Because of the Implication

Ajouté : 2026-05-23

[00:00:00]steps up with that day to finesse.

[00:00:02]Selling it right. No room for a guess.

[00:00:05]The mics are hot. The beats hit hard.

[00:00:07]Fans throw their hands high in the yard.

[00:00:10]Deals and impressions like a verbal sprint. Breaking it down with an ade.

[00:00:22]We've gone for main engine start. We have main engine start.

[00:00:33]Ladies and gentlemen, adtech ad talk.

[00:00:54]>> Where is it? Room. Oh my god. Where am I?

[00:01:00]>> You dropped during >> a second. I'll fix you. Just just a sec.

[00:01:03]>> Welcome everyone to Tech Ad Talk. Adam's here. He's he's in the the blank box. I guess he's in timeout right now. But thank you so much for joining. Ah, there he is.

[00:01:12]>> He's back.

[00:01:15]>> Sasha, you gota you got to throw those ones out there. That was a good one.

[00:01:18]Sad. Adam was he became his worst nightmare for a moment. A black box.

[00:01:23]>> Yep. But I'm back to my transparent self. not in the office this week. I'm in a hotel room in Tampa, Florida, where I'm attending a family wedding. But I would not miss this show. I'm Adam and I represent the buy side.

[00:01:36]>> I'm Gareth and I represent the sale side.

[00:01:39]>> Welcome.

[00:01:41]>> Thank you for joining us everybody. Um, exciting news in adtech world. Uh, I think that today we're just gonna do news, right, Adam? We'll just run news top to bottom.

[00:01:53]Uh oh. Is that a frozen atom?

[00:01:56]It sure is. Um, well, we're going to run news today. So, we'll definitely talk about Live Ramp because we have lots of feelings about Live Ramp. Um, and then there are a few other articles that we wanted to cover, right? So, there are some interesting things, interesting perspectives that came from agency holding companies. There he is. He's back. Um, and we're kicking off with news.

[00:02:16]>> [ __ ] this hotel Wi-Fi better better work. I paid like 12 bucks for it.

[00:02:20]>> Oh, man. I mean 12 is, you know, it's about the going rate. It's what it costs in airplanes. So, first article to talk about today, making marketing more modular. What Agentic AI can learn from the shipping container revolution.

[00:02:35]>> A, we recommend everyone reads this article because shipping containers are just cool. Like Adam, you've talked about shipping containers a lot. Like this is like one of your favorite things to talk about in supply chain is shipping containers. Um, but the notion and the the the brunt of the beginning of the article to tell people why shipping containers are cool.

[00:02:51]>> Um, is >> prior to the 1960s, the size and weight and everything about stuff that was shipped internationally was not standardized. So, people would just take stuff out to a boat and be like, "Hey, load this on the boat for me." Uh, and when that was the case, loading things onto the boat took a really long time and things would get damaged and things would get stolen and there were all of these issues that emerged. And a dude who used to be a trucker in the 1950s or 60s, I forget what the article said, was like, "Hey, what if we just put everything in boxes that were identical?" And that way the only thing that needed to be moved on the boats were these identical boxes where we knew exactly how big they would be, what dimensions they would have, etc., etc. And they're locked up. So they are locked up while they're in transit the entire time. And we'll kill a bunch of birds with one stone where we'll be able to maximize the space on the shipping containers. We'll be able to load and unload way faster because we only have to worry about getting one thing on and off. and will reduce theft and damage because these things are just shut. They're shut when we get them.

[00:04:07]They're shut when we get rid of them.

[00:04:09]We're not supposed to open them. Um, and it worked. Like it really worked.

[00:04:16]>> Like they uh I think the stat in the article was they went from like it took like 48 hours to load a boat down to like one or two. Some crazy like efficiency gain um in things moving.

[00:04:29]>> 70% efficiency gain. Yeah, >> I believe I believe it's 300% cheaper.

[00:04:34]Uh >> but uh >> you want to put it another way to calculate it. It's another way actually.

[00:04:44]But uh yeah, the definitive article on shipping container as the model for the AI marketplace is by an investor named Jerry Newman. He published it in Colossus and he spoke about it on the OddLots podcast. Um it's uh a great theory that Dan Hogan from Pavas is running with here saying what happens to a market when not only does it become much more efficient but it becomes predictably much more efficient in ways that everyone has access to >> and uh of it that I've been been touting as per Jerry Newman is that the adopters uh capture most of the value. It's not the shipping contain Thanks Josh. It's not the shift. It's not the uh the middleware as we've seen in ads companies in the middle uh or even like the agentic companies necessarily. It's the adopters. And there's another book that I would add called the box uh that's about how Walmart uh took shipping container and the efficiency gains to the extreme and switched the retailing from a salebased uh business to what what Walmart calls everyday low prices. So in other words, they changed their business to pass the efficiency gain onto their consumers by correcting for the error that causes overstock that causes discounts and say, "All right, we're going to have zero error. All the shipping will be as efficient as possible and everything will be on sale all the time." Um, so I think that's how that's one way that you could think about how this changes our market is like, right, everything could be on sale all the time if you create a service that turns this efficiency gain to the advertisers's advantage or the publishers's advantage.

[00:06:29]>> I'm ju I'm just unpacking that like explanation of what everyday low prices are because I didn't know that >> and it's like it's a mindblowing theory, right? The emergence of sales is actually to address an inefficiency in stocking. It's like not out of the like good nature of anyone's heart. It's like we have this [ __ ] on the shelves. We need to get it off the shelves and we made a mistake and we're going to take a hit. Like we're like we're going to make less money from these because we just need to get them off our shelves. And Walmart was like we fixed that and everything's on sale all the time. The other critical piece, guess what? Guess what technology? The effic shipping efficiency was one. And guess what?

[00:07:11]Technology, I'll give you a hint. It also changed casinos was critical for Walmart being able to execute on everyday low prices.

[00:07:21]>> Changed casinos, too.

[00:07:24]Was did they monitor inventory with cameras?

[00:07:28]>> Four. No. Close. Four letters.

[00:07:30]>> Technology with four letters. It tells you where things are.

[00:07:34]>> RFID. H.

[00:07:36]>> Yeah, that's it.

[00:07:38]>> So, they also they massively adopted RFID, so they always knew where the goods were.

[00:07:44]>> That's the other huge problem with shipping is like even if it's more efficient, you didn't actually have visibility into where >> you didn't have real time. So my my last company was called RTK >> and RS stood for release the Kraken >> but the real term was real time kinetics >> and it had to do with precisely this monitoring the location of objects in real time and yeah know where the know where it is when's it going to be here how long how much >> and you could you could extend the AI metaphor with that too is like uh chalice has gone through this or we can add enormous value for the advertiser and perform really well. But another critical piece is giving visibility into that and what it's doing where it's finding value and trying to figure out why is the yeah the analog to RFID is saying well what okay we found value but where is it?

[00:08:42]>> I mean it's really interesting. So I have a a few thoughts here. One is what I actually immediately thought of when I read this article was open RTB. Um and the adoption of open RTB bundles did a lot to standardize and commoditize inventory in adtech which is >> what are open RTB bundles Garrett?

[00:09:06]>> Ah so the the foundational element of programmatic advertising is the open RTB bundle and an open RTB bundle you can think about it as a little snippet of text with a bunch of different fields in it. So, one of those fields could be the site domain is CNN.com or let's do one of my clients. Um, let's let's do golf.com.

[00:09:30]>> Um, and the user ID is XYZ and the device is an iPhone. Um, and so what OpenRTB did is prior to the existence of OpenRTB which is a standard. So, OpenRTB is a standard developed by the IAB. Um, prior to its existence, everybody had their own little way of structuring all those little fields. So, everybody's text bundle was different. And that meant for companies to integrate with one another to buy and sell ads from one another, every integration was custom.

[00:10:07]They had to, and it was just like the shipping containers. They had to figure out like, okay, well, which fields do you have? What do you call this? I'm going to map this to this. How big is it? Is it bigger than ours? Is it smaller than ours? We got to shed these things because we don't use these. We don't do this. Um, and every integration took forever and was a mess because it was totally bespoke. And the creation of OpenRTB said, "No, no, no. We're all going to use the exact same taxonomy, right? The exact same method for describing all of these little fields in the bundles that describe the ad that you're going to buy. And when everyone's using the same thing, integrations get a lot easier and the sale of ads gets a lot easier and scales a lot better and all of this stuff. Um, now >> why is it called a bundle? Bundle of text that goes over into as the bid request.

[00:10:59]>> Yeah, that's that's that's what I think about it. It's it is a bundle of text.

[00:11:03]It is not a bundle of ad opportunities, which I had another weird shower thought about this weekend that we can discuss later. Um, but it is it is a bundle of text. That's that's I call it a bundle.

[00:11:14]It's it happens in JSON and JavaScript object notation. Um, and >> okay, >> so yes, this is the foundation of like internet advertising and the proliferation of ad exchanges and DSPs and all of these things is I think largely from this. It did remove a lot of integration friction from starting those companies. Um, and it was I mean it was it was great when it happened man. It was like awesome. Get everyone getting on Open RTB was wonderful. Now everyone uses different versions and that created its own problem where it's like well I'm on version 2.8 or what do you want on I'm on version 2.5 doesn't work the side representation why why I want to make sure everyone understands why we needed that it's because they were trying to scale machine learning >> yeah% >> people talk about agentic standards this is not our first rodeo all of us in adtech have gone through the same process that they're now talking about agentic standards of making the machines speak the same language so that you can train a predict ictive machine on doing valuations. Like this is this is not our first time through this process and there's a bunch of pieces of it that are predictable if you you know paid attention to the to the first time and how it went and uh the unintended consequences and the addedon fixes like that's that's what the real drama is about >> when we turn to today's big news. So Gareth, what happened on Sunday? I had text.

[00:12:42]>> Oh my gosh, it was on a Sunday. I thought that was remarkable. I got a text from my team. Um, so Live Ramp got pot by none other than Publix, the agency holding company based in Oh, yeah. Paris.

[00:12:56]>> Facebook. French company.

[00:12:58]>> French company. Arthur Sadun. Um, >> yeah. Were there any posts about this?

[00:13:05]>> I I didn't see any. Did you?

[00:13:08]The post about this is this is a landmark in not just in adtech or in the show but in my life. I have never been so overwhelmed by something happening in the media as as the posts about this because there's so much going on, right?

[00:13:23]It's so much it's finance, M&A, adtec are topic, but this is the first time that we had tons of AI assisted and AI written materials on an event as it's happening. And I I mean I cannot I I could go three hours talking about it, but I'll just say at the start that uh if you're bad, if you do not have a good take on something, AI is not going to help you. That was my big takeaway. This bad takes I saw a lot of bad takes made worse by AI and was it's a little disturbing how bad.

[00:13:58]>> Well, and it's funny because a lot of people are just negative for the sake of being negative. Um, what did get me was there's a buddy of mine who's not in the industry >> and he he fed my quote to his chat GPT and it formulated a lovely response explaining why my perspective existed and I was like this >> I know you don't I know you don't know what you're talking about like you you clearly have not poisoned your AI with your bad thoughts you poisoned it with my good ones. Um >> as as many people said commenting about this I should say as they many people incorrectly said in comment it's all about the data and it's like in a way >> what data what data what do you what do you I mean yeah >> what do let's put on >> it is technically correct that it is all about the data if we want to be technical about it is technically correct >> uh go back to Terry yeah go back to Terry I liked Terry's take >> Terry this is the definitive uh take because Terry's a the public master of adtech M&A and he's wasn't the banker on the steel so free to talk about it and he oh Adam was about to start reading it so I will read it AI agents and workflow automation are dramatically improving efficiency across campaign planning media buying creative production optimization reporting and analytics and I really want to dig in on this we'll come back to it uh historically advertising has depended on fragmented systems and labor intensive coordination among agencies, platforms, publishers, and adtech vendors. Agentic AI reduces this friction by collapsing workflows into softwaredriven orchestration, thereby creating a more efficient advertising supply chain. It also creates margin pressure for the agencies and platforms that were built on managing the workflow complexity. Moving forward, workflow automation is unlikely to provide durable differentiation because agenta capabilities will quickly become table stakes. And I really agree with the first sentence and really agree with the last sentence. and the se and the middle part I'm kind of like um but the the reason I really agree with the first one is because I deeply believe that the agentic part of adtech that a lot of people are talking about that's not system construction right not artfal does not what gamera does but the agentic part where we're taking things that humans did by hand and we're letting machines do them that is squarely an agency function. Um like agency people did lots and lots and lots of things by hand. Um they just they just did that is their life. They are the agent of the advertiser. It is therefore their job to do things on their behalf. And the hallmark of a good agency is they do smart things and they do them well and they do lots of them.

[00:16:52]and a bad agency does dumb things and doesn't do very many of them. Like that's that's like the the differentiation there. So there's like really a lovely application of agents where if you're a smart agency person, you can be like, "Oh, which of these functions that I'm doing can I have agents do for me and which ones are value ad where I can't have an agent do it and my creativity is required and my like knowledge is required and uh yeah, it's my it's my agentic spiel. I'll pause for a second, Adam."

[00:17:24]>> I mean, all that's TBD, but uh yes, thank you. I mean, you're welcome.

[00:17:29]>> The Yeah. What was I What was I saying when I got cut off? I don't know how much you >> You're about to start reading the post.

[00:17:35]>> I just All right. A lot of the commentary under the post was I thought misinterp interpretation of what Terry just said, which I thought was very funny. A lot of people were like a lot of people are on this thread and others are hypothesizing that Pubis is turning into an ad network, which is ridiculous and laughable. Pubis already bought several ad networks and have them operating inside Pubis if that's what you want. People don't understand how big Pubis is and how many clients they serve. So any theory that's based on oh now all Pubis customers are going to get this is is ignorant and doesn't know how a holding company operates like they they can the whole idea of the holding company is to contain multitudes and offer different service under different banners and different flavors of of offering right. So, so it's it's they don't turn it into a monolithic offering just because of this purchase. Um, and you don't have to go that far to explain this, right? Live ramp is growing faster than Publois. So, that's enough to explain it. Pubis has a lot of capital and if you could turn capital into a growth rate higher than your current growth rate, that's considered a good deployment of capital and investors will reward you. And you don't need anything more to explain that. And it's not even that much more.

[00:18:49]It's like Live Ram's growing at like 9% and Pubicus is growing like 6% last year. So it's like oh yeah they're going to get you know essentially if you volume adjust it it becomes like half a point of growth.

[00:19:03]>> Yeah. Do you know their uh >> is is Pubis' valuation better than Live Ramps? Because Live Ramp trades at a very low multiple on their their revenue. Um >> uh yeah it's much higher. Yeah, that right a couple of billion and focus is >> in the tens of billions above 10.

[00:19:20]>> Yeah, it's a great deal. It's amazing.

[00:19:23]And and then there's the >> amazing uh so the one my my take on this deal has always been live ramp's magic is its ability to reduce friction. Like that's what it is. Everyone uses live ramp.

[00:19:43]Every time I do an integration with someone, every time I start working with someone, if there is a discussion of user ids, the discussion is, oh well, do you work with ramp ID? Do you tie this thing into ramp ID? They are everywhere.

[00:20:00]And what that means is now that they're part of publicist is twofold. One is it's that publicist clients now have a lovely integration with everywhere.

[00:20:09]Their portability remains. It's still there. if they have clients who don't use live ramp today, they can take the live ramp solution to them where if those clients come in and say, "I have this firstparty data. How do I and where do I activate it?" They just say, "Great. Slap it into Live Ramp and Live Ramp will help you activate this first-party data wherever you want to.

[00:20:29]We can go out and find it and we can do other magic with the marketplace, whatever it is." But the other thing they get is if they decide to pull the rug out on non-publicist clients where they're like and which they said they're not going to do, but like the tacet to uh to quote it's always sunny in Philadelphia, it's the implication.

[00:20:50]Um and and the idea is right like hey if you really like using live ramp for all of your firstparty data portability where it's mapped to everything and your segments can be everywhere and you can buy your users everywhere with at scale.

[00:21:05]Um you should probably talk about how the rest of publicist's assets can help you keep that business running smoothly and happily. Um, and they basically have like complete optics into anyone who is not a publicist client what their data practice on live ramp looks like. And that's just so valuable. Like, oh my gosh, right? Because if somebody, let's say like a huge brand, and this is actually a question, a good question for you, Adam. Let's say a huge brand decides they're going to switch their agency. They're going to switch their holding company. There's got to be a bunch of like shoehorning and all sorts of stuff to get the data into the new one, right? Like there's got to be a whole thing. Um and I would assume that's like part of the part of the switching cost and that's part of the stickiness of working with an agency is like well you've done all this work to upload all this stuff into us. Um you don't want to do that again, do you?

[00:22:03]like just fix the relationship with us.

[00:22:06]>> Yeah. I mean there's not a great case for moving that. some some some agencies I mean some advertisers will but the fact of it if I could put put a bow on it is enterprise advertisers who use live ramp regarded as theirs like live ramp early on built in built an environment where you can log in and see your CRM data which means any any email that you collected for any reason right some of these are customers some of them are former customers some of them maybe just signed up to be your sweep stakes or whatever right any emails you collect Ed are in this live ramp environment and you can do things with them, right?

[00:22:45]Onboarding is what they're most known for, which would mean that you can, you know, target that audience in Tik Tok or Reddit or any anywhere you want, like a a thousand places. Uh, and and in Open RTB, the publishers have also have live on that side. So, essentially, every publisher in the world will let you match to these lists. But the point being is that advertisers regard this as theirs. Live ramp has achieved a kind of plumbing integration. You should think of it as a plumbing integration where they've enacted what lives in the advertiser's house to not just any one market but all the markets >> all of them. It is the tool that you use to get your data from your house to activation.

[00:23:28]>> Right? So that said, I think what I think a lot most people don't realize that advertisers regard this as theirs and that's part of the value to publicists and I wouldn't extrapolate further than that as saying that that's what makes it valuable, right? You don't want to jump from that to their creating an ad network or an agentic framework.

[00:23:45]>> No, you don't need to.

[00:23:47]>> It's enough that advertisers consider this theirs and their environment. Um, but then I was gonna >> No, I mean I my tweet about this was I'm on the sell side, right? Like I'm not like I've never done a first-party data onboarding for a major brand. But I get this like I get that this is a really sticky hard business to be in. And one I thought that Live Ramp traded at a really low multiple considering its strategic value. Like Live Ramp was trading at like less than 2x revenue or less than 3x revenue. Um, and I was like, "This doesn't make any sense. This company occupies a like really strategic position that's going to be really hard to displace." Um, so I thought that was strange.

[00:24:32]>> Yeah, it was poorly understood. I think it's too complicated.

[00:24:34]>> You said strategic value and >> and let's talk about the defensibility, right? Because the >> people are conflated two values. One is the graph. Live ramp has an ID for every email. That's all you really need to know is live has an ID. It's actually more complicated than that because they've also created the connections between like the household IP address and the phones that are in that household every night. That's that's what they really mean by a graph that has like cross device graph on it. But you can think of it as every email and the um online ids that attach to those.

[00:25:08]So they have that graph. But the value of live ramp is the plumbing >> the plumbing.

[00:25:12]>> Right.

[00:25:13]>> Right. So both those things lots of companies have one lots of companies have the have matching.

[00:25:20]>> Yeah.

[00:25:21]>> And there are lots of companies that have different kinds of scaled plumbing but live is the only one that has both in an open way.

[00:25:30]>> And I I also think it's the the quality of their graph. There are a lot of crummy email to IP address probabilistic graphs out there. There are a ton of those.

[00:25:40]>> How do you know how good it is? How do you know which ones are good?

[00:25:44]>> I mean, you don't >> the cell side, >> right? Then why you say it's good?

[00:25:49]>> I Well, no, no. I know that they're crummy because I know that the data is >> hold from sketchy places.

[00:25:55]>> You said live has good quality. Did you mean it's perceived as having good quality?

[00:25:59]>> Oh, no. You could see from side.

[00:26:02]>> No. So, I I mean live ramp's deterministic one. So they're onboarding one where when they work with an advertiser, the advertiser says, "Hey, this I have this email address and live ramp has code on page on the checkout page or whatever it is." And they stitch those things together with households and then go and find them. The original Provenence data set, they have a probabilistic one too that I don't trust because I don't trust any probabilistic email to IP address data sets. Um I don't know if Live Ramps is any better or worse than anyone else's. I mean, I would assume it's better if their seed data set's better, but I don't really know. Um, but the that first point of integration, right, the point of the advertisers working with Live Ramp is like a an undeniable >> it truly on board their CRM. And on the sales side, when you enter your email address to have something sent to you or to get a notification that you actually want, there's how good is the chance that Live Ramp is collecting that email address and matching it to your to your cookies or IP? What do >> you know of e-commerce and >> on the sell side? So, when people like log in on a website, like sign their newsletter or something.

[00:27:10]>> Yeah, I actually So, it depends on the publisher, depends on how good of a job they're doing. Um, most of them don't hook it up to Live Ramp if I were guessing. Most of them are using an email service like Beehive or something like that. And I'm sure Beehive views Live Ramp not as a competitor, but certainly as a company that they uh like, >> but as a separate data asset, >> I'm a newspaper. I want to set be able to send out the live ramp ID in my bid request or have live ramp transcode that bid request. So, >> so fancy publishers who know what they're doing will work with live ramp to send in hashed user ids. That is a thing. Now, if they use live ramp for the probabilistic matching has its own thing. I know live ramp and live intent as well. So, live intent is another email company, right? They got bought by Zeta um for a lot less money but does something similar, right? Like live intent is an email management system and they built a probabilistic email to IP address matching tool and they would charge publishers.

[00:28:11]But the reason I'm so skeptical of all these is because they made money from the more IDs they could send that buyers wanted to buy, the more money they made.

[00:28:19]The incentives were bad.

[00:28:21]>> Uh they remain bad >> being on a little better.

[00:28:24]>> Exactly.

[00:28:25]>> Being on both sides. Okay. So that's a reason to believe that it's quality. But I want to I actually want to testify as a live ramp customer for the quality of the plumbing tech. Like we re like we could tell any customer, oh if you have a seed you want chalice to mo model off of go to live ramp connect and chalice is a destination and there and in artf our our only predecessor of doing containerized work uh cross company is is uh is live ramp sidecar which is transcoding in the hot class. One millisecond of transcoding a bid request ID into a live ramp ID is enormously useful for matching impressions and conversions for a company like Chalice which is not operating a scaled platform. So, and it's it's a matter of speed, right? It's they have to do you just have to do that really quick so that by the time we get the vid request it has an ID in it that we can read.

[00:29:16]>> It's decrypted. Yeah, >> so that's that's plumbing. Other companies can build that but they just haven't. and live ramp spent years and millions and millions of dollars building it. And yeah, I agree with you that the value of of all of that work was not really reflected in their but I think for perfectly good reasons that it wasn't it wasn't clear from a live ramp earnings call what they had built or why it was worth so much or what they charged for it.

[00:29:40]>> Yeah. Even though they made $800 million a year, like it wasn't a small company by any stretch of the imagination. Like >> it was >> but it's funny if I could I love that company uh we I go way back with Scott how all the way to Avenue the Live Ramp CEO was very active at Avenue the company where the Chalice founders uh met 20 years ago.

[00:30:04]>> Cool.

[00:30:05]>> Yeah, a lot of history and we love the company and we love our live ramp team.

[00:30:08]They're critically important partners to us and will continue to be so. But if you want to see something funny is is tell Claude to like look at Live Ramp's earnings all their earnings calls and tell >> Oh no the punch line >> during the punchline tough.

[00:30:31]Yeah. Um I'm sure it was going to be good. What else does Scott How sell?

[00:30:37]Scott how Scott how this is his second like data company exit. The guy's a machine. Um >> I didn't know that actually.

[00:30:48]>> Oh yeah. So his last one was another huge one. It was Axiom. Yeah. Scott how sold Axiom to publicist, right? No. Dominic.

[00:31:00]Yeah. Um he's the man.

[00:31:05]>> What's the last thing you heard? I'm not running the call off my phone.

[00:31:08]>> No, it's Last thing I heard was the punchline. So, go into Claude and Oh, it's running off the phone.

[00:31:17]>> Uh, go into Claude and ask Claude based on the based on live ramps at last X earnings calls. What comes next?

[00:31:30]>> Basically, how does this company make money? You say what products do they have? What do they charge for the products? and how much margin do they each make? You could ask that for a huge company like Microsoft and it will all come right up. Microsoft discusses and all, but with Live Ramp, it's it's just like, well, we sell subscriptions and we make $800 million from subscriptions. And as a customer, those subscriptions, like you have you have an MSA that covers some things and then you have a bunch of other contracts for the other things you've asked to do. Like it's it's very uh innovative. I'm gonna use the word innovative because it's it's market demand that has driven this.

[00:32:08]Like people want the end customer wants a lot of these costs to be invisible.

[00:32:12]They want the plumbing to be >> sure >> not think about it, right? Like there's a c an overall cost on it, but um there's never been, you know, they they've successfully avoided the kind of margin pressure that other IT based plumbing companies go on go are subject to.

[00:32:29]>> Yeah. And I just think back to what you originally said, which is if you have a new client and they want to work with you, just work with us in the live ramp like connections marketplace and make us a destination. And it's like, man, if you're already in there where you have you have now have a lovely network effect, right? Where the advertisers are already using this to onboard their CRM data that's already done. And if it's not, when they go work with Chalice, Chalice goes, "Well, it would be real easy if you had your CRM data in Live Frame." Like, that would be that would make life very easy. Um, yeah. And my tweet was, "I can't believe it took someone this long to buy it if they if you could buy it for this much." It is it is the ubiquitous connective tissue of IDs in adtech.

[00:33:15]>> Um, and connective tissue between the people that matter, the advertisers and the agencies. Like I don't care about ad exchange ID syncing and probabilistic IDs and all this. I don't I don't care about any of that. I care about advertiser has firstparty data that they care about. Where does it live? Where where do they store it in to get it out elsewhere? That's live rabbit. That's what they use. Um >> yeah. Yeah. It's it's it's you can't think of it as storage because they might store it in uh in Snowflake, right? It's it's it's the applications.

[00:33:49]It's the combination storage and applications that we keep calling plumbing. They mean, >> you know, it runs right up to your house if you're if you're the brand. Like it's you you see the end point >> and that's critical. Now, let's talk let's talk about Pulis. Another company that I love I've worked with in the past as a customer of Chalice today. Tons of respect for Pubis. Like they've just demolished their competition over the last five years. like you know they won an unbelievable I would say unprecedented series of pitches and uh they've they have more share of the market than any holding company well until until uh Omnicom and IPG merged they were dwarfing building a company and they sort of forced that merger just to have a competitor as big as them so they are you know they're the champions but this is champions of what they do and uh they didn't give a lot of tea leaves into uh what this is for right the commentary was a nod toward AI and Agentic Media. Uh, let's read the comment. Go ahead, Gareth.

[00:34:49]>> Uh, Live Ramp is the only vendor that integrated for IDs with Google ADH, too.

[00:34:54]I don't know what ADHD is.

[00:34:56]>> Adds Data Hub. That's just saying they had engineering resources to spare.

[00:35:01]There's no revenue in ADHD. I mean, it's easy to believe there might have been some. All right. So, so what is Publix doing? And I think it's hard to comment on this if you don't understand holding companies, which you don't unless you've worked at one at the executive level. Uh, and it's hard if you don't understand the enterprise advertising market, which is very different than the small business advertis the longtail advertiser market that produced most of of the Walt Garden revenue. Uh, so another thing to understand is epsilon, right? Pubis is paying in this last round of M&A when Pubis bought epsilon and IPG bought Axiom. It started when Densu brought bought Merkel. So this was about 10 years ago. They decided they all needed household data sets except WPP and Omnicom at the time. Uh but three of them decided that it was worth billions to have these data sets. Pulysus was successful partly because their data set came with the old valueclick ad network stack uh bundled with epsilon and they made some offerings that are that are uh bundled offerings not like the text bundles you were talking before but like uh tech and service bundled together uh that could work well with a principalbased media buying strategy.

[00:36:20]Pulis is vast and parts of their business operate with the strategy today although no one knows what share of revenue come from this or what share of su of pulis success it's due so why buy live ramp when they have all this going and there's I think a lot of crazy theories about it well where I would start is with this enterprise small business split right there's no small business live ramp right there's no small business with substantial CRM files to the extent that you need live ramp or want this plumbing hooked up to your house. This is about the enterprise. And one thing that I think it's worth thinking about is enterprises that are that don't love the principal conversant model that exists within Pubosis uh but want some scaled AI solution. I would say that live ramp aims at something sort of in between that starts to feel like uh what a lot of people in the enterprise are saying is advertisers can do with this a lot of themselves and don't need to pass over decisioning to an ad network. Uh having live ramp in house to me moves in the direction of an offering that's both scaled and enterprise.

[00:37:31]>> I um he caught a lot of [ __ ] for his agent stuff Arthur. Um, and I actually thought that those takes were really wrong. I was like, who's telling you people what a gentic is? Because >> if you have something right that now handles um I'm gonna let you take this question.

[00:37:50]I'm just going to process.

[00:37:53]>> Um, so but I uh I think that the the agentic thing makes sense if you've just bought wedding, right? If you've just bought plumbing between houses and the job of agents is to go out and pick, right? Let's say that in an agentic world, we have something that is picking. Okay, I want to use this solution for this and this with this and I want to test this and this and I want to do each of these things. Having the plumbing there is what enables the agents to do that. Without the plumbing, the agents cannot do it. If the data is not portable between places, you cannot easily have an agent allocate between them. And and that was that was my read on this. He was like, this is for an agentic future. It's not because, you know, it's doing an ADC CP direct sales ad thing. That's not it's not what it's doing. What it's doing is it's creating a lower friction environment where when agents are making selections, they can just work.

[00:38:50]>> Um, and I was like, oh, that actually makes perfect sense to me. like that that that totally tracks um >> yeah the lever ID it's a sort of lingua franca right if you want if you want your systems speaking to each other they have to speak the same language and like protocols are some of the same language but this is the language of user identification which is bound to be a critical element of the energetic advertising future and now we have that common language for user for addressibility lives within probosis so yeah to say that It's has nothing to do with Aentic.

[00:39:24]Seems crazy to me.

[00:39:26]>> Yeah, people really went hard on that. I was like, I actually think the dots connect pretty well. Um, >> but you know, everyone wants Agentic to mean automated direct sales. That's that's what everyone seems to want it to mean. And that's just >> it's just not it's not >> Did you Did any did Martin Sell comment?

[00:39:44]I missed Martin Sell this week because when stuff like this would happen. Do you remember Martin? Are you old enough to remember?

[00:39:50]>> I remember him. I remember Martin. I've never met Martin.

[00:39:54]>> I don't know. I mean, his commentary, he would comment in public and you'd get usually a video clip. He didn't like post on LinkedIn or anything. You'd get a clip or he would speak to the press.

[00:40:03]It's an old school guy. He's still active. But he would he would come in.

[00:40:07]He was the only one, and now there's no one in the holding company leadership world who would come in and talk as an operator of one of these companies and speak like from an economic financial view. This is where the value is. this is how we make money, these are the margin implications, these are the growth implications. He's the only one who would lay it all out publicly about how to think about it. And you know, I just always paid attention to what he said because it was the only way to get that lens on what's happening. Um, and you know, now yeah, I think I think he's missed in in not coming out and providing it though. I'm sure he had I'm sure he has the takes and somebody else >> I think he post I think he posted something about the deal. I think it was a snippet. I'm I actually just I was just trying to find it on on X. Um >> yeah, >> I didn't see it.

[00:40:52]>> In his his uh postWPP life, he's become >> I'd say a little seems a little more emotionally motivated, right? He's mad at WPP for for how they laid him off or whatever. But um so sometimes that comes very little I can the uh being salty out of a desire for revenge but his analysis like mine is is very incisive long-term thinking and crystal clear.

[00:41:21]>> Yeah. I mean I uh well a you don't get like kned without like you know >> kned. Yeah.

[00:41:28]>> Kind of know what we're talking about.

[00:41:29]Um but I uh he built he built it right like he he he is the one who built WPP into what it was like that was it's his baby right is there was he was he the founder um >> he took it over I think when he took it over it was still it was still called wire >> like the W you know what WP stands for >> not like some wire like some wire company that the innovators of WPP like used as the holding company for their agencies. That's crazy.

[00:42:01]>> It stands for wire something and something. Yeah. So, it's it's really funny. It was just like a literal placeholder for something to hold all the agencies. But yeah, before Publoist mopped the floor with everyone and dominated WVP under Surell did that and it was through largely through foresight over where the business was going. And you know part of that strategy toward the end of his career was the Zaxis strategy which was that the rec the recognition that open web programmatic in general and ABN Nexus in particular was creating the ability to have an outcomesbased scaled machine learning offering that would deliver outcomes for clients at scale without getting nickel and dimed so that the value accretion would be could be captured by the service company without the advertisers complaining.

[00:42:54]That absolutely worked uh at the time and still talk about today as something that might you know when people say this is what publicist is doing like one reason you know it's not true is because sell and and his team did that with with Zaxis and it ran its course eventually majority of advertisers rejected the model that Zaxis was operating under.

[00:43:17]So, it's it's no longer the dominant >> agency trading desk operation.

[00:43:21]>> Ah, all right. I have a new one for you.

[00:43:24]Um, they got mad when agencies did it, but I will point to someone who is an old buddy of mine.

[00:43:32]>> Um, who has had a lot of success recreating Zaxis in other places.

[00:43:37]>> Yeah.

[00:43:38]>> Uh, and that's Mark Grether.

[00:43:40]>> So, Dr. Mark Grether.

[00:43:43]>> Dr. Mark Grether. So, Dr. Mark Rether was one of the founders.

[00:43:48]>> Yeah. Yeah. He's an old buddy.

[00:43:50]>> Um I tried to sell RTK to him when he was running Seismic. Um and through no fault of RTK's um it was not a time for Seismic to make it stressful over there. Um but I I got to know him a little bit. He's a super smart guy, super nice guy. And um but I think that the reason he's been so successful first at Uber, right? Like the the commentary is that Uber became a profitable company because of their ads business >> and like that was that was Mark.

[00:44:24]>> Um and now he's at PayPal, right?

[00:44:28]>> Um >> PayPal. Yeah.

[00:44:29]>> Doing it again, running the playbook again. I think that building an outcomesbased ad network, right, and what's essentially large retail media networks works when you have a remarkable data asset. And when there's a remarkable data asset in there that you can couple with unique inventory, you can build this consolidated advertising product. And it's a playbook now. Um, and I think that he got the initial inklings, and I'm spitballing this because I just had this thought, but the initial inklings of that were at Zaxis where it's like, okay, I have our data assets over here, right? because WPP has stuff. Um, but we have these advertiser relationships and I'm going to build this like internal network and Uber was able to deliver on it like with flying colors because they had both inventory and data and PayPal is probably going to be the same. Uh, right like PayPal has pretty spectacular user data. They have an unbelievable identity graph. they have all of these things that work to their favor and to build an internal outcomesbased retail media network/ ad network is viable uh when you're at that scale and I think he's going to do it again I wouldn't be surprised if PayPal is wildly successful executing it um it's just >> my message is >> my my message is >> yeah that model still exists uh AI doesn't the existence of AI doesn't significantly change it >> um when you're say data asset, you really mean conversion data. When you say data asset, you mean conversion data. I I would contest a little bit on whether conversion data is what makes >> What do you mean conversion data?

[00:46:05]>> Like the retail the retailers could do it is because they have sales, >> right?

[00:46:10]>> You meant by remarkable data.

[00:46:12]>> Yeah, I think um well Uber and PayPal are like different than retailers to an extent. Um >> yeah, they have interesting data.

[00:46:24]>> There's just interesting data, right?

[00:46:26]There there are relationships in there.

[00:46:28]There there there is purchase history certainly and I guess retailers do have purchase history, right? So a retailer knows this person has bought these 25 things. Um and therefore I and and I know their email address and I know their cookies because they come to my website or they log into my app. Um and they you can stitch those things together. That's what I mean when I say data asset. I mean so yeah, so retailers I think do have it. Um, maybe PayPal and Uber are just slightly different than your average retailer, but I I think that's right.

[00:46:56]>> David, go back. David, give us the graphic of that guy. His name was Abi.

[00:47:02]>> Oh, yeah. Yeah. No, that one.

[00:47:04]>> Not yet. We'll have to next week.

[00:47:06]>> This guy's name is Abi Yadav. I've never heard of him. Forgot what country he's in. Uh, but he commented under Terry's post something that sounds pretty similar to my thesis. is like one of the first times that someone I do not know at all. I don't know his company or what they're up to though it might be competitive with Chalice. Uh but he his his summary at the end was live ramp may accelerate the opposite move brands deciding they're done renting decisioning. I am a 100% agreed with with this prediction which is what what you were saying what you just said was access and you said that still goes on and what my question what I love to raise is are these enterprise offerings and as enterprise players >> good question >> publicist live rand and retail media networks like Walmart as they move into this type of offering don't they sort of push the enterprise advertiser out like we were just saying the enterprise advertiser views that live ramp conversion data as theirs and their environment. And Chalice's idea that we found to be very sticky is that the decision algo should be yours and should live in your environment. It should have full visibility into what it learns about what's valuable for you out in the market. Um, doesn't this exacerbate that split? And you know, maybe you could see it as as a a hedge on that in various ways that we might see from various platforms. Uh but I would caution to anyone who's saying, "Yeah, the biggest ad network wins." It's like that that game has sort of played out like Google has grown as much probably as much as it's going to grow. Uh it's its growth has slowed. Uh and Meta is growing like crazy on what seems to be mostly longtail.

[00:48:58]uh and and the growth numbers that we're talking about here in publicist live ramp are in the high single digits. Um so it it definitely leaves a lot on the table for the biggest brands, the biggest publishers and the biggest advertisers to chart their own growth trajectory outside of all of this.

[00:49:16]That's what this guy Abby is is referring to. And I think he's right.

[00:49:20]So I think that a a really important part of this conversation is the spectrum of sophistication when it comes to the measurement of incremental sales.

[00:49:31]The the enterprise requires this because for them to understand the ROI on their marketing efforts is a herculean task, right? They they are spending >> probably not >> okay but to to me it's >> conventional Yeah. Um, and you know what I I'm sure you know like there are some people who really understand that you know PNG spending a few billion dollars a year.

[00:49:58]Here's how we do the econometrics to understand what the ROI was. But for an ad network, right, the the reason the ad networks fail there is because ad networks need relatively straightforward attribution in order to like extract the value from data. I think that this is the this is true of meta. That's why it appeals to the long tail and I think that your thought is very much correct is that ad tech ad networks work up to a certain point. Uh and I remember when we first started chatting you asked me a really a question that I didn't understand why you were asking me >> and I was like this question doesn't make any sense Adam. Uh, and the question was, how do you optimize a campaign where the system doing the buying isn't aware of conversions or like what's working and not what's not working? How do you do that? Um, and I was like, what do you what do you mean it needs to be aware of a conversion in order to optimize? That's that's how it works. But I think for really big buyers, right, it's it's not that simple. Um, it's just it's more complicated. They have to do more math in the back end to understand okay this channel delivering to these users with these attributes we believe drove this much incremental like lift in our sales and that and we have a whole system that exists in the back that's not a oh show ad add ad make thing happen system that's just not how it's built um >> and I think that that is the limitation that they run into here which is to build things to service those advertisers it it's way more bespoke, right? It is you have your own algorithm that lives in your thing. You have your own data that lives in your thing and you manage that centrally with tentacles going out to all the different places that your ads are running. Um, and >> I would say I not a little Yeah. What I my discovery after six years at this is it's not a tweak on the existing architecture. It's not a tweak on Zaxis.

[00:52:00]It is a different architecture at the foundations, right? Like it's um >> it makes sense that it would be >> and and the idea that and the idea that you that you tacked on here that brands can't tell or don't appreciate the difference. I'll just I'll just bring up finance. I started doing optimization for in credit card optimization, right?

[00:52:20]Credit and then we went to home loans and you know before the financial crisis, it's not not a coincidence. uh these companies were giving out as much credit as they could with with network like with to mostly to the ad networks.

[00:52:33]>> Yeah.

[00:52:34]>> Uh and the crisis shows that obviously that's not the goal. You don't want to give out as much credit as you can. You want the credit you're giving out to be calibrated to the person paying it back, right? That that's that's finance. It's really clear. But this is true in every business is the the actual value is not the value of the conversion, right? A quote does not really have a value attached to it, a credit card application. And it's the same thing with the sale of a can of soda or the sale of a car. It depends. The actual value to an enterprise depends on a whole bunch of things. The easiest example to understand is getting a credit card. It depends on if the person what the person's going to spend on the credit card and whether they're going to pay it back. And you cannot tack that data on later. That's been the solution.

[00:53:21]Like, oh yeah, you know, we're in that network and we could deal with that.

[00:53:24]just send us a feed later on of which of these selections were adverse, who didn't pay back their credit cards, and we'll adjust. No, you have to have that built into the value prediction at bid time. It that makes all the difference. You cannot rearchitect these things uh retroactively. Um so that's what this quote means to me. It's like the reason you don't rent decisioning is because your valuation, your true valuation is essentially private. It's what finance people call alpha information other people don't have. Uh and if you can't bring your alpha to market without it being averaged up with some platform valuation, then you then your your money is not being well spent. uh a lot of brands are realizing this and it's arriving at different in different ways at different companies. Uh but to say that this is an argument about measuring incrementality is is like a bunkered view like inside the marketing organization there people there people are ringing their hands about oh should we use the MMA should we do an incrementality test should we hire this company or that company in the CEO CFO IT conversation they're like spending it's much different they're like the line Adam is a billion dollars right the media line dollars And within three years, we're going to bring accountability to that through our combined experience, tech and experience in in finance and technology, and the marketing is going to follow along. But there's a certain physics of this where we're going to get to the actual rules through which our valuation becomes more accurate and it pays off. Um there's certain inevitability to the to that that's hard to see inside of a marketing organization that's saying should we do MM or recreationality test or both at once like like yeah that that could fill up your world if you work at a holding company you deal only with the marketing department it seems like the whole world but there is another world bearing pressure down on that uh through stock valuations that have already priced in some massive gain from from AI uh that's that's you know it's not going to resisted from inside the marketing organ.

[00:55:42]>> But what happens when a lovely Google rep >> gets the phone number and maybe it's not even a Google rep. How about this? The CEO's brother-in-law is a who's it at Google Marketing.

[00:55:57]>> Yeah.

[00:55:57]>> And goes, you know, we have the best data scientists on earth.

[00:56:03]>> They invented data science >> and we have more data than you do by far. You are but a single advertiser. We work with a billion advert.

[00:56:12]>> Oh yeah. One banker told me >> the conversation also included at a bank marketer. And by the way, we have $250 billion on deposited in deposit at your bank.

[00:56:26]>> Oh my god.

[00:56:30]>> A significant.

[00:56:32]>> You want to argue with that?

[00:56:34]>> Oh my god. Well, it reminds you of the the whole joke, right?

[00:56:37]>> Move it. We're going to move it. We're going to take that money and move it.

[00:56:40]And your 3% on 250 billion is going to go away.

[00:56:44]>> Exactly. Like if you owe the bank a million dollars, you have a problem. If you owe the bank $100 million, the bank has a problem. Like that's it's that's what it is. And um >> yeah, so those calls are going to come.

[00:56:57]>> Yeah.

[00:56:57]>> Uh Google is not going to rearchitect into custom valuation.

[00:57:01]>> Um they already they already have decided their growth is going to be in cloud. There'll be a product suite in cloud. And I think they have years to play a sort of I don't know what the game is where you give people a little bit and back up a little more and give them a little bit but you know like if there is a CF CEO CFO saying okay Google we'll use your solution and we'll do a custom valuation in cloud and apply it through Gemini to the stack you know and you have one year to show results you know that easily becomes two years three years or you know you have we have some results but not others and other things have to be brought on for and this you know like a lot of it will be lost in the sauce >> but I do think what it'll come down to right is the R word results so like >> once people can agree upon results that's that's when it comes to that's why I brought up MM and MTA and all that stuff earlier is because somebody's going to make a judgment at some point hey this billion dollar line item was worth it >> or wasn't worth it >> right and like someone's going to do that >> um and that's that's that's what people will compete I I'll go I'll close with one other thought is like these problems are problems of of of um lack of shared context >> right the marketer what the marketer is looking at and what the CFO is looking at and what the head of IT looking at don't connect in any meaningful way uh and when and if you talk about agentic and AI coming to advertising what I say to watch out for is look for someone to use these tools to bring this whole picture into focus for all aspects of the agency, for all aspects of the enterprise advertiser where they could start to clarify this is where your money is going. This is how your dollar is broken down and this is the return.

[00:58:48]And it's not a question of M&M or MTA or or Lyft tests. It's a matter of all the data that's available being effectively summarized and transported in shared context into the into the enterprise. I think people are overestimating their ability to hide the ball in the era that's coming.

[00:59:11]>> So did you >> I did that thing.

[00:59:16]>> All right.

[00:59:16]>> I like that. That's a good closing thought. Overestimating their ability to hide the ball.

[00:59:21]>> That's an important one.

[00:59:22]>> All right. The enterprises. All right.

[00:59:24]So, in closing, we have a we have a video today at atte talk went out on the street in Brooklyn. G, this is a surprise for you. I did the camera work here. I hired a park slope personality named Atari.

[00:59:38]Perfect name for Atat on the street and we decided to go get opinions >> on AI from regular people. So, we could cover this in detail next week. Send your questions and comments here for the first time on adtech ad talk. Uh, eight minutes of atat on on the street.

[01:00:00]I >> think so.

[01:00:01]>> Amazing. Fantastic.

[01:00:05]>> Okay. Sh, how's the festival going for you?

[01:00:08]>> Um, it I always love the Fifth Avenue Festival and the food's really good and like there's different like culture spots that I really like.

[01:00:16]>> Oh, yeah. cuz like you get to experience different cultures. And the bouncy houses are also really fun, but they get a little hot when like the sun goes on the rubber for a little too much time.

[01:00:27]>> Do we have permission to put this on YouTube?

[01:00:29]>> Yeah.

[01:00:29]>> Amazing. Okay. I'm going to ask you a question about AI. What are your thoughts on AI?

[01:00:35]>> Um, I actually think it's a great tool.

[01:00:37]I think, um, I don't know too much about it and I'm still learning and I wish I knew more. Um, there's so much learned.

[01:00:44]I actually just used it this morning.

[01:00:46]Um, and I used it last night. I kind of gave it a list of all the ingredients in my pantry and I was like, "What can I make for dinner?" And I ended up making something. But, um, I think if you use it correctly and use it as a tool, it would be It's great. Yeah. What are you using? ChatBT Claude only because, um, that's the only thing I really know and it's the quickest. Like, I just downloaded the app and I and I use it that way. Um, if there's something else that you guys suggest, I would be open to hearing more. But yeah, >> I'm using Claude.

[01:01:20]>> Okay. And who is Claude?

[01:01:23]>> Who is Claude?

[01:01:24]>> Do I do I have your guys permission to put this on YouTube?

[01:01:26]>> Yes, you do.

[01:01:27]>> Amazing. Thank you.

[01:01:29]>> Um, yeah, I think the same thing cuz like most of the teachers there's like do not use AI in your work, but they like assign things with AI. It's like really weird and it's like not like the best, you know, for the ecosystem at all.

[01:01:45]>> What's your favorite food to eat at the festival?

[01:01:48]>> Well, last year me and my dad got these like kebabs. It had like beef on it and it had like the perfect amount of like seasoning and sauce and they were like so good.

[01:02:01]>> What culture does kebab come from?

[01:02:05]>> Um, I actually don't know.

[01:02:07]see in my teachers take photos of the work to grade it like using like chat GBT. So, I think it's a little crazy and I don't like how far it's getting and all the videos I see on like my preview page. Um, I don't like like how it's affecting our environment either. I don't think it's amazing to have, but it can help in some cases. I just don't think we should let it get as far as it's h like it has been.

[01:02:29]>> What does AI safety mean to you? Well, it should mean a lot, but I don't exactly know what it means. I keep on hearing this thing um going around like we shouldn't support chat GBT and AI is the devil and all this stuff, but I don't really understand why >> and maybe there's there needs to be more education on what is happening. Like how could to me I don't really understand like how could a computer be the devil or like how what are the differences between Claude and Tat you know like if they do the same thing um I'm just not understanding the differences. Yeah, >> I think it's like bad for the environment. It's also like kind of in infrastructuring on like daily life is like not really good because like the data centers being built and stuff. It's like just like bad for like the environment and for like society. And I think societyy's coming so reliance on AI that they just stop functioning themselves. Like people use it to like write emails and they don't know how to write it like themselves. And that's like kind of an issue because like as humans like you need to be creative and be able to function as a human instead of having a robot function as a human.

[01:03:33]>> Oh. Um, I like tangulu. It's like um like fruit covered with a candy layer and um I really like getting it. It's really good.

[01:03:46]>> It's from Asian culture.

[01:03:49]>> You have a very sophisticated palette.

[01:03:51]I'm deeply impressed.

[01:03:53]>> Thank you. Um I personally think that um there are certain jobs that are going to be um in AI's grasp that humans have less ability to like be able to work with um without AI. So I think a lot of jobs are going to be solely based in like AI and data and just computer sciences rather than jobs that humans typically would do. So I think it's kind of dangerous in that sense.

[01:04:22]Roxy, what are your thoughts on AI?

[01:04:27]>> Anything you're scared about in relation to AI?

[01:04:33]>> What does the future hold with AI?

[01:04:39]>> Is there anything that you wouldn't look up on chatbt for security reasons?

[01:04:45]Um there's plenty of things that just like yeah I think anything that goes into a computer personally I wouldn't like give all my financial thing like but that's just me like I also recently just saw someone on TikTok say um if you want to help boost your credit score drop in like your tax return like all this stuff and it will help you boost your credit score. What or maybe it wasn't tax returns it was probably whatever else it was. Um, and I was like, "Oh, that makes sense." But I hope everyone's blocking out, you know, their routing numbers and stuff like that. Um, just in case, I guess for a data, what is it called? Like a like a hack when people like hack it. Yeah. So, >> that'd be the reasons why I wouldn't share those things. So, >> totally. That's so reasonable.

[01:05:35]>> I think the future looks endangered endangered with AI. Yeah.

[01:05:39]>> Okay. Sh, we're going to switch gears.

[01:05:42]What are your thoughts on AI?

[01:05:45]>> Well, I think AI is um really smart and it could help people, but some people like use it to make like like like weird videos and like like swap and like I don't like that, but I like how it could like help people learn new stuff.

[01:06:02]>> Yeah, I think it really is in danger cuz like I've seen videos where it's like an AI robot, but they're not even doing anything. like they're taking over the jobs of like people just like sitting on the couch all day.

[01:06:14]>> The children that have access to it and like you're saying the the terrible things that people do think about and like can easily surge up. I mean that gives me goosebumps. Like I don't think that way. So like I that's not something that comes to mind, but I guess you're right. There are sick people in this world and it's actually kind of crazy to think that they have something at their fingertips um that they can use in such a negative way. I don't think the like world is going to be very good in a few years. Like we're not going to have a lot of clean water and I just like we're AI is going to take over our jobs and I just like think it's just not going to be very good.

[01:06:53]>> I'm scared with our I'm scared for our future with AI honestly.

[01:06:59]>> Can we talk more about the slop videos?

[01:07:01]What do you mean? Can you describe >> like they make like brain rotted things and like it it's like sometimes they're a little funny but it it's still like good for your brain and >> totally totally totally. What's your favorite chat box?

[01:07:17]>> I don't use AI.

[01:07:19]>> Um I sometimes use Goth to do my homework in all honesty. What is it called? Um, so it's an AI that like if you take a photo of your homework, but it it will explain how to do it and it like teaches you how to do it basically.

[01:07:32]>> Cool. Very cool.

[01:07:35]>> I don't really support AI major like majorly.

[01:07:38]>> What's your favorite chat box?

[01:07:41]>> I don't have one. I don't really like use >> Yeah, I don't like to use AI. It wastes too much water.

[01:07:47]>> So clawed all the way. Is that what we're we're going for?

[01:07:51]>> Claw. Um, well, there's this one channel that I like to watch and it it is AI, but it like it talks about like different things in history and like what you would experience in that time and that's what I like.

[01:08:08]>> What do you guys think about AI? Um, I'm not a big fan of like the AI being used in like schools and stuff because a lot of my teachers have been using AI on making assignments, but they don't they wouldn't let us use AI for answering the assignments. I like the use of AI for like helping like medical stuff in medicine, but not with like the general use.

[01:08:32]>> Oh my god, this is great.

[01:08:37]>> You know, >> what do you think? The AI doomerism really gets to me.

[01:08:42]>> Yeah, those team >> really gets to me. Those team like they're all they've all been convinced that the world is going to be destroyed by AI. Also, the water things like the water things are ridiculous. Like the water concerns are so silly. There there must have been really good graphs. Like you know how much water we waste on almonds? Like almonds consume a profound amount of water. Like a absolutely profound amount of water. so that everyone has almonds. If we just if everyone stopped eating almonds, if we if we if we banned golf courses in almonds, we would like the >> it's a nick the AI.

[01:09:19]>> So much water.

[01:09:20]>> It's so much water. So, it's just not an argument, right? It's a it's a it's a weird propaganda thing. Um makes me sad.

[01:09:28]Like this makes this makes people sad and anxious and sad and anxious people do do bad things. Like it >> Yeah, it's bad. I had an idea. I had an idea for Chalice when Anthrop when what are they called? I say anthrop anthropic didn't want to said that the new Claude model was too powerful for them to release. I thought like Chalice should do this and be like we just made a we just made a model for for Hershey's Reese's. That's it's too powerful. If you release this, everyone would just be eating Reese's Peanut Butter Cups all the time.

[01:10:00]>> The Roas was too good. The Roas was too good person. They're gonna spend every dime they have on marketing.

[01:10:09]They're gonna make $10 back.

[01:10:11]>> I had a I had a good talk with that young woman and I I recommend anyone bring it to uh to Claude if you can to just ask. People should know why did the anthropic founders split off from open AI and what's the difference between constitutional AI that anthropic founded and uh the training mode used at open AI? Then you can answer her question.

[01:10:32]What's the difference? Maybe we'll cover that in a f future episode. But uh let let us know your thoughts on >> on ad out on the street. Oh, Roxy the dog. Thank you, Ryan. Roxy the dog was, you know, at least at least she wasn't wrong, right?

[01:10:49]>> More more right than little.

[01:10:53]>> Dog the dog knows. All right, thanks a lot for coming everyone.

[01:10:57]>> Amazing.

[01:10:58]>> We'll see you next week.

[01:11:00]>> Thanks everyone. Take care. Hallelujah.

[01:11:03]Hallelujah. Hallelujah.

[01:11:11]>> Darn. That's the end.

[01:11:12]>> Yo, the crowd is back. The lights are dim. Front rows buzzing. It's about to begin. and tech. And talk