I do remember when phones sounded like that. We're going to take a break and come back and talk about some other news that uh maybe has something to do with Google, but mostly not. That's coming up here in a moment.

[00:43:48]All right. Uh let's see here. We've got an Andre Karpathy joining Anthropic.

[00:43:56]This is the OpenAI co-founder. Andre uh Karpathy. Um yeah, becoming an Anthropic employee, one of the more respected AI uh AI researchers uh working today. Like I said, founding member of OpenAI, former director of AI at Tesla. Um, yep. He coined the term vibe coding. So whether you love it or hate it, he's a visionary in this. Yes.

[00:44:20]>> Yeah.

[00:44:21]>> So he's going to work on uh helping Claude uh it's using Claude itself to accelerate pre-training research.

[00:44:28]>> And this is according to Axio an increasingly important frontier as AI companies race to automate parts of AI development.

[00:44:35]>> So this is the step toward AI making AI.

[00:44:37]>> Yeah. Having the having the model help improve itself, help build itself.

[00:44:42]They're all doing this to some degree.

[00:44:44]And I think even Google yesterday mentioned something about um I can't remember what it was tied to about how um anti-gravity was used to create one of their other things. I wish I had written it down. I can't remember off the top of my head. But yeah, they're all doing this. They're all using their eye to their AI to somehow iterate, somehow, you know, check the logs or, you know, they're integrated in multiple uh different ways. And so that sounds like what um Andre is going to be doing.

[00:45:14]He says uh on X quote, "I think the next few years at the frontier of LLMs will be especially formative. I'm very excited to join the team here and get back to R&D and uh yeah, it's pretty significant hire. I mean, he's a he's a notable name and Anthropic is a notable company right now. Um they have the poll to bring him in and put him to work." So could yeah I'll be curious to see what that leads to >> uh for Anthropic. Speaking of Anthropic, uh acquired Stainless. Stainless is a New York startup founded in 2022 uh by a former Stripe engineer Alex Retree Ratree. Um it's being used by OpenAI, Google, Cloudflare, Runway, Replicate. Um which a number of those, you know, clearly that's Anthropic's uh competitors. And now Anthropic owns this company, acquired it for a reported $300 million, though I don't think that they announced an official amount on this. Uh they are winding down their hosted products so competitors can keep the SDKs that they already generated. they lose access going forward. Anthropic says Stainless has powered every official Anthropic SDK since the early earliest days of their API. So basically they're pulling this key piece of infrastructure uh out of common access and kind of locking it down into their own walls and that means that Anthropic's direct competitors are going to have to build or find alternatives uh as a result. So, right, that's a that's a power move right there. But I guess if they're using it and relying on it so much, I guess it makes sense. They've they've got an insane war chest right now. So, yeah, I don't know what that means for all the all the companies that have relied on this on this product suddenly having to to work around it, but that's what they'll have to do. Uh >> yeah, it's interesting these two stories do go together because it's buying people and buying apps by companies that they're they're using this VC money to as a territory grab. Uh which is okay. I get that. But um Google's buying less and right now doing more. So we'll see.

[00:47:36]Yep. Uh let's see here. Boop. Gonna vamp for just a quick second because I want to God. trying to figure out how I'm I'm realizing my podcast art is on a different machine than the one that I have here. Okay. Well, what I wanted to show off real quick is because we were talking about anthropic claw design, which we mentioned a number of weeks ago when Claude first started rolling it out and I had never used it and I had the credits and everything.

[00:48:07]So, I was like, I want to redesign the AI inside podcast art, which okay, this is a this is a horrible way to show it off, but uh there it is. It's just really small and I can't make it larger.

[00:48:19]So, uh, well, maybe maybe I can >> We're going to put on a postage stamp soon.

[00:48:22]>> Hey, there we go. I'll just zoom in on it. It's not the best way to show off the new podcast art, but this is where we got. But at the beginning, I was like, you know, the podcast art, it's like there's no personality. It's just the logo. It's really dark. It's got this like dated AI circuitry board behind it. Like, there's a number of things that I don't like about it. and I just wanted I'm curious to know what Cloud Design would would inform me, you know, to do. And so I launched Claude Design and started going back and forth with it. I just basically said like, I want to redesign this podcast art. Here was an idea. You know, this was like an initial idea that I was playing around with uh when I first started kind of thinking about what I do. It's like, okay, I'll build you a proper design review. Um, let's see if I can get to There's got to be Oh, yeah. So the first, so this is what I kind of fed it the first draft. Uh it said cluttered circuit board cliche. Then it came back with something that informed me into this direction and then ultimately you know I landed here and then eventually I took it to Patreon and ended up putting a circle around the logo. Sorry for audio listeners. This is very visual. Um but what I want to show and this is not showing me the information. Let's see here. Was this it? Okay. So, this was my first review. No, that's not it. That's not it. Where is it? I'm sorry. I probably should have planned this a little better than what I did. Okay, so when I fed it this information, it gives me the image and it points out all the different things that it's analyzing.

[00:49:57]And that's what these numbers represent.

[00:49:59]There's different dotted numbers on the image that I gave it to say these are things to look for. The green ones are kind of like, I like this. Um, the logo has >> three. The guy with the beard is too old. Why are you hanging out with him?

[00:50:11]>> That it said nothing about that, Jeff.

[00:50:14]Don't worry. You're good. You're good.

[00:50:15]It liked the logo and hated everything else. Basically, it was like the faces are killing the logo. The two most important elements collide. You know, we're off. We're cut off on the side. Uh circuit board background is dated and cliche. The cyan rim light reads like a Photoshop cutout. The the kind of tracing around us. And so it gave me some real actionable information where I was like, "Oh, okay. This is like I could work with this, right?" And then it starts to get real weird because then it's like here, let me let me show you some examples of what you could do and it starts pulling back examples that are just Jesus. Uh they're just chopped.

[00:50:55]Like this one's I guess a cohesive idea, although it doesn't even integrate the logo that it said it liked. This I'm not sure what I'm looking at. It's like a cutth through see-through AI with us poking through, but it doesn't have any sort of sorting of direction. Let's just say the ideas it came back with probably because I gave it a flat image. Um, you know what I mean? Like what could it what could it do um immediately with a flat image and it just it came back with some hilarious results.

[00:51:24]>> What's interesting is that it recognized the discrete elements of what you gave it with the numbers >> but then >> but then it just went haywire.

[00:51:31]>> Right. Right. Totally. And and it went off. Yes, exactly. So, this was the first one. I took the suggestions from what it gave me and I kind of came back with something like this. I was like, "What do you think of this?" I actually thought this was an interesting approach and I I kind of liked it and it was like, "Okay, well, here's all the things I like. Here's the things I don't like.

[00:51:48]You know, your word mark got too small.

[00:51:50]That's the hero. That should be larger."

[00:51:52]And in this case, you know, the uh the images of you and I were just too big, apparently, according to Claude. And so, it kind of, you know, walked me through again all the things that I did horrible. And then again, it gave me all of these like really choppy blocky kind of examples of what I could do instead.

[00:52:09]But ultimately, I mean, I got to say, you know, I ended up getting there. I'm I was pretty happy with the way I turned turned it into kind of like a final uh art that is a little different than what you're seeing here. I just thought it was an interesting experiment, like what could Claude do uh actually do in the in the realm of design. And I think it stumbled in some ways, but it also had really good insight to share. And so I just focused more on the insight than I did on any of the layers and and uh designs that it gave me because those were absolutely useless. But I liked the the advice and I got to a place I liked.

[00:52:45]So there you go. That's Claude Design.

[00:52:49]You going to use it for your next design project, Jeff?

[00:52:52]>> I don't design anything you don't want me to design.

[00:52:54]>> You're like, that's not my cup of tea.

[00:52:56]look around. I mean, when I look at So, so, by the way, I have the first copies of my book, >> right?

[00:53:04]>> So, I have the priv Yeah. Uh, I have the privilege this I love this cover, right?

[00:53:09]Ben He is the uh is the designer. He's the actual designer >> to watch the iterations and the process of what he went going through and and getting the machine here and figuring out how to do all that and this and this sense of how to use the space and how to use this gap in the subtitle.

[00:53:23]>> Yeah. Yeah, for those of you who are watching, um, for those of you want to look like look like, go to jeffjarvis.com and you can see the cover. Um, I have tremendous respect for designers.

[00:53:32]>> When I started the magazine Entertainment Weekly, the design was, um, universally, almost universally hated >> because I, as the editor, didn't know how to manage designers and they went too far and we had to redesign within 15 weeks. So, no, keep me away from design.

[00:53:47]>> I'm not good at it.

[00:53:48]>> I always think I can do more with design than I'm able to do. like I have I have a I I can get halfway there again with everything in technology. I know just enough to be dangerous and uh it's definitely the case with with design.

[00:54:03]>> So, but I appreciated the kind of like collaborative iterative process with with the cloud design experience. Like I thought it was informative from that perspective. I just none of like literally zero of the examples that it created as an idea I should try were useful. So, as you all saw if you saw the video version. So, there we go.

[00:54:25]Anyways, that's my experience with the quad design.

[00:54:29]Um, we let's see here have next up the Amazon Alexa plus uh podcast feature.

[00:54:36]So, basically with your ALEXA device, which I just probably fired off one of your devices out there.

[00:54:44]Apologies. Uh, Alaxa podcast is the name of the feature rolling out to US customers today. No, no, no.

[00:54:52]>> Not a fan.

[00:54:53]>> I mean, I Notebook LLM's podcasts were a parlor trick.

[00:54:58]>> They were cute. I can see some utility if I were a student and I said, "Oh, review this for me. Just let me let me listen to it while I'm driving."

[00:55:05]>> Okay, I get that. Uh, but no, I think this is going to be slop of slop. We have video slop, we have visionary, you know, we have we have visual slop and now and now we have audio slop. No.

[00:55:17]>> Yeah. Kind of loweffort. Uh yeah. Yeah.

[00:55:22]I'll be curious to hear it uh to to be able to kind of cast my final verdict on it, but um I mean it basically you ask uh your device to create a podcast about a specific topic. It researches it. It um gives you a summary of what the episode's going to cover. you can get in there and tweak the length uh the tone what it focuses on and then it generates that podcast in air quotes uh with AI uh generated voices and there are partnerships here as well. Amazon made partnerships with Associated Press, Reuters, Washington Post, Time, Forbes, Business Insider, Politico, USA Today and more than 200 local newspapers.

[00:56:05]very interested to see if this is any good and whether people actually want this. I think there are some people that are that want this, but I don't know how many of those some people there are and how good the quality will be is the real determining factor. If it if it sounds if you if I feel like I'm listening to Siri read me uh podcast, um I'm not going to be interested. No, >> like I, you know, it has to be incredibly convincing.

[00:56:35]>> Yeah, >> absolutely torture.

[00:56:36]>> Yeah, it would be torturous.

[00:56:39]>> Uh, don't really care much about this story, but I put it in here anyways because it's done. Elon Musk lost his lawsuit against Sam Alman and OpenAI case was dismissed. Um, his team, by the way, framing it uh with in public statements as a win, saying that the lawsuit itself surfaced damaging information about OpenAI's nonprofit to profit conversion. So yeah, >> he's not 100% wrong.

[00:57:04]>> Yeah, that's true.

[00:57:05]>> Because uh I think there was some dam reputational damage to Altman particularly and Open AAI. Uh embarrassing, but they'll get past it.

[00:57:14]They're going to get past it as soon as they start cashing the IPO checks.

[00:57:18]>> Y >> um interesting to me was that the jury didn't even bother with poor jury sat there through all this stuff and they didn't even bother to get to any of the substance of it. They just said you're too late.

[00:57:29]>> Yeah. technical >> limitations. That's it by >> we we waste enough time on this.

[00:57:34]>> Um and so there was no >> long either.

[00:57:36]>> No, two hours. 90 minutes.

[00:57:37]>> Two hours.

[00:57:38]>> So I think it was mainly filling out the forms. So there was no um uh nothing was nothing was determined.

[00:57:46]No court of law said anything about the actual dispute. Too late. Screw it.

[00:57:51]>> Which by the way surprises me that the judge didn't say there's a statute of limitations here and you're over it.

[00:57:56]That's I I think that would have been something to be do preemptively. But in any case, >> something you catch early, someone catches early. I'm, you know, kind of surprised that Alman's team didn't catch that early. Like, >> well, I'm sure there was a preemptive I'm sure there was an effort to get them to throw it out on that basis, but for whatever reason, the judge didn't >> and uh so the jury ruled on that. It would strike me as an odd thing for the jury to rule on. But fine, that's what it is. The case is over. Uh they all should be embarrassed. Uh this is why people don't like AI. the the uh this is why commencement speakers are getting booed as we'll discuss in a second. Um I think this case kind of represents why AI has a bad rep.

[00:58:35]>> It's not the technology, it's these kind of people.

[00:58:39]>> Yeah. Yeah. All of it together. Yeah.

[00:58:41]Well, I mean we can talk about that, right? Uh Wall Street Journal has a piece. There's two basically two articles, but they kind of are are framed around this Eric Schmidt um experience at at University of Arizona.

[00:58:55]um where uh it was the com commencement um event going on and students just weren't having it. When Eric Schmidt uh was speaking about AI during his speech, they booed very loudly. Um he's been >> he has been an aggressive uh you would say advocate for AI. um you know both investment and also kind of competitiveness uh for uh the US in AI and everything and yeah the Washington uh sorry the Wall Street Journal writing that this is you know one of of many symbols of a growing backlash to AI uh people pushing back organizing around you know kind of their their this fatigue around AI and I mean I I don't I don't uh dispute like I've certainly talked to people that I can tell like that is a sentiment that is shared by a lot of people.

[00:59:50]>> It exists.

[00:59:51]>> Yep.

[00:59:53]>> Uh and definitely, you know, and and as we've talked to be a graduate right now and to be kind of in this moment of uncertainty around what AI and what this new technology means for all of the work that you've done to get you to where you are. It's got to be a weird got to be a weird place to be to be a student right now. So, Eric Schmidt got to be on the receiving end of that uncertainty there.

[01:00:18]All right, we are uh going to take a quick break. Uh quick before we do, we do have a YouTube channel. If you are not aware, go to youtube.com sorry uh go to YouTube, search for AI inside or AI inside show. You will find it. You can subscribe. You'll get all the video version. You'll be able to see that cloud design thing that I was talking about. Hopefully I didn't bore audio listeners to death uh with that, but no, >> I think it's worth I think it's worth looking at. So go there and take a look because if you if you're curious what it can do, boy, that's one way to look at it. Uh we're going to come back and and do a little speed round. So that's coming up in a moment. Stay tuned.

[01:00:59]All right, speed round time. Next Era Energy is acquiring Dominion Energy and a deal that um brings uh some pretty important players together um for powering AI data centers and everything else. CNBC uh is at least this article anyways is covering this as a major power infrastructure story that does relate to AI, right? Data centers consuming all this electricity and so this is really um kind of uniting two behemoths together. Um, this was a big deal. I mean, people seem to be um I don't know. What did you think of this?

[01:01:35]This was uh just Monday, right? Yeah, I think this was on Monday.

[01:01:38]>> Um, yeah. Uh, again, it's it's it's a lot of of acquisitions going on all around in this field.

[01:01:46]>> Mhm.

[01:01:47]>> Um, I I don't think we know yet whether the data centers are going to be needed to the extent that we think.

[01:01:55]>> Yeah, that's good. Um, and I and I'm and I'm fearing that we're going to have an overbuild as we have with the internet, >> you know, the crash and and and so uh the capex they said at Google yesterday was up I think either six or seven times over a few years. These are multiples of capex.

[01:02:13]This is why the stock market's not wildly happy because of the spending. Uh it's keeping the the economy afloat while everything else is getting messed up with tariffs and war and such. Um, so it's playing a role right now, but it's a huge, huge investment. And meanwhile, what we saw, what we hear all the time from Justin Wong, what we saw yesterday from Google, what we hear from Yan Lun, uh, everybody is the the main goal is to find more efficiency >> in the hardware and software so that you're constantly trying to do the same with a lot less or do more with less.

[01:02:48]And so, do we need all of these um data centers at the size which they're being built? Uh the presumption right now is yes. We can't get enough of it. We keep build build. Um u Ben Benedict Evans had his annual uh Mondo spreadsheet uh not spreadsheet slideshow come out and and one of the slides is so telling is the data center building has now surpassed office building.

[01:03:14]>> Oh. Oh, interesting. Isn't that so building for machines is bigger than building for butts.

[01:03:20]>> Uh and so yeah, the economy is responding to this demand, but I'm not sure that it's going to be a sustained demand. And and meanwhile, of course, 70% of Americans don't want them anywhere near them.

[01:03:33]States are trying to pass laws because because they have cooties. If you can't get rid of Eric Schmidt talking about AI, you can get rid of the data center.

[01:03:40]So, um uh I don't know. I don't know what to think about about this kind of consolidation in data center building.

[01:03:48]>> Yeah, that's interesting. Yeah, there there's a big movement happening. Uh Pope Leo I 14th will publish his first encyclical called Magnifica Humanitus on May 25th. Anthropic co-founder Christopher Olah will uh join the launch panel at the Vatican. The enc uh the encyclical is expected to address AI and humanity directly and yeah having an anthropic founder sitting on that panel is >> amazing one I hadn't really heard of much.

[01:04:22]>> Yeah true.

[01:04:23]>> Um yeah he's the pope is Leo is signing this on the same date as his namesake Leo I 12th signed his encyclical which was about the industrial revolution.

[01:04:34]>> That's really interesting. It really is really interesting and I think we'll see that um uh what he cares about is humanity in this.

[01:04:42]>> Uh I just read a um u in German with the help of Google Translate. I read a Q&A with Paulo Benanti who is seen as the Pope's AI guy. He wrote a book called Homo Faber.

[01:04:55]Uh and it was interesting because I think I was trying to get some signal of where the where Leo was going to go with this. And I was gratified to see Bonanti saying what what I often say which is that it's not the technology, it's the people.

[01:05:07]>> It's us we got to be concerned with.

[01:05:09]It's how we use this, what we do with it. That's what we get concerned with and maintaining human dignity uh in the next age of technological revolution and with the lessons we have from the past with the lessons we have from the industrial revolution where does this go? So, I'm I'm really really eager to read this when it comes out. And so, we'll we'll actually in a rare moment here on a tech podcast, we'll tell you about it.

[01:05:32]>> Yeah. Right. Yeah. I'm eager to to find out more about that, too. Lionus Torvalds uh says AI powered bug hunters have made the Linux security mailing list, in his words, almost entirely unmanageable.

[01:05:47]Uh and he says uh many researchers are just using the same AI tools you know to find the same bugs and then they're flooding the list with a bunch of duplicate reports that their team had then has to sort through. Uh Torvald's words uh he put it this way. He said enormous duplication due to different people finding the same things with the same tools and uh creating what he called unnecessary pain and pointless makebelieve work.

[01:06:15]I mean, this is the this is the other side of AI is so good for fi at finding bugs. A is it actually finding bugs? B does, you know, is they're all finding the same things and then submitting it.

[01:06:26]And so there's just an avalanche of these bug reports and then you got people on the other end um having to sort through that and find, you know, signal to noise ratio. What does that look like? and he's just basically saying if you want to actually add value, read the documentation, create a patch two and add some real value on top of what that AI did. So there you go.

[01:06:47]>> Yeah. And I I I imagine Torvol is far from alone in how people are getting bombarded probably. It's not just slop, but it's just people being able to do the same thing is what's interesting here.

[01:06:58]>> Yeah.

[01:06:58]>> It's not it's not for bad motive, >> but enough enough. Okay, you can all do this.

[01:07:04]>> Yeah. How do you ddupe at that level?

[01:07:07]>> Yeah. Yeah.

[01:07:10]>> Because if everyone's using the same tools, >> there was a post that I that I saved the other day. I didn't put in the rundown uh just of the of the increase in content. And we know this, right? But but it it quantified it a bit. And what that struck me most is the really almost um hockey stick increase in scientific papers.

[01:07:30]>> Now, it's not just because it can write.

[01:07:31]It's also because it can help with research. It also can do other things, right? But but there's a there is a huge increase in speech >> because of the speak talking machine.

[01:07:41]>> Uh and again it's not necessarily for bad ends in this case but it's just enough.

[01:07:47]>> Yeah. Yeah. Uh real quick before moving on I just want to call out uh Lee who's watching live show says I work for Next Era Energy.

[01:07:57]>> Oh >> it's a pretty huge deal for us >> from inside.

[01:08:01]>> I imagine so.

[01:08:04]Thank you, Lee, for watching and for commenting. Appreciate that. Uh let's see here. Meta bringing virtual writing to everyone uh with Meta Rayband display glasses. So, this is neural handwriting uh via apps uh that are coming to the Meta Rayband display glasses. I think uh God, I feel like we heard this maybe it was early news of this, but I feel like we heard this for something else. Maybe it was for Android XR or something. I can't >> cuz I don't remember this. This is This is new. You've You've played with this, right? Have you Have you played with the wrist thing?

[01:08:39]>> Have you had the chance to play with that? No.

[01:08:41]>> Oh, this is with the wrist.

[01:08:42]>> That's the thing. This is the wrist.

[01:08:44]>> Oh, I see. I I understand now. No. So, I haven't heard this news. Um, no, I actually haven't. Um, not not this particular one. When I was in God, was it Mobile World Congress? I think uh last year I played with a different company's neural band and uh yeah it was that was very interesting but I haven't I haven't done it in kind of connection with a pair of like smart glasses that I'm wearing >> right so but you know you got to remember to wear the thing I guess if it's a wristband yeah it's not a big deal like I do have some smart glasses that come with a ring that you can wear but I don't know the ring's kind of cheap and flimsy and like I wouldn't want to wear the ring. A wristband like that could potentially, you know, kind of blend in with things.

[01:09:33]It could be more of a fashion kind of accessory sort of thing that that isn't so cheap. So, I don't know. Maybe maybe it's useful. Uh interesting timing though because we had Google's audio glasses announcement at IO. So, there you go. Uh Elon Musk's X AAI unveiling its first coding agent called Grock Build. Yes, they gotta they got to get on it.

[01:09:56]>> Everybody's got to catch up.

[01:09:58]>> Everybody's got to catch up. Cloud code doing what it's doing. Developer focused tools having some pretty significant traction right now. And uh so you know, Grock's got to get in on that action.

[01:10:10]And then right before Showtime, I saw a uh exclusive Wall Street Journal um article saying OpenAI is preparing to file for an IPO very soon. The article says possibly by the end of this week.

[01:10:22]Don't really know any more details beyond that, but um could happen.

[01:10:26]>> And now that the case is out of the way, >> that's what enable that's what's enabled now. So they were just waiting for that case to be to be >> finished. Okay. So that actually makes makes a lot more sense than >> for them to be like, "Okay, cool. We can uh put that behind us. Let's uh let's get back to business. Make this happen.

[01:10:43]>> Um if you had a whole bunch of spare money, would you invest in Open AI?

[01:10:50]>> Oh boy.

[01:10:53]Oh, that's God. I I I don't want to answer that because like I do not listen to me for any financial advice. That's for sure.

[01:11:01]>> I don't know. Like if if maybe maybe I don't know. I don't invest in things like I don't invest in in companies or anything like that. So >> well judgment in that.

[01:11:12]>> If you have a mutual fund that ties to the NASDAQ or anything like that invest I Yeah, that's true. That's a good point. Um but I don't I don't like monitor the the market and decide like I'm going to invest in them because I hear that I don't know. Would you?

[01:11:27]>> Um >> would you?

[01:11:28]>> I I think it'll rise very fast on the first day.

[01:11:31]>> Yeah.

[01:11:32]>> Then I think gravity will take it.

[01:11:34]down. I think some reality with I think there'll be there'll be some some pentup demand for it. Uh but they've got to catch up and I think there's going to be a lot of tough analyst and journalist uh writing about open AI and the trouble that it's in right now.

[01:11:52]>> Uh if they could only have gone public a year ago, they would have been in better shape. But now they've got Anthropic ahead of them, Google ahead of them.

[01:12:01]Um, I I might I mean, theoretically, I'd buy it on the on a big dip, >> but not right after the open after the IPO.

[01:12:12]>> And uh don't take our our >> No, do not take our No, do don't don't do this at home, kids.

[01:12:17]>> Do not do not You can do whatever you want, but don't do it because we told you yes or no. That's for sure. At least I can speak for myself. That that is absolutely the case for me. Uh, cool.

[01:12:28]Oh, well, we have reached the end of this episode of AI Inside with new artwork and and all sorts of fun things to announce and everything. Uh, Jeff and Jeff, you have the book in hand now.

[01:12:37]It's tangible. It's real.

[01:12:40]>> That's got to feel good. That's a that's a cool milestone to get to. It does.

[01:12:44]>> You know, to open that box and be like, "Ah, there it is."

[01:12:46]>> Yeah. Well, actually, the funny story is the publisher found a typo on the back cover.

[01:12:52]>> Oh, no. So, I think they're going to have to reprint the cover and rejacket the books in the warehouse. Okay.

[01:12:58]>> Which will delay the arrival in the UK, not in the US. It's August 20 pub date.

[01:13:03]So it instead of right time it or right moment it said W GHD. And how the heck did that happen? And the publisher looked at what he approved and didn't have it. And I looked at what I'd approved didn't have it. You know how that happened? after investigation ah technology when they tr they used uh inesign and when they transferred to make the final PDF for reproduction inesign a lot of the commands the publisher told me use w so apparently when it was done the cursor was in the wrong spot there was an errit w and it went into the text >> nightmare kind of stuff that happens >> that sucks >> it's not a blurb >> is it better is it better that it happened that way and not a typo inside of the book.

[01:13:47]>> There's always typos inside his books.

[01:13:49]>> There's always going to be something.

[01:13:50]>> Yeah. Yeah. And you can you can Well, you could correct them on the next printing now. That's pretty easy to do because they're they're >> PDFs.

[01:13:56]>> Uh but the cover uh just as the new book comes out in a blurb about the guy that I uh yeah idolize >> book history. Yeah. I let him know about it just to know that.

[01:14:08]>> Yeah.

[01:14:09]>> You got to have >> So, it's great to have it, but these this is the problem with Adams, folks.

[01:14:14]>> Yeah.

[01:14:15]It's not as easy to change as bits.

[01:14:17]>> At least you caught it. Yeah. At least it was caught and and it's being >> The publisher saw it on on the train on the way home and I can't imagine how he gulped.

[01:14:25]>> Oh, I I can only imagine.

[01:14:28]Oh, boy. Well, jeffjarvis.com. You can see the cover that has no typos on it whatsoever. And you can pre-order also Gutenberg Prthesis magazine and the Web Weave. And uh new book series, Intelligence, AI, and Humanity coming uh someday soon. uh from Bloomsberry and Jeff Jarvis. Thank you, Jeff.

[01:14:47]>> Thank you both.

[01:14:48]>> Uh as for me, sure, pod tuneup.com for podcast consulting. I'm also doing AI consulting. Lots of consulting happening right now in >> because you should consult with Jason because he's really smart.

[01:14:58]>> Yeah. Well, thank you.

[01:14:59]>> And he knows his stuff.

[01:15:01]>> If you need to contact me, go to Pod Tuneup. Even if it's for the AI consulting, you you just go there for right now. I don't know how I'm going to there's a lot of uncertainty around like how I how I promote this stuff. So, anyways, just go there. You can find it.

[01:15:13]>> He's being too humble.

[01:15:15]>> Uh, ainsside.show with the brand new podcast art. It's funny that earlier I couldn't I couldn't figure out how to pull up the art. Oh, right. Go to the website.

[01:15:24]That's right. That would have been a really good easy way to do it. So, there we go. AISight.show for everything you need to know about this show.

[01:15:31]patreon.com uh inside show. If you want to support us on Patreon, you can do that. We have some amazing executive producers. Dr. Due Jeffrey Marachini, Radio Asheville 103.7, Dante St. James, Bon Rick, Jason Knifer, Jason Brady, Anthony DS, Mark Starcher, and Karsten Somatki. And maybe you, I don't know, you know, we're doing daily podcasts now, so maybe we can get you uh into that tier as well.

[01:15:57]>> Thank you.

[01:15:57]>> And to those who come live uh who are in the comments, Sab Daniel Croft, always reliably here. Uh I hope I don't pronounce this. Sabakanisha who left a nice comment before. Jesse Scott uh who else is here? WX ROR and Lee Woods talking in from Next Energy. Thank you for coming in and thank you for joining the chat too. We we really like that >> indeed. Yeah. Thank for the for the call out there.

[01:16:21]>> Uh so great to have you all here and uh good to see you Jeeoff. We will uh see you next time on another episode of the AI Inside Podcast. Have a wonderful week. Bye everybody.