Ads in AI: Karsten Weide on Advertising, Agentic Commerce, and the Future of AI Monetization.
- Jun 11
- 64 min read
What happens when the technology that powers advertising becomes the thing consumers interact with directly?
In this episode of Signal & Noise, Rio Longacre and Brett House sit down with industry analyst Karsten Weide to explore one of the most important—and controversial—questions facing the digital economy: how AI will ultimately make money. From advertising inside chatbots to agentic commerce, autonomous media buying, and the future of the open web, Karsten shares why he believes AI will reshape advertising more profoundly than any technology shift he has witnessed in more than three decades covering media, technology, and digital advertising.
The discussion also tackles one of the industry's most debated topics: advertising inside AI products. Will OpenAI, Google, Anthropic, and Perplexity eventually embrace advertising as a major revenue stream? What will AI-native ad formats actually look like? And can advertising coexist with the trust users expect from conversational interfaces?
Along the way, the group discusses:
Why AI adoption in advertising is still in its earliest stages
The rise of autonomous media buying and agentic workflows
How Amazon, Google, Meta, and emerging ad tech players are approaching AI-powered advertising
The future of SMB advertising and self-service media buying
The battle between open standards and proprietary AI ecosystems
Why the programmatic supply chain may look dramatically different in five years
The opportunities and risks of AI-driven creative production
Whether AGI is actually closer than most people think
How AI could reshape agency operating models and industry employment
If you've been wondering how advertising, commerce, media buying, and consumer behavior evolve in a world increasingly shaped by AI agents, this episode provides one of the most thoughtful and grounded discussions available today.
The future of advertising may not be humans buying from brands. It may be agents buying from agents. And the implications are enormous.
Watch the full episode and join the conversation.
🔑 What We Cover💡 Key Takeaways🎯 Why This Episode Matters
Read the full transcript below.
Brett House (00:01.06) Hey everybody, welcome back to Signal and Noise. I'm Brett House joined by my co-host, Rio Longacre. And today we have Karsten Weide, did I get it right? On our podcast, who is a principal and chief analyst at W Media Research. A lot of you might know him from, you were formerly at IDC. And you advise clients on advertising media and technology trends, and you write some very interesting reports. The most recent is the top 10 trends in advertising. Karsten Weide (00:01.486) You got it right. Which we read yep Rio (00:06.968) Hey there. Rio (00:12.056) Nailed it. Rio (00:26.924) which we read Yep. Brett House (00:28.718) in 2026, which I thought was super interesting. over the course of your career, seems you've had analyst roles, you've been at the Yankee Group, IDC, which I mentioned, you're widely recognized for your work covering digital advertising, connected TV, media economics, increasingly agentic and generative AI, and you're a linguistics major. And I know we talked about that ahead of time. From Leibniz Universität Hanover. Karsten Weide (00:31.398) So over the course of your career, seems you've had panelist roles. You've been to the Yankee Group, IDC, which I mentioned. You're widely recognized for your work over in digital advertising, connected TV, media economics, increasingly agentic and generative AI. And you're a linguistics major. I know you've been talking about that ahead of time. From late in his university life and over, you joined the crowd. Brett House (00:54.726) Join the crowd, join the family of sort of liberal arts majors. Karsten Weide (00:56.476) Yeah, we've had like comparative literature, political science, we named across the ages. Everybody in the industry is a master of arts, so there you go. Rio (00:57.26) Yeah, we've had like a comparative literature, Journalism, political science, you name it, across humanities. Brett House (01:05.507) Yeah, yeah. Yeah, yeah, and I actually think that one of the most interesting, kind of boring, but most interesting courses I took in university was Greek and Latin Elements in English. And so it broke down the language into its component parts, and it just changed how I thought about the language. And then I was studying other languages as part of my major. I'm trying to get my kids to do that, but kids nowadays just don't read very much, so. They don't read anything at all. Rio (01:09.432) Love it. Karsten Weide (01:19.036) super interesting. Rio (01:32.078) not gonna read anything. Brett House (01:35.046) So, yeah, your report, the top 10 trends in advertising in 2026, it was super interesting. Appreciate that contribution to the industry. And you're focusing on a lot again, as I mentioned, on how generative AI is reshaping advertising and commerce. So we think you're a perfect guest. We just spoke to Scott Brinker in one of our podcasts on the topic of MarTech and... Karsten Weide (01:58.3) in one of our podcasts on the topic of Mark tech and just the deep research we really appreciate and I think the audience appreciate it. So welcome to the pod. Thank you very much for having me. Yeah. And did I miss anything in your background? Brett House (02:02.626) And just the deep research we really appreciate and I think the audience appreciates it. So welcome to the pod. Yeah. And did I miss anything in your background? Let the audience know kind of who you are, where you're coming from. And for those not familiar. Karsten Weide (02:16.988) You know, I'm a, so I've been in the industry for a long time. I've done media and entertainment for like more than 30 years, I think now. So it's been a long time. I've been a journalist, I've been an analyst, I've been a product manager at Yahoo. So it runs the gamut in my career. And now I'm just a advertising technology influencer. I stole that from Ari Paparo and a consultant, right? So I serve. Brett House (02:42.949) Yeah. Karsten Weide (02:45.936) I serve mostly vendors, a few publishers thrown in for good measure. Once in a while, there's an agency that I work with, but it's really mostly vendors, advertising technology vendors. And so I help them with- Maybe that's true. Yeah, I know. Brett House (02:58.384) I think once you hit the 30 year mark in your career, you can no longer say that you're an influencer. Rio (03:02.83) Incap it, yeah. Influencer. Well, maybe an influencer among the older folks, right? Like the seasoned people like us, right? Karsten Weide (03:09.564) No, no, I'm among the sages of the industry. That's what I am. Brett House (03:09.708) Yes. Yes, yes, that's that's what I realized. We've been at it for 25 plus years and we're like, we're to stop cutting it 25, maybe 30, some people I know say that. But yet we have some wisdom to share. We've been around, we've been through multiple iterations of digital advertising of the entire data ecosystem. So it's certainly fun to share our knowledge. Rio (03:14.84) I said that. I said it. Karsten Weide (03:21.98) yeah Karsten Weide (03:32.444) Yeah. Multiple reinventions, multiple crises, and all that stuff. Yeah. most of the stuff that I do is just to complete that. it's content marketing, go-to-market support, or custom research. Just like forecasts, TAMs, what have you. So that's me. And just if you research, just like the IDC, you guys do market analysis and vendor analysis to understand how different vendors fit into the space. Yeah. Yeah, yeah. Rio (03:32.942) Yeah, multiple reinventions, multiple crises and all that stuff. yeah, you get used to it after you see patterns. Brett House (03:51.278) Yeah. And in just to be research, just like IDC, do you guys do kind of, market analysis and vendor analysis to understand how different vendors fit within the space, whatever category they play in. Karsten Weide (04:02.18) Yeah, so I'm doing now basically exactly the same that I did at IDC. It's advertising technology and advising clients. Well, Karsten, we're thrilled to have you. I know we've talked about this for a little while. We run into each other very frequently, whether it's at, I think last time we chatted a little bit, it was possible. Possible, yeah. Possible. Brett House (04:14.31) It's awesome. Rio (04:14.542) Cool. Well, Karsten, we're thrilled to have you. I know we've talked about this for a little while. We run into each other very frequently, whether it's at, I think last time we chatted a little bit, was at ramp up. I, no, possible. That's right. We saw each other, saw each other possible. And then I know we saw each other at ramp up. So this is a recurring theme that we run into each other at conferences and events where we're doing similar things. We're meeting with people, we're making content, all that good stuff. So thrilled to have you. And thank. Karsten Weide (04:30.342) we saw each other. So this is a recurring theme that we run into each other at conferences and events where we're doing similar things, we're meeting with people, we're making content, all that good stuff. So, thrilled to have you. I'm glad you came now to talk about what is a very interesting topic that is getting a lot of press, and that's advertising in AI. There's a lot of debate about it. Obviously, it is happening. We're already seeing within the, certainly the frontier models and the big, let's say, not the... Rio (04:41.75) I'm glad you came now to talk about what is a very interesting topic that is getting a lot of press and that's advertising in AI. It's just a lot of debate about it. Obviously it is happening. already seeing within the certainly the frontier models and the big, let's say not the business, but the consumer facing AI tools that have chat interfaces. You're now seeing advertising for the free models. So for the free version. So it's already happening, but the question is how big is it going to be? Karsten Weide (04:59.184) business but the consumer facing AI tools that have chat interfaces you're now seeing advertising for the free for the free model so the free version so it's already happening but the question is how big is it gonna be where's it gonna go is it gonna be good is it gonna bad the entire proposition here you see companies open AI Gemini or including including advertising as I mentioned so I think personally it's gonna become a big revenue stream I think that is without a doubt that may be Rio (05:10.722) Where's it going to go? Is it going to be good? Is it going to bet? Will it corrupt the entire kind of proposition here? And you see companies, OpenAI, Gemini are already including advertising, as I mentioned. So I think personally, it's going to become a big revenue stream. think that is without a doubt that may be, I mean, that's my opinion. That may be up for debate, but I also think it will be incremental rather than existential for these companies because you look at the revenue they're making, whether it's charging 20 or 200 bucks a month. Karsten Weide (05:28.668) I mean, that's my opinion, that may be up for debate, but I also think it will be incremental rather than existential for these companies because you look at the revenue they're making, whether it's charging 20 or 200 bucks a month, these are big businesses already, they're scaling super quickly, and philanthropic something on the business side, but certainly on the consumer side, OpenAI is generating a lot of revenue. Google's already an advertising business, right? So I think it'll be a big part of it, an important part, but I don't think it's the whole thing. So this is we're gonna talk about, what did you get your view? Rio (05:40.27) big businesses already. They're scaling super quickly. Anthropics done it on the business side, but certainly in the consumer side, OpenAI is generating a lot of revenue. Google's already in advertising business, right? So I think it'll be a big part of it, an important part, but I don't think it's the whole thing. this is what we're going to talk about, wanted to get your view, as well as talk about some of the interesting reports you put out recently that we're going to highlight too. So welcome. Karsten Weide (05:57.82) as well as talk about some of the interesting reports you put out recently that we're going to highlight too. So welcome. Yeah. Well, thanks for having me again. Yeah. So how big is it going to be? I think it's going to be big. And so I think that five years from now, I think most of the value chain will be automated by agentic AI. And the reason why that is is there's two reasons. So one, money talks. and bullshit walks. So there's a huge competitive pressure on just about anyone in the industry. So therefore, there's a strong incentive to help anything that would make you more efficient. So that's one. But then also, the complexity of the programmatic ecosystem has reached a point where it's just not possible for humans to manage it anymore. You had a briefing with a vendor. few weeks ago and they had actually done the math of multiplying out all the different factors that go into a programmatic campaign, which platform, which channel, which advertising format, which pricing model, yada, yada, yada. And once you do that, they claimed you get to like more than a trillion permutations, right? I don't know if it's a trillion, let's say it's just millions, but even then... Rio (07:06.734) Thanks a lot. Brett House (07:20.485) Yeah. Karsten Weide (07:24.476) It's just impossible for a human being to manage that. And therefore, going into automation is just do it or die. Do or die, right? So it's going to be big. the other question is, where are we today? And today, it's really the earliest days in the industry in terms of adopting agentic AI. Rio (07:34.222) It's the only way to do it. Brett House (07:52.775) Yeah, it's an exciting place to be. I feel like we're on the cusp of a tsunami that's, you you better be prepared with a surfboard that's capable of riding a tsunami. Karsten Weide (07:59.674) So yeah, so Rio (08:01.294) Well, it's early, we already have pretty, it's early, but you already have pretty good adoption considering how early it is only a couple of years in. Right. I mean, I think if he looked, you your rewind looked at, let's say even using web browsers, it would have taken years or even search or social, any of these things took years of them to get even close to the market penetration that you see with AI right now. Or, I mean, I mean, I, could pull the stats, but open AI has hundreds of millions of users. Right. Karsten Weide (08:26.844) So the adoption of AI is, of course, enormous, right? Both on the consumer side and on the enterprise side. Obviously, there's more money in the enterprise side than the consumer side. But I'm really just talking AI in advertising specifically, right? Now, AI in advertising in general has been around for a long time, at least starting back in 2012 or something like that. So it's been around for 10, 15 years. Brett House (08:42.565) Yep. Karsten Weide (08:57.052) mostly machine learning based optimization of whatever you have. Gentic AI, that's new. That's really software that can go out and do stuff like a human could, but either see my autonomous sets the situation today, or in the long run later on, completely autonomously. And there, it's the early days. Brett House (09:00.25) Yeah. Yeah. Karsten Weide (09:27.822) Mostly, agentic AI at this point in advertising is internal point solutions. So of course, have the World Garden session. There's Google. There's Facebook. There's Amazon, all of which have agentic AI solutions internally. In indie space, it's a different story. Pubmatic is there. They have a great PR campaign. mean, kudos to them. They've done a great job. being on the forefront of agentic AI. Hey, we have the industry first. We have the industry first, fully integrated agentic AI campaign. I suspect there's still a bit of manual labor involved. But hey, good for them. I think you've also done a good job. Viant is there. was in front of you. Agents workflows. Viant is there. Yahoo is there, if you want to call them as an indie. There's a bunch of startups. Rio (10:13.336) Yeah, them and Viant, think, have also done a good job in terms of embedding, embedding, like, agentic workflows and more AI tools. Rio (10:27.527) Amazon DSP too. mean, they've been rolling out. They're kind of, I wouldn't think it's end to end, they have some like video and content generation tools that are AI enabled that are pretty good. Karsten Weide (10:27.836) Amazon DSP too, they've been rolling out. I wouldn't think it's end to end, they have some video and content generation tools that are AI enabled. Yeah, Amazon, the stuff that they do is just amazing, right? mean, let's say you are at SMB on Amazon, you're selling stuff and you wanna have a video ad promoting your products. All you need to do is to input the Amazon product ID and a one sentence prompt and the damn thing will spit out a whole commercial for you. And it works. So they're doing that for video. I remember even the precursor to that before they launched the video capability. They initially launched the background image one, which sounds kind of like, what's so great about that? But you realize these companies that have really complex. Brett House (11:08.73) Yeah. Rio (11:09.592) But it's funny, I remember even the precursor to that before they launched a video capability, like they initially launched just a background image one, which sounds kind of like, okay, what's so great about that? But you realize these companies that have like really complex, like product catalogs, they have hundreds or thousands of products. Getting those products with the right background image, queued up and ready within a retail media experience. Karsten Weide (11:24.188) product catalogs, have hundreds of thousands of products, getting those products with the right background image, queued up and ready within a retail media experience, you wouldn't think it's unimportant, but actually makes a huge difference in terms of conversion rates. If you look at actual tying it to sales and tying it to transactions. so having a background, a lifestyle background image actually converted, I think it was like 15 to 20 % better overall. And converting white backgrounds to like, Rio (11:34.68) You wouldn't think it's important, but actually makes a huge difference in terms of conversion rates. If you look at actual tying it to sales and tying it to transactions. actually, so having a background, a lifestyle background image actually converted, I think it was like 15 to 20 % better overall. And converting white backgrounds to like lifestyle images was very challenging. So I remember they were explaining to me in this meeting, I thought it was wild how a retailer could just upload their images to white backgrounds they'd use for normally conversator rip through these in just a couple of minutes. So we'd have all of them with these. Karsten Weide (11:51.004) lifestyle images were very challenging. So I remember they were explaining to me this meeting, I thought it was wild how a retailer could just upload their images to white backgrounds they'd use for normal conversation, rip through these in just a couple of minutes. So we'd have all of them with these lifestyle images, some little blurred out, but it actually made the images much more appealing and boosted conversions for even little things like that for an SMB or even an enterprise who instead of hiring photographers and taking weeks or months. I remember one brand estimated it saved almost $3 million. Rio (12:04.898) lifestyle images, some little blurred out, but it actually made the images much more appealing and boosted conversions. So even little things like that for an SMB or even an enterprise who instead of hiring photographers and taking weeks or months. I remember one brand estimated it saved almost three million dollars in total costs. Brett House (12:21.574) Yeah, yeah. Well, that was one of your key themes in your report, right? Was the generative AI sort of turning creative from scarce to abundant. I think you had a fact that 90 % of short video ads are going to be built by generative AI in 2026 this year. Yeah. And just the versioning, to your point, Rio, the fact that the versioning piece of this, especially if you're a retailer with a massive inventory management system and e-commerce catalog, just the sheer volume. Karsten Weide (12:31.735) Yeah, I yeah, because why wouldn't you? mean, Brett House (12:51.526) from a personalization perspective, it became kind of untenable for humans to manage. And so if you're just swapping out simple things that align more to the person, either it's product alignment or it's sort of context alignment within the ad itself, it's gonna have, you're doing this at scale. So even if it's small incremental improvements, it's gonna have a larger improvement when you start to really scale this out across millions of views. Karsten Weide (13:04.756) Yeah, so Amazon stuff is quite impressive. A big part of the agentic story is it opens up commercials to the mid-market and even to SMBs. Which... They don't have the resources, they don't have the money, they don't have the know-how, and now they can create a commercial within a couple of minutes. It's just, it's like magic, yeah. Brett House (13:31.28) There's a democratization piece to this. Rio (13:32.75) I think especially SMBs, right? Because they're the ones who really struggled to pay for photographers even or pay for videographers, Brett House (13:45.627) Yeah, it's really turned that production model, which kind of started in the television video world, And kind of started to break into digital. It just breaks that whole model and says, you don't need heavy duty production facilities and budgets and people to produce these things, especially at, and this is a relatively recent development where you could create thousands of variations of video. Karsten Weide (13:53.072) to break into digital, it just breaks that whole model and says you don't need heavy duty production facilities and budgets of people to produce these things. Especially, you know, and this is relatively recent development where you can create thousands of variations. But question, and with both of you, so it's getting easier to make video, the cost of it's going down, the quality by AI generating videos going up, the quality is incredible. Rio (14:09.966) But question, and especially for both of you. So it's getting easier to make video, the cost of it's going down, the quality by AI generated, like videos going up, the quality is incredible. Karsten Weide (14:24.056) I think one of the issues if you look at how local television ad sales works, right? It's you have like all the legacy MVPDs, Like Charter, Spectrum, Cox, whatever. They all basically have reps who run around, stop in pizza shops, stop in bowling alleys, and they sell them on local television. That's how it works, right? They've tried to roll out these sole service platforms, kind of like Metahaz. I think one of them quoted me, they had a six or 7 % adoption. Rio (14:24.238) I think one of the issues, if you look at how local television ad sales works, right? It's you have like, like all the legacy MVPDs, right? Like charter spectrum, Cox, whatever, you know, they all basically have reps who run around, stop in pizza shops, stop in bowling alleys and they, they, they sell them on local television. That's how it works. Right. They've tried to roll out these self-service platforms, kind of like Metahaz. I think one of them quoted me, they had a six or 7 % adoption of the last one. These are a few years ago. Brett House (14:44.827) Yeah. Karsten Weide (14:52.572) the last one is a few years ago. I don't think it's really AI enabled at all. No one's using it, right? They still rely on the local reps. My hypothesis is just because they haven't pushed out the right tools. Meta's proven, Google's proven, if you have really great self-service tools, SMBs will use them and they'll spend a lot of money when you aggregate it all together on them. Who do you think's gonna own that? Because I don't see the big guys moving quickly. I don't see them changing their model. Do you think it'll be like the Pubmatics device? Who's gonna actually? Rio (14:54.302) I don't think it's really AI enabled at all. No one's using it, right? They still rely on the local reps. My hypothesis is just because they haven't pushed out the right tools. Meta's proven, Google's proven, if you have really great self-service tools, SMBs will use them and I'll spend a lot of money when you aggregate it all together on them. Who do you think is going to own that? Because I don't see the big guys moving quickly. I don't see them changing their model. Do you think it'll be like the Pubmatics device? Who's going to actually... Will it be Amazon? Like who's going to be the ones who actually get these tools in the hands of these SMPs so they're using them? Carson, let's start with your thoughts. Karsten Weide (15:20.186) Will it be Amazon? Who's going to be the one to actually get these tools in the hands of these SMBs so they're using them? Carson, let's start with your thoughts. OK, good. So it's definitely Amazon. It's Google, and Google because they already have an established relationship with SMBs in search. So they already have the distribution. It's going to be certainly Facebook or Meta. Brett House (15:41.114) Yep. Yeah. Searching local listings. Yeah. Karsten Weide (15:49.84) Because, yeah. And then in the indie segment, yeah, there's really nobody there who specializes in SMBs, right? Even though, you know, the trade desk, for instance, right? So growth is slowing. So I think, shouldn't they go into the mid-market at least? Well, possibly. But I'm not so familiar with what Brett House (15:50.18) Yeah. Well, Meta's been there forever. mean, Meta and Google built their businesses on long tail advertisers. Rio (16:12.876) Isn't Yahoo doing that? Brett House (16:13.413) Yeah. Karsten Weide (16:19.238) what Yahoo is doing in terms of SMB stuff. But you could argue that the trade desk should tackle that, right? Just take the trade desk platform, shrink it down, simplify it, super automate it, don't have too many levers and screws that you can turn. they will argue, okay, we target the Fortune 500, and if we go mid-market or even smaller, then that will dilute our brand. I going to say, that's almost the opposite of their strategy today. I've run into them at big global farmers. A trade desk representative will be... Sometimes even one client would actually have desks and they were coming every day and they do their annual plannings with them. it almost seems like I agree with you, that would be an interesting opportunity, but it seems completely contrary to the way they're doing. Well, said for heaven's sake, spin out a different brand, so you don't damage your core brand. Rio (16:47.798) I was going to say that's almost the opposite of their strategy today. I've run into them at big global farmers. will have a trade desk representative. Sometimes they even have one client. Remember, they actually had a desk and they were coming every day and they do their annual plannings with them. it almost seems like I agree with you, that would be an interesting opportunity, but it seems completely contrary to the way they're doing it. Karsten Weide (17:15.854) or even spin off a different company, right? mean. Brett House (17:18.072) Yeah, well, it's a different delivery model of the service, right? As opposed to kind of a managed services model that's supported by the agencies, right? You'd have to have local agencies potentially managing those local clients and being able to self-serve within the platform. Karsten Weide (17:25.34) Yeah, Rio (17:27.596) It have to be. Rio (17:33.998) That's a good point, Brett. It's not the big, even if it is part of a whole code, it'd like, let's say the local agency, this part of like, or it would be an indie local agency. They're the ones you're right here. Because these local agencies, if the local businesses, if they work in agency, it'll be a small local agency, it won't be a big one. Brett House (17:42.788) Yeah. Brett House (17:51.483) Yeah, because there's going to be small local agencies, small local companies that are never going to log into a platform, regardless of how many years you look out, forecast out into the future. They're still going to have local agencies that have expertise. We did this, you know, I did this at a startup called Sign Up that competed with Yax for local listings. And the majority of their business got killed in COVID, but the majority of their business was local agencies managing the mom and pops that all required Karsten Weide (17:54.748) that are never going to log into a platform, regardless of how many years you look at forecast that in the future, they're still gonna have local agencies that have expertise. did this, you know, I did this a startup called Sign Up, the computer, the X for local listings. And the majority of their business, got the children's in COVID, but the majority of their business was local agencies managing the mom and pops, they all required up-to-date local listings so that even when the restaurant was open or the... Brett House (18:18.566) up to date local listings so that you knew when the restaurant was open or the retail store, the laundromat, hours, holidays, because the local companies just didn't have the level of sophistication or interest. They're dealing with the day to day of the business. Yeah. Karsten Weide (18:33.679) I think there's probably hundreds of thousands of little mom and pop shops who don't even know how to log into a platform, right? So they always need handholding. But I think there's a huge segment of SMBs who can. And for them, the world's their oyster. Brett House (18:42.479) Yeah. Brett House (18:48.346) Yeah. Karsten Weide (18:51.994) They can suddenly they couldn't do stuff that was impossible before, right? So that's a huge part of the story, think is SMB and mid market. Brett House (18:59.918) Yeah, yeah. And so you think that that's a growth area for some of the DSPs, the trade desks, et cetera, to potentially move down market with self-service tools. Karsten Weide (19:05.296) Yeah, so what I'm saying is more growth needs to come from somewhere. So what can they do, right? Go into apps, go mid-market. I mean Brett House (19:13.286) Yeah, there's only so much penetration into the Fortune 1000 that you could make. Yeah. Yeah. Rio (19:14.413) Yeah. Brett House (19:22.694) Yeah, Rio, what's their global growth look like in terms of like, let's say the big DSPs, but let's use the trade desk as an example. And then how global are they? You we looked at live ramp, we talked about them yesterday and live ramp is the definition of not global, US only 99 % of the business. Karsten Weide (19:34.876) the Rio (19:39.086) Oh yeah, their ex-US businesses were described to me by someone as like, what they call it, I think they call it a charity or something like that. And they had 90, was 95 % like the US and I know they had offices in other places, but they just did not have the signals or the market awareness or anything in any other countries. That's a good question, Brett. It's funny. I actually have a bunch of CPG clients that are global and they're all heavy on the trade desk. Brett House (19:46.98) Yeah. Karsten Weide (20:04.134) They're all heavy on the trade desk. They do tons of open web work. The trade desk is their main. Yeah, they will get many different DSPs. In fact, you look globally, they might work for like 18 DSPs. It just sounds like a crazy number, but you realize there are some DSPs that might specialize in APAC and specialize in this one specific sector in APAC. There's a couple in EMEA that are just aimed at publishing, for example, that you've never heard of here. So that number is not crazy, but most of them do work with the trade desk. Brett House (20:05.509) Yeah. Rio (20:06.104) They do tons of open web work. The trade desk is their main. Yeah. They will. will get many different DSPs. In fact, you look globally, they might work on like 18 DSPs, which just sounds like a crazy number. But when you realize there are some DSPs that might specialize in APAC and specialize in this one, like this one specific sector in APAC or, mean, there's, there's, there's a couple in, in EMEA that are just like, just aimed at publishing, for example, that you've never heard of here. So that number is not crazy, but Brett House (20:30.139) Yeah. Rio (20:31.99) Most of them do work with the trade desk. so I would imagine that's a good portion of the business. I don't know, Carson, if you know anything about it. Karsten Weide (20:33.98) And so I would imagine that's a good portion of the business. don't know, Carson, if you know anything about it. So they are, what state's international business, like 25%, 30 %? So it's not super huge. Actually, that's one growth opportunity that they have. I do think- 25, 30 outside of the US. Yeah, correct. And I think that- 50 % of ad spend in the US, that's actually not too bad. can think about it. Rio (20:51.054) So 25, 30 outside of the US. Rio (20:57.048) Well, 50 % of ad spends in the U.S. That's actually not too bad when you think about it. Brett House (21:01.22) Yeah, when you think relative to other markets, yeah. Karsten Weide (21:04.188) Yeah, I mean, you look at Google, it's more than 50 % is abroad. And yeah, it's 55 % is broad. That I don't know the top of my head. I think the bottom line is, I mean, I think there's potential for somebody like the trade desk. And I do think that they, in particular, they've decided to, oh, we need to get the US business right. Brett House (21:08.184) Yeah, more than 50 percent. Okay, that's interesting. Rio (21:13.804) Meta is the same, think, too, Karsten Weide (21:32.696) First, for that we need all of our resources and all of our attention. We take opportunistic business in Europe and in Asia, but that's it, right? Whereas I'm saying, hey, maybe focus a little more on international expansion, right? But that's kind of detour a little bit, right? Coming back to AI and what, agentic AI and what it can do today, you're looking at, as I said, internal point solutions, but even they are highly, productive, right? mean, it's, first of all, you have two senses, right? One sense is because of AI, energetic AI, you become more efficient, right? Brett House (22:02.979) Yep. Brett House (22:13.7) Yeah. And this is internal organizations building, you know, in some cases it's Python developed products, but they might be specifically oriented around a skill or a use case internally developing these things as efficiency, productivity enhancers. that? Yeah. Karsten Weide (22:22.908) Yeah, that's correct. that's the effect that the AI part has, but what it also has is it's a natural language interface. So, you know... Many, many marketers are not very tech savvy, right? Brett House (22:39.64) It democratizes that ability to compound the value in productivity. Rio (22:47.822) Okay. And by the way, just quick, quick point here before we keep going. This is very, this is, did, I, you know, this is very interesting. This is not what I thought. U.S. and Canada revenue for Metta are between 28 and 39 % of revenue. Rest of world is 61 to 72%. So they are like, like much heavier international. Obviously ARPU is much higher in the U.S. versus international, but the majority of their business is international, which I think is very interesting. Brett House (22:51.27) She looked it up. Karsten Weide (22:52.54) This is very interesting. This is not what I thought. US and Canada revenue from Metta are between 28 and 39 % of revenue. The rest of the is 60 to 72%. So it's much heavier international. Obviously ARPU is much higher in the US versus international, but the majority of their business is international. It would be interesting to see the country break down because they're banned from China. Brett House (23:04.59) Okay. Yeah. Brett House (23:15.332) Yeah, it would be interesting to see the country break down because they're banned from China, right? So there's a huge population miss there. They're probably huge in India and parts of Southeast Asia. Europe, Europe's small and it's, there's probably say heavy, there's certainly heavy penetration there, but population wise, they're not quite. Karsten Weide (23:23.644) They're probably huge in India and parts of Southeast Asia. Europe is small. There's probably some heavy penetration there, but population-wise, they're not quite as Well, Europe's about half the size in terms of advertising spending. It's about half the size of the US business, so it's significant. But of course, the problem is that it's so fragmented, right? Right. And we've got the three big markets, UK, France, Germany. Rio (23:45.87) facture do. Brett House (23:46.084) Yeah, yeah. Karsten Weide (23:53.403) than Spain or Italy or Greece. Turkey is big, but the other ones are small. So. Brett House (23:56.357) Yeah. Rio (24:01.816) So Karstead, actually, I think your hypothesis is, it sounds like it could be correct. mean, if Meta is able to do it so effectively, Google is too. mean, open web business is global, right? I in theory, the trade desk could focus on that. I'd almost say that's low hanging fruit for someone to nail it. Karsten Weide (24:21.21) Yeah, and in terms of localization, we just said it's fragmented, right? In terms of localization, AI can help with that also. So yeah. Brett House (24:30.086) So pivoting the, let's take a dive into your report. So the top 10 trends of an advertising in 2026. the central sort of thesis was that 2026 is not going to be an incremental year compared to all the other 30 years that you've been in the business, Karsten. What makes you, what are the forces? I mean, you talked about a bunch of different forces hitting it once. What do you think is most directly impacting? Karsten Weide (24:35.772) and that was in 2026. The central sort thesis was that 2026 is... Compared to all the other 30 years that you've been in the business, as you know, what makes Brett House (24:57.306) this transformational change that's hitting us right now. Karsten Weide (25:01.392) Yeah, so I've been in business for a long time, right? I've been in &E for over 30 years. I have to tell you, I've never seen something like this. mean, the rapidity of the change and the volume of the change, I think is just really unprecedented. It's not just the coverage of the change. It's not just breathless like hype and hyperbole. This is actually the literature where things are changing. this is actually happening. Brett House (25:17.892) Yeah, and it's not just the coverage of the change. It's actually what we're seeing on the on the. Yeah. Rio (25:21.418) It's not just breathless like hype and hyperbole. This is actually legitimately things are changing. Brett House (25:26.744) Yeah, and you categorize it as sort of AI, there's privacy, there's supply chain restructuring, right? Identity loss, measurement pressure, there's a whole bunch of topics that you cover in that report. what do you think the most important things are? Karsten Weide (25:30.172) There's supply chain restructuring, identity loss, measurement pressure. There's a whole bunch of topics that you're covering. What do you think that So I think the two most important ones are AI and the collapse of the supply chain, basically. So the supply chain will change a lot. You mean programmatic supply chain? Yeah, yeah, yeah. Rio (25:54.19) Programmatic supply chain, assuming, not like the product supply chain. Okay. Brett House (25:56.645) Yeah. And are we talking about an elimination of the internet, an elimination, but a consolidation of that intermediary layer between sort of the data owners, the publishers and the brands and advertisers that are spinning against that? Karsten Weide (25:57.661) Are we talking about a consolidation of the intermediary layer between the data owners, the publishers, and the brands and advertisers that are spinning it? So one major issue that's happening there is what I would call the phenomenon of unified advertising platforms, as I call it. And there's like five different names for that out there. maybe my UAP creation will make the race. That'd be great hurrah for me. Maybe that too, yeah. But you look at, I know, unidentified advertising. Area phenomena, I guess. Brett House (26:32.002) Isn't that the new term for unidentified aerial phenomena? That's like the new word for UFOs, right? UAP. Aerial phenomena, I think. It's not always aerial. Apparently they're in the oceans as well. Rio (26:49.312) UAPs. You're watching too much Joe Rogan, The second half of this, we're going to talk about the secret chambers under the pyramids. Karsten Weide (26:55.868) So I love this. I love mysteries. we can talk about that too. But so what you what you see is that, you okay, so you've got the trade desk and the trade desk says, or other DSPs, they say, Hey, we're going to connect directly with the publishers, we're going to cut out the SSPs. If the SSPs can add value, great, then they can demonstrate that to the publishers and they have the Brett House (27:00.614) They keep discovering them! Rio (27:04.824) Catacombs, right? Yeah. Brett House (27:23.394) Like with Open Path, yep. Karsten Weide (27:24.666) and they have a reason to exist, right? But if not, then too bad. And conversely, then the Pubmatic says, Trace asked if you can do that. We can do it too. We can play that game too. So we go direct to the buyers, right? Which they do. For instance, Pubmatic, from what I hear, the majority of their campaigns is now going direct. Rio (27:37.72) We can go direct, yeah. Brett House (27:38.982) Yep. Rio (27:52.398) I'm not surprised. Karsten Weide (27:53.307) So what you have is at the earliest stages, you have basically the springing into existence of walled gardens in the independent advertising system. So the trade desk will become end-to-end, Pubmatic will become end-to-end, OpenEx. That's also a different, another wrinkle to the story, right? Brett House (27:53.414) Yep. Brett House (28:10.096) Publicist group has become end to end with live Ram's acquisition. Arguably. That's a whole nother topic. Yeah. We talked about that yesterday about the publishing, you know, and, and WPB trying to build their own walled garden ecosystems with the data advantage, identity data, you know, matching advantage. Karsten Weide (28:26.454) I think you might get a situation where you look at publicists acquiring live ramp. I think you might be looking at a situation where all of a sudden they're like, hey, we're not an agency anymore. We're end to end. We're like Google. We're like Amazon. And they control billions of dollars of advertising. Brett House (28:36.592) Yeah. Rio (28:38.434) Yeah, yeah, yeah, yeah, yeah, yeah, yeah, Brett House (28:39.492) That's the trajectory, right? That's the competitive threat. Yeah. Rio (28:45.07) Yeah, and they worked directly with the brands and control the ad spend. And what's interesting too is there's that article about agentic trading and Omnicom's experimenting. And in the article, I'm not divulging anything here. They talked about how it would quote unquote cut out the entire ad tech supply chain. would actually agents negotiated with agents. I don't know if that's actually happening yet, but it is interesting to think about, Karsten, your point about Karsten Weide (28:55.076) I'm not divulging anything here. They talked about how it would quote unquote cut out the entire ad tech supply chain. I don't know if that's actually happening yet, it is interesting to think about, Karsten, your point about. Rio (29:10.424) things changing and this would be a complete rewriting of it, Karsten Weide (29:10.588) things changing and this would be a complete rewriting of it, Yeah, so I've already hearing the first voices that say, if I'm Procter and Gamble and I have a super smart buy site agent that is so smart it can replicate what PubMedic and the trade desk are doing, why can I not connect directly to the publisher? Brett House (29:33.272) Yeah, you're getting more out of every media dollar you spend, right? And you're also, and you're also actually, we talked to Joe Root at Permative just at this point, and top of yours was not to pile on, but you're also getting access to much richer audience data on the publisher side, which is not being passed through the DSPs right now because it is cost prohibitive because of the compute. That direct linkage is super powerful. Karsten Weide (29:42.044) But you're also getting access to much richer audience data on the publisher side, which is not being passed through the DSPs right now because it's cost prohibitive. Karsten Weide (29:58.493) Yeah, no, I agree. agree. Now, we're not even close to a scenario like that. Interesting, I had a call today with a little startup called Kova, QOVVA. The CEO is Kyle Dozman, the ex-CRO of Pubmatic. Very smart guy, of course. And they're like, hey, Carson, we are developing a buy-side agent. Rio (30:12.174) Don't know them. Brett House (30:17.978) Okay. Karsten Weide (30:26.972) But interesting, so which automates everything for for buyers automates just about everything except for the for pulling the trigger on buying inventory. That's still the human domain. And and and I saw I basically I said I gave him this grand vision of everything's automated machines do the buying and selling right. They're like yeah I know we're doing everything except for the transaction layer that's Brett House (30:30.619) Yeah. Brett House (30:39.021) Okay. Brett House (30:52.57) So everything would be figuring out your goals and objectives from a campaign, breaking them into very specific tasks, generating creative, audience segmentation and assembly. And then the launch, they can't, yep. Karsten Weide (30:59.26) Media planning, reporting, optimization. The system can even come up with a solution for, you can ask natural language questions. can say, hey, campaign XYZ that I'm running, what are the errors that you find? What can I improve? The system will come back and say, you can do this this and this to improve the campaign. Rio (31:26.606) Karsten, question about that. So I can see it's one thing to take AI to, because it can ingest information very quickly. It can make sense of it quickly, structured. It doesn't have to be totally normalized. It can actually take unstructured data and it can very quickly understand and give recommendations on it. then whether you let it implement those recommendations or not, you have a human to loop, that's another story. But it's one thing, like putting AI on top of existing systems. But what I hear some people talking about, Karsten Weide (31:29.244) It's just one thing to take AI to, because it can ingest information very quickly, can make sense of it quickly, structure it, can, it doesn't have to be normalized, right? It can actually take unstructured data and it can very quickly understand and give recommendations on it, right? And then, know, whether you let it implement those recommendations or not, I mean, that's another story, but it's one thing like putting AI on top of existing systems. mean, but what I hear some people talking about, Rio (31:55.054) is actually rewriting the entire supply chain using AI right now. That's very interesting. I don't know if it's actually possible at least yet, because you think about how complex it is when you look at the supply path is very complicated. All the different intermediaries, all the hops, even the things that we take for granted that these tools do. like, what do DSPs do? We've got to work with DSP. How do you handle identity? How do you handle frequency caps? How do you handle? And they're not getting ingesting all the information from... Karsten Weide (31:55.197) is actually rewriting the entire supply chain using AI right now. That's very interesting. I don't know if it's actually possible at least yet, because you think about how complex it is, and you look at the supply path, it's very complicated. All the different intermediaries, all the hot things that we take for granted that these tools do. Like, what are DSPs? Well, we've got to work with DSP. How do you handle identity? How do you handle frequency caps? How do you handle? And they're not getting just all the information from Bitstream. It's only a... Rio (32:22.958) from bitstream, it's only a small portion of it because they have to act so quickly, but they are providing an important and serious capabilities that enable programmatic advertising. there was that recently, I think it was someone from WPP actually have the article here, was Emily Lai from WPP Media Group is saying that it's too complex for AI to handle at this point, right? That it would actually crash the models because I guess it's just, Karsten Weide (32:25.134) of it because they have to act so quickly, they are providing an important and serious capabilities that enable programmatic advertising. mean, there was that recently, I think it was someone from WPP, actually I have the article here, was Emily Lye from WPP Media Group is saying that it's too complex for AI to handle at this point, right? That it would actually crash the models because I guess it's just. Rio (32:51.756) it's asking you to do too much. I mean, that could change over time as the models get better, but thoughts about that. mean, are we talking rewriting of it? Are we just talking to AI layer on top of it to like make things work better or is it, or do we not know? Karsten Weide (32:58.907) Yeah, here's how these things work. I don't think there will be ever a complete rewriting of the value chain. So everything that you do will sit on top of legacy platforms, code, data, and that's how it is. Rio (33:19.726) including the auctions today, right? Because you have to understand how much inventory costs and there has to be a bidding process, right? So you're not suggesting that goes away. You're suggesting like there's a layer on top of that that understands it better and maybe makes better decisions. Karsten Weide (33:19.868) Including the auctions today, right? Because you have to understand how much inventory cost has to be bidding processed. So you're not suggesting that goes away. No. You're suggesting that there's a layer on top of that that understands it better and maybe makes better decisions. Yeah. So the same way that many elevators today are still run by MS-DOS, right? It's just the way it is. And well, the New York City sub-links, which is still run, I think, the MS-DOS. But the moral of that story is of Brett House (33:31.365) Well... Brett House (33:38.822) that's interesting. Yeah. Yeah. And well, in the New York City subway station is still run, I think, by MS-DOS, right? Karsten Weide (33:49.053) of that anecdote is you always put on top of what's already there, which of course limits what you can do, but it is what it is. it also sounds like, you know, those three key protocols that sort of the, that may become the agenda of internet staff, right? Ad CP, right? Let me talk about that. Brett House (33:53.253) Yeah. Well, but it also sounds like, you know, those three key protocols that may become the agentic internet stack, right? Add CP, right? Love you to talk about that. The advertising context protocol, which sort of acts as the orchestration layer, not to belabor that word orchestration. The user... Karsten Weide (34:11.104) Yeah, so Brett House (34:15.43) context protocol, UCP, which sort of is your audience layer and then your ARTF, agentic RTB framework. Are those the three, kind of the big three when it comes to creating an agentic internet advertising stack? And can you describe those and define them for the audience? Karsten Weide (34:34.49) Here's the more interesting part of the story is that, so these are the top three that are always mentioned when you talk about agent to AI. There's frameworks, protocols, however you want to call them. There's a few others, like the agentic advertising management protocol, MCP is around, agent to agent is around. So these are just industry-wide attempts at standardization, right? Which is great. And my hope was, Rio (34:40.62) These are just protocols too, right? mean, that's all we're really talking. Brett House (34:42.118) Yeah. Brett House (34:52.453) Yep. Karsten Weide (35:00.732) we're going to have from the get-go, we're going to have an industry standard and everybody's playing by the same rules. But then you get the news, know, PubMedic has their own agentic OS, the trade desk is going to have their own agentic OS. And of course, what you see is, I think that the vendor's concern is that if I go with the industry standard, I'm going to lose leverage, right? And I may even eventually be disintermediated. So I have no interest to go industry standard. Instead what I'm gonna do is. You're inventing your own, right? Correct, correct. And the consequence of that is that for the next two or three or four years, you're gonna have like a dozen different protocols that compete with each other. And eventually there's gonna be a shakeout. Brett House (35:31.396) Yeah. Yeah. So it's almost a wall. Rio (35:33.834) You want to create a mode and the mode would be having your own protocol. That's wild, right? Brett House (35:39.278) Yeah. Well, yeah, it plugs right into that wall garden model, right? You're like, hey, we've got something that's specific to our ecosystem. Karsten Weide (35:59.357) and it's gonna shrink down to a few, but that's gonna take time. that's an interesting part of the story is that because of this automation via agentic AI, it really gives the independent ecosystem a lease on life in terms of competing with Google, instance, or Amazon. Amazon is really, actually I think that Amazon is... Brett House (36:02.81) But the open web, the open internet needs to have some way to compete with the walled gardens, right? And so shouldn't they be all aligned around a common set of protocols? Karsten Weide (36:29.18) greater threat to the independent ecosystem than Google is in some ways. Because it actually connects into the open web and whatever else is in the programmatic ecosystem. there's... Yes, even though I talked to some vendors and they're like... Brett House (36:42.084) Get with their DSP. Yeah, it connects into the television ecosystem with Amazon Prime. Although if it is anybody's. Yeah. Rio (36:46.902) And they're really investing in offsite too. mean, they're aggressively expanding in open web. Karsten Weide (36:57.02) I think that's the wish is the father of the thought is that they're going to shut down the network business. And I don't think it's going to happen anytime soon. would, if they did shut it down, would be good. But it's just. No, no, no, no, That's what I'm saying. No, no, no. Brett House (37:07.922) Interesting. Yeah. Well and not not to not to mention that they they have one of the largest ecommerce platforms on planet Earth They own Whole Foods Rio (37:15.974) So the rumor you heard is they're going to shut to all the networks, basically all the, like any all off, so anything that's not like, no, they're going to jettison that. would they? Karsten Weide (37:26.844) I think Brett is right, or you're right, Rio. I think you said it. They're doubling down on the network business. Doubling down on what? Doubling down on the network business. like third party inventory that goes through the Amazon GSP. Really hard. All of their AEs are being sent in to sell. No, you talk to DSPs. The DSPs will say, hey, that's a Rio (37:34.392) That's what I see. Yes. Yeah, for sure. Brett House (37:36.976) Doubling down on what? Okay. Got it. Rio (37:39.682) The network business, yeah. Rio (37:43.95) They're pushing that really hard. I mean, all of their AEs are being incented to sell off state inventory. I know that. Brett House (37:45.903) Yeah. Karsten Weide (37:55.719) it would be good for us if they shut down the network business because it would mean more inventory for us. But also the reason why they think it might happen is because they think that the bottom line is better for them with their own and operated inventory. the thing is, only so much inventory. When we talk to us about Dr. Fu, we're like, well, remember the question was, why would Meta have so much? Rio (38:14.712) Well, sure it is, but the thing is there's only so much inventory. When we talked to us about Dr. Fu, we were like, well, remember the question was why would Meta have so much, let's say, audience extension, network business when they're paying, they only keep 50 cents in the dollar for that Karsten, for all of the offsites versus everything on Facebook. Realist in that. So they're only keeping a fraction. But the thing is, when you look at the amount of money going through the pipes, there's no way there's enough inventory on Facebook to ever spend that. So they have to go. Brett House (38:19.76) Yeah. Karsten Weide (38:23.164) so much, let's say, audience extension, network business, when you're paid, they only keep 50 cents on the dollar for that, Karsten, right, for all of the assets versus everything on Or less than that, yeah. Less than that, right? I so they're only keeping a fraction. The thing is, when you look at the amount of money going through the pipes, there's no way there's enough inventory on Facebook to ever spend that. So they have to go, It's a strategic share of right? They want as much share of wallet for the advertising if there's not enough in... Brett House (38:40.836) Yeah, it's a share. It's a strategic share of wallet play. They want as much share of wallet for the advertiser. And if there's not enough inventory to come to to qualify for that within a network like Amazon, surprisingly. Karsten Weide (38:48.68) Yeah, so, so AI. Rio (38:53.134) Yeah, would you rather keep 50 % of a billion dollars or 100 % of a million, right? Brett House (38:55.654) Yeah, even if a lot of that inventory, know, and Fu's point is that once you go outside of the walled gardens into the open internet for these audience extensions, the inventory quality just goes into the toilet. Right. And so you're getting these kind almost kind of empty impressions in a sense. That was kind of his point. Karsten Weide (39:16.124) collapse of the supply chain there's a couple other things so one is advertising on chatbots so so so this relatively small at this point but I think it's gonna be big and I think that it's a it's a few hundred million dollars all told Brett House (39:24.422) yeah, yeah, I wanted to get into that for sure. Brett House (39:34.138) Yeah. Rio (39:34.146) How big is it right now? Because an opening eye has some of it for... Brett House (39:40.74) Yeah, their annualized revenue is about $25 billion OpenAI specifically, right? Karsten Weide (39:44.699) Yeah, so it's a fraction of what they make via subscriptions and via enterprise. But it's the super early days. Also in terms of the ad tech that they have. I mean, you look at chat GPT and their advertising. Let me switch this off here. But you think about it, they have 25 billion and it's climbing pretty quickly. I it's growing 10 % a year. I think it's growing probably 10x or something like that. And then like it's, and that's with charging people. Brett House (39:50.0) Yep. Rio (40:00.942) Well, but you think about it, they have 25 billion and it's climbing pretty quickly. I think it's growing 10 % a year. No, I think it's growing by like 10X or something like that. And then like it's like that. that's with charging people, charging consumers $20, $200 a month and charging businesses for the APIs, right? I mean, it's, that's a huge business is growing very quickly and, advertising is adding incrementally to whether it's 5 % or 10%. Karsten Weide (40:23.728) Yeah. Correct. let it rip once they perfect it. And all these people who have been saying, I mean, I was an 11th friend of the party, know, I think he's been a, he's been a critic when it comes to advertising and AI, there's no way these companies won't let it rip if they can. Rio (40:25.464) There's no way they're not going to let it rip once they perfect it. And all those people who have been saying, like an Aero's 11 friend of the pod, I think he's been a critic when it comes to advertising and AI. There's no way these companies won't let it rip if they can. And I don't think we know what the format of the ads will be. mean, we have no idea, right? Karsten Weide (40:44.218) I don't think we know what the final format of the adverts the ads. No, it's like it's like ads super primitive at this point, right? I mean, yeah. Brett House (40:49.732) Yeah. Yeah. And I, I, and I sort of, I do appreciate the fact that they're, being careful about how they introduce it because it is a different consumer experience. it's in, in, in, yeah. And, yeah, well, privacy, also just, you know, I don't, I am going to a chat bot for a different reason, for an answer to a question that I have. And I don't want to be shilled some product that's not relevant to what I'm asking about. Right. Karsten Weide (41:00.444) Privacy is a big issue. privacy is a big issue. Karsten Weide (41:19.287) I was like, oh, they're going to make a lot less revenue. But they didn't. It's more. So that was kind of like unexpected. Brett House (41:28.379) Yeah. Brett House (41:32.474) Yeah. Rio (41:33.368) Yeah, Google's continued to grow. The Google stock's gone up, mean, like not only are they monetizing AI now, but the search business has gone up as a result of this too, Carson, you're right. Brett House (41:36.505) Yeah, they've Karsten Weide (41:38.885) Yeah, no, I agree. so, yes, all of these outfits are desperate for money, right? And even a small incremental advertising revenue stream is helpful. So that's why I think that guys like guys who said they get out of advertising, like, like, like, entropic and perplexity, I think they're going to be back. And here's one reason why. Brett House (41:48.741) Yeah. Brett House (42:01.625) Yep. yeah. Karsten Weide (42:05.602) I mean, one might argue, hey, we're just going to go IPO. We're to have all the capital that we need for the next five years. But if you look at the size of these IPOs, just one of them is more volume than all startups financed by VCs over the last 10 years, right? Which also means whoever goes first soaks up all the money and the others are like, that's too bad, right? Rio (42:30.784) Open eyes, open eyes, open eyes, filing right now. mean, the thing is, I want now that SpaceX is IPO and getting so June 12th, right? mean, open AI. Karsten Weide (42:36.752) I haven't checked the S1 yet, but I will it's interesting, you know, it's Yeah, I think they're they're gonna come back they're gonna be back Rio (42:40.546) Yeah, mean, opening, I was going to be right behind it. I mean, and then they'd be crazy not to be. And then my guess is Anthropics not too far behind them. I mean, I think they all have to go. They have to go this year. Brett House (42:49.606) Yeah, and the expected valuation is what, one trillion for OpenAI? Or either in 60, like close to either 900 billion, one trillion, somewhere in that range. It's gonna be one of the largest. Karsten Weide (42:59.727) Yeah, it's crazy times. Well, they have 800 million active users, right? As a couple of months ago, that was just an insane number, right? And you think about how quickly it scaled up. And not everyone's paying, right? But do you know what is the, is it? Our poos gotta be pretty high. I think for Google on it was for search, was like three, 350. Brett House (43:04.518) 500 billion to 750 billion, that's what I just saw. Yeah. Rio (43:05.624) Well, they have 800 million active users, right? As of a couple of months ago, that is just an insane number, right? When you think about how quickly it's scaled up and not everyone's paying, right? But do you know what is the, is it? Our poos got to be pretty high. I for Google on it was for search, was like three, three 50. Brett House (43:10.82) Yeah. Rio (43:26.072) per month, I think was the average revenue per user. Right. So it was not a lot, but obviously everyone's using Google. So, but it's like, was, it was the greatest business ever created. Right. These. And that's what I found interesting. So Google needed to monetize search. They found a way to do it. Right. You could, you could argue they've made it a shitty experience by having pages of blue links. People still use it. Right. And SharePoint early Carson, they've rolled out AI answers. They've included that they've included paid results in that. And it not only is not Karsten Weide (43:26.246) per month, I think, was the average revenue per user, right? So it was not a lot, but obviously everyone's using Google, right? It was the greatest business ever created, right? And that's what I found interesting. So Google needed to monetize search. They found a way to do it, right? You could argue they've made it a shitty experience by having pages of blue links. People still use it, right? At SharePoint early, Carson, they've rolled out AI answers. They've included paid results in that, it not only is not Rio (43:56.622) people have not bailed on it. More people are using it and it's monetizing even better. Within Gemini and the app, have ads and OpenAI for the free version has ads and they're experimenting. this is, going to be, so I don't think it needs to be the whole thing. Like with Google search, they needed advertising to work. Otherwise there was no business, right? They tried selling other types. They tried selling like a marketplace business. They tried, never worked. So they went into what we see with, what we see with the sponsored links, right? But for AI, Karsten Weide (43:56.765) people not bailed on it, more people are using it, and it's monetizing even better. Within Gemini and the app, have ads, and OpenAI for the free version has ads, and they're experimenting. this is, gonna be, so I don't think it needs to be the whole thing. With Google Search, they needed advertising work, otherwise there was no business, right? They selling other type, they tried selling like the marketplace business, they tried, never worked, so they went into what we see with the sponsored links, But for AI, Rio (44:27.032) They're making so much damn money just on the subscription business. This is incremental to your point before. I think it's going to be big, but it doesn't have to be as big as search was. Karsten Weide (44:27.164) They're making so much damn money just on the subscription business. This is incremental to your point before. I think it's going to be big, it doesn't have to be as big as search was. Yeah. Well, no, I've got some R food numbers. So open AI, the approximation is $2.10 to $2.40 per weekly active user. For weekly? Yeah, per weekly. That's what it's about. So I think Google's $3,350 a month. Yeah. And that's per week. So it's already. Yeah, $3 $3.50 per user per month. Brett House (44:36.038) Yeah. Well, no, I've got some RPU numbers. So OpenAI, the approximation is $2.10 to $2.40 per weekly active user per month. Yeah, per weekly. That's what it's saying. It's sort of... Rio (44:45.656) per weekly. So if Google's about, I think Google's $3,350 a month, and that's per week. it's already more than 2X without advertising really being anything. Brett House (44:51.278) Yeah, yeah, but yeah, three to three three fifty per user per month for Google. Yeah. Karsten Weide (44:55.772) So it's already more than 2X without advertising really being anything. So it's, but then also, as we said, it's super primitive, right? So they have like, so chat GPT runs on Criteo right now. It's not like Criteo has a platform that could, so let me take a step back. So this chatbot advertising thing, I mean, the promise is, it's going to be a whole new form of advertising. Brett House (45:00.591) Yeah. Brett House (45:04.197) But. Karsten Weide (45:25.676) going to be a step change in terms of quality and accomplishment and effectivity. none of that is there. Brett House (45:31.342) Yeah, closer to real user intent. Rio (45:36.94) It'll probably just be a bunch of banners. I mean, I'm not kidding, Brett House (45:41.85) Totally unrelated to what I asked about. You I can see if you're doing product questions or questions about travel, some sort of purchase experience, that's when the recommendations sort of fit in. Like I'm planning a vacation and then you get recommendations just like you do in Gemini, in Google search. Karsten Weide (45:50.013) That's when the recommendations sort of fit in. I'm a vacation, and then you get recommendations to do a chat with them in general and do the search. So yeah, that's, you know, so the AI advertising on chatbots, that's going to be big, I think. Again, Rio, you're right. I think it doesn't have to be as big a search. All they need is an additional revenue stream, But then fourth factor that turns things upside down is That's kind of a more longer term proposition that's, that's agent commerce. So. that's going to happen. That's going to happen. It's going to happen. Cause I don't want to spend an hour on, the internet to just to find size 10 red Brooks sneakers. Right. I tell my agent, go out, get a deal, find the best price. want free returns. Rio (46:25.518) Do think that's legitimate? Eric Sufer is always saying it's a mirage. People are morons for thinking it's going to happen. Explain. Brett House (46:31.802) Yeah. Karsten Weide (46:48.489) and the thing goes out and does it for me. Brett House (46:51.268) Yeah. Well, the universal cart is going to make that a lot easier too. I mean, I think, I think to me that seems like when I started reading the releases around the universal cart from Google, that to me really made a lot of sense where you could sort of allow that to go and buy, especially core. Karsten Weide (47:01.148) really made a lot of sense. sort of allowed that to go in by this. I didn't even find out what it is. Brett House (47:07.992) like items that you buy on a regular basis, like CPG items, toothpaste, things like that that you buy on a recurring basis. The fact that it can go to multiple e-commerce sites and you don't have to deal with the friction of having to enter in credit card information, et cetera, once you're... Karsten Weide (47:12.826) Yeah Rio (47:22.862) Or even learn how each site works and how to navigate. It'll just do it for you. Right. That's a huge time saver. You know, you know why I think it's going to work. And the people who were saying, well, I just don't like rewind 25 years. No one bought anything online. I remember I bought a book from Amazon in 1996, right? That was the first online purchase I made. I remember I didn't even want it. I didn't have a credit card. I don't think of that, but maybe I did. Like maybe had a debit card. I don't remember. Right. But I remember I. Brett House (47:25.838) Yeah. Brett House (47:36.39) Yeah. Karsten Weide (47:42.141) I remember I didn't even want to, I didn't have a credit card. don't think about maybe I did. I can maybe I had a Derek card. I don't remember. Right. But I remember I bought it online and you had an option to do COD cash on delivery, right? So I actually paid the UPS guy when he showed up. I don't think I'm listening. You know, I, I, I still remember the trepidation that I had when I did my first online purchase on CompuServe buying a U S robotics modem. I'm like, Brett House (47:45.945) Yeah. Rio (47:49.358) bought it online and you had an option to do COD, cash on delivery, right? So I actually paid the UPS guy when he showed up. I don't think that exists anymore, right? Rio (48:02.925) Yes! Rio (48:08.812) It's probably really probably scared to shit out of you. And the thing is, and I remember like people like I heard this story, apparently at Amazon in the early days, people were so afraid to put their credit card in that they would actually they would call in and give them their credit cards. They could keep them by phone, which is no less secure. Right. Someone actually has to write it down or they would email it. Brett House (48:10.736) Yeah, yeah. Karsten Weide (48:11.548) I remember people, like I heard this story, at Amazon in the early days, people were so afraid to put their credit card in. They would call in and give them their credit cards. could come in by phone, which is no less secure. Or they would email it. I think you're totally right, I think you're totally right, Rio. So it's going to happen. that's not the big question for me. To me, the big question is, what does it mean for advertising? When bots buy. Brett House (48:26.21) And the cash out delivery, that is hilarious. That was a prank we used to play when we were young. We used to call like, I'm sorry. Brett House (48:39.717) Yeah. Karsten Weide (48:41.328) What does it mean for advertising? Because if the bots buy it, not a human being, then why would you advertise? so. I this gets into the question of what is an agent. Rio (48:50.03) Well, yeah, I remember, because this was, Brett House (48:50.182) Yeah, well, there's human agency behind the bot buying, right? There's a plan, a strategy, a prompt. Rio (48:56.266) No, no, no, no, no, Like, what if like, like would bots advertise just to other bots? Right. The thing is like, and I guess this gets into the question of what is an agent Karsten, right? So you think about an agent is someone who represents you and has fiduciary responsibility to you. Right. So if an agent's being representing you, it's being influenced by advertising. It does that. actually compromise. I don't know. It's a very interesting philosophical question, but so. Karsten Weide (49:05.99) Right so you think about an agent is someone who represents you and has to do share responsibility to you, right? if an agent's being representing you is being influenced by advertising does that I actually compromise other it's a very interesting philosophical question Brett House (49:06.938) Yeah. Brett House (49:12.378) Yeah. Brett House (49:21.23) It's like an egyptic twin or a digital twin. Rio (49:23.186) And certainly advertising wouldn't look the same. It wouldn't be banners, but you'd want to make sure the agents know about the latest prices if they're looking for you. Karsten Weide (49:31.066) So think that there's three items there, I think. So one is it's bad news for Google, think, because so direct response advertising will directly be affected because you don't have to convert humans anymore. You have to convert agents and buy purchase bots or shopping bots, right? So that's the second factor is there's going to be botvertising, right? So advertising to bots where you say, here's a special deal. Here's stuff that others don't do on top of the Brett House (49:31.237) Yeah. Karsten Weide (50:00.733) product. So that's going to be that. Brett House (50:01.669) Yeah, Yeah, and if your personal shopping agent is anything like just the early days of us getting involved in sort of setting up, you know, your GitHub and setting up rules and instructions with all of your, you're basically programming your own agent or your own LLM. Yeah, around how you want, like, no, but how you write, how you, you know, so you're gonna continue to train these agents on. Rio (50:18.158) Just store your credit cards, just storing all your frequent flyer and frequent. Brett House (50:27.659) what you like to purchase, what your favorite colors are, your propensities. You want to get this to know as much about you and to be as much a digital twin. Karsten Weide (50:31.972) Hey Rio (50:34.51) Right, and they'll get better and better. The more they do it, the better they'll get. You think about it, like the first few times you're probably like, like right now I would not want an agent buying a plane ticket for me, but let's say you had an agent go out and like 30, 40 times, look at tickets, give you recommendations. And after it learns when you like to fly, it learns what seats you like. It knows it has all your credit cards. Why wouldn't you just let it do it? You can always return it within 24 hours anyway. Brett House (50:53.87) Yeah. Yeah. Karsten Weide (50:56.412) My friend, I'm so bad at booking travels, there's no way that an agent could be worse. so Brett House (51:01.766) No, but you'd give it that instruction booklet. You'd give it that markdown file and be like, you know, now, so to me, that's when agents can start to, like, if you've got an agent on your behalf that you've trained, that you've built, sure, because it's going to, you know, you get a human agency involved where it comes back to you and asks, like, is this okay? Please confirm. Rio (51:02.239) Ha ha! Rio (51:07.886) Give it to Markdown for Lyle, yeah. Rio (51:22.99) Yeah, but after 10 times asking you saying it's fine, do it, you're just gonna say, buy it for me. Send me the receipt. There's no way this doesn't happen. There's no way this doesn't happen. Brett House (51:28.196) Yeah. Karsten Weide (51:28.407) So, yeah, it's gonna be it's gonna be a development. So, but the third item for advertising is it's gonna be a renaissance of display advertising or brand advertising at large, right? So, because the bots need to be programmed by their users and how can you influence the bot? By influencing the user who Brett House (51:34.522) Yeah, in one form or another it is. Rio (51:42.433) Interesting, okay. Karsten Weide (51:52.829) who programs the bot, right? So there's gonna be display, video, all of this is gonna be important because you can still influence humans before they customize their bot. So this is going back in time, like the Renaissance, old time display. It's the funniest thing because the talk in the industry is, hey, everything's going performance, performance, performance, right? And it's gonna be more squishy again, right? In real life, but it's more important too, because I give people. Brett House (51:54.787) Yeah. Rio (52:02.623) yeah, so this is going back in time, like a Renaissance of old time display and like, Brett House (52:12.292) Yeah, yeah. Rio (52:18.062) In in real life, it becomes more important too, because like if people, agents are doing stuff, you're probably going to be spending less time in front of screens, least less time shopping in front of screens. You might be doing other things, right? So if that's the case, where can you influence people? Maybe it's in real life, maybe out of home as a big resurgence. Karsten Weide (52:22.172) If you're agent you're probably going to be spending less time in front of screens, or least less time shopping. You might be doing other things, right? If that's the case, where can you influence people? Maybe in real life, maybe out of home as a big researcher. Out of home is one of the most underappreciated channels that there is, just like audio. Super effective, relatively small, but effective. And it's kind of ignored by the industry. Brett House (52:27.717) Yeah. Rio (52:39.618) That and audio. Yeah. Rio (52:52.928) IRL in real life. Karsten Weide (52:54.068) Yeah, it's true. Yeah. So there you have it. Sure. How do quick hits work? Oh my God, I'm going to I'm going to embarrass myself. Rio (52:57.262) to be more important. interesting. Brett House (53:02.84) Yeah. So should we get a, should we move to quick hits? Rio (53:06.872) do quick hits. We're going to rip through them and you give us your quick opinion on them and we'll move on to the next one. Brett House (53:17.688) Yeah, yeah, yeah. Hopefully we don't stump you completely. We're not going to give you complete softballs, but, right. So what are what? Yeah, exactly. We're like. We're going to test your brain capacity in a Friday afternoon. Most yeah, most overrated assumption about ads in AI. Overrated, underrated. Let's do overrated, underrated. What's the most overrated? What's the most underrated assumption? Karsten Weide (53:21.052) know, I know. You're widely like that. Overrated is... Rio (53:23.47) We saved the hardest ones for quick hits Karsten, sorry. Guests are completely stumped and look foolish. Rio (53:33.762) Before a holiday weekend, right. Karsten Weide (53:49.125) taking the vision for reality. just not there yet. It's super early days. The promise is big, it's gonna happen, but we're not quite there yet. Underrated is that AI has been around for a long time. Everybody acts like the agents are like the first time that AI is happening in advertising, but it's not. It's been around for a long time. Machine learning, yes. Yeah, Brett House (53:56.804) Yeah, underrated. Brett House (54:07.736) No. Yeah. Machine learning has been around for a long time and that's just a subcategory of AI as a general overarching category for sure. Rio (54:17.666) Which AI company is best positioned to dominate and come close to perfecting AI advertising of any form? Karsten Weide (54:17.729) Which AI company is best positioned to dominate and can close to perfect in AI advertising of any form? Of any form? Of any form, it's not Google. I thought you were going to say Google. Google's not. Well, if you include YouTube, maybe, yes. I didn't even think of YouTube right now. But yes, you're right. Rio (54:33.1) I thought you were going to say Google. Brett House (54:45.382) So what is what? It's Yeah. Yeah. So what is one practical step advertisers should take in the next, let's say, 12 to 24 months, although 24 months seems like an eternity in today's world, to prepare for AI-driven advertising? Karsten Weide (54:45.412) It's Google. Let's cut this before it's It's totally going to be Google. On YouTube, yes. Rio (54:47.758) Let's Google. Rio (54:53.614) On YouTube. Karsten Weide (55:06.492) The advertisers are in a good shape. don't have to do a whole lot. mean, the agencies and the DSPs are doing all the work. unless you have in-house stuff, right, then you have to do what DSPs are doing, what agencies are doing. have to, first of all, we have to hire people that know what they're doing. And I think it's going to be hard for brands to hire people like that. Because if I'm a If I'm an engineer who makes $500,000 for AI, I'm going to work for Google or a startup. I'm not going to work for Procter & Gamble or Walmart or Target. I'm going to work for Google or a startup. It's more interesting. It's sexier. Brett House (55:42.372) Yep. Yep. that's interesting. Yeah. Because of the capacity to pay or the eventual payout when it comes to it. Yeah. Yep. Rio (55:51.224) Doing cool work too. Yeah. Yeah. It's funny you mentioned that that's the one advantage that agencies have always had over in-house teams. Why every, every couple of years is like all the in-housing is going to destroy agencies. And the biggest challenge, the biggest challenge they have is recruiting and retention of top talent at these in-house teams. It's just doing the same thing year after year. Karsten Weide (56:04.028) agencies and the biggest challenge they have is recruiting retention of top talent. And a few years ago, you know, few years ago, there was all this talk, oh, everything's going to go in-house. I mean, there's some big outfits who do in-house stuff like Procter & Gamble, but I don't think it's that much of a widespread phenomenon. Every five to seven years, there's a wave of in-housing that goes back out. Exactly. Brett House (56:11.387) Yeah. Rio (56:26.536) Every five to seven years, there's a wave of in-housing and it goes back out. I mean, it's just cyclical. Brett House (56:26.63) You think that you think the job becomes inherently less interesting? Yeah. Karsten Weide (56:33.788) It's like, do you see, every five years there's a slew of new startups. They all get acquired. Then it's quiet for another five years. And then you have another, you know. Brett House (56:38.552) Yeah, it's probably the diversity of the job. within an agency, you're working with multiple clients rather than just one client in-house and the potential to embrace a bunch of new innovative tech and steps. Rio (56:50.242) And you can jump to new agency, right? The agency, can go across the street, New York, especially, right? You can switch agencies, Brett Sherpoint, you can switch accounts. And you're doing them more, just, yeah. Karsten Weide (56:50.48) and you can jump to new agency, right? The agency, can cross the street, New York especially, right? You can switch agencies, that's your point, you can switch accounts. And you're doing the work yourself, right? It keeps you a little bit more on the bleeding edge. Brett House (56:59.554) It keeps the diversity of the job, right? And it keeps you little bit more on the bleeding edge of tech and innovation where you may not be embracing that stuff as quickly within a big brand, for example. Rio (57:09.134) Yeah. And an agency, may start out as a 25 year old, just traffic or just doing like, like in platform doing maybe less interesting. Some people like ad ops, like, you, like, you do graduate and there's the ability to do more interesting things would be more in strategy side. I think with it, when in-house teams, it's pretty static, right. And that's the thing. they, they end, fact, you look at tenure, these in-house teams actually very long because people stay, but I think attracting the top talent is more of the challenge than retaining less than top talent. Karsten Weide (57:09.756) You can see, may start out as a 25 year old trafficker, doing like platform doing maybe less interesting. Some people like ad ops, like you do graduate and there's the ability to do more interesting things, be more strategy side. I think with an in-house teams, like it's pretty sad. And the fact you look at tenure, these in-house teams actually very long because people stay, but I think attracting the top talent is more of a challenge than retaining less than top talent. Brett House (57:36.656) Yeah. Rio (57:40.12) Karsten, what's the biggest misconception people have about the economics of advertising and AI? Maybe even people have misconception of people in the industry or outside of the industry. Karsten Weide (57:40.285) The biggest misconception about, can you repeat the question? Okay, I'm gonna answer this question, but you may have to cut it out for political reasons. Rio (57:53.73) The biggest misconception people have about the economics of advertising and AI or just anything about advertising and AI, maybe something that people don't, aren't thinking of, get wrong. Rio (58:07.854) This is an independent part. Brett House (58:08.474) There's no editing of this website, of this podcast. Karsten Weide (58:10.204) So all the, I mean, of course, all the proponents of AI say, layoffs, no, no, there's going to be no layoffs. Instead, we can reassign people to more qualified jobs. And I mean, who are we kidding? I mean, you look at the agencies, there's still super headcount heavy and technology poor in many cases. Not everybody is publicist or WPP, right? And so what's going to happen is you have a lot of junior people who Brett House (58:35.141) Yeah. Karsten Weide (58:39.452) do a lot of manual work right now, and in a couple of years, they're going to be gone. And they're not going to be reassigned. They're going to be unemployed. And that's the biggest misconception is that it's going to be all rainbows and unicorns. It's going to be a pretty rough ride, I think. I don't know, but I see all these breathless. And Microsoft says something today about like 60 % of white collar jobs will be gone three years. Rio (59:00.814) Yeah, but I see all these breathless, like Microsoft says something today about like 60 % of white collar jobs will be gone in three years. mean, this people have been saying this like, yeah, there are a lot of layoffs. I don't think the AI is causing layoffs. Yeah, I think the layoffs are caused because to your point, these companies are bloated and have too many people anyway. Karsten Weide (59:08.54) That's not gonna... mean the the layouts that you see right now are not because of AI. AI is the excuse to right-size bloated enterprises. Brett House (59:16.92) Yeah, it's more of a business model problem. Like there's certain things that just should be automated. Brett House (59:30.17) Yeah, yeah, yeah. And that's a business strategy and a business model perspective on the FTE side with agencies because that model just doesn't scale and it doesn't compete with the walled gardens and the tech companies. Yeah. Rio (59:31.064) But e- Karsten Weide (59:35.192) But, but, Rio (59:41.102) It's not a model that will work in the future. When you think about like at some point in the future, there's no way this model makes any sense. But like you look at these companies, even tech companies, like 40 % of the people in some of these companies, I don't even know what a lot of them do. Like I remember we used to joke when I started consulting, we'd consult with like big banks and stuff. Like you go in like literally half the people, they could get rid of them and like they would have zero impact in the company. And I think with these tech companies too, a lot of them over hire. So I'm not surprised you've seen waiver. It sucks if one of people gets clipped, right? Karsten Weide (59:41.169) It's not a model that will work in the future. When you think about like, at some point in the future, there's no way this model should make sense. Look at these companies, even tech companies, 40 % of the people in some of these companies, I don't even know what a lot of them do. I remember I used to joke when I started consulting with big banks and stuff, you go and literally half the people, they could get rid of them and they would have zero impact on the company. And I think with these tech companies too, a lot of them over hire. So I'm not surprised you see waiver. It sucks if in field, it's clipped, right? But with layoffs, I probably getting closer to the point. if you want to make decisions, you better have correct information. so sticking your head in the sand is not going to help. Now, to your point of, oh, Microsoft said, all these knowledge workers are going to go away. Well, maybe. Rio (01:00:11.074) We have layoffs, I think, probably getting closer to the point. At some point, you cut too far, though. Karsten Weide (01:00:31.132) but certainly not based on models that are LLM based. I mean, these things cannot do the most primitive things, right? I mean, you try to accept all the company names out of this document and make an alphabetical list, comma, separated. Can't do it. mean, like very simple tasks they cannot do. And we are being told, intergalactic super intelligence is just around the corner. I mean, it's ridiculous. And then they show, oh, Brett House (01:00:38.799) Yeah. Rio (01:00:48.502) Screw it up. Karsten Weide (01:01:00.55) this chatbot solved a math problem that has been around for three centuries. And it's so irrelevant to anything. mean... Brett House (01:01:05.076) no, I saw that, no, it was it. No, that was a quantum computing example, right? From Google. I think it was the quantum computer that Google had. And it was a math problem that would take a thousand years to... Oh, is this different? Rio (01:01:13.422) No, no, no, no, no, think a couple of days ago, Brett. No, no, no, no, no, no, said, I think one of the of the one of the LLMs solved some math problem that I mean, they were saying how this was a breakthrough. This is the first time it actually had solved the new problem if it's had right. So yeah. Brett House (01:01:31.05) yeah yeah not not the quantum computing problem which is Karsten Weide (01:01:34.299) But AI is eventually going to get there. But AGI is never going to happen based on LLMs. Brett House (01:01:41.029) Yeah. Karsten Weide (01:01:44.317) But by the inherent design, they're just not capable to do that. It's possible that you put other pieces of AI software on top of them. That's a different story, but that's not happening so far. I mean, there's some of that. They're building shells around the models to like sieve out inappropriate responses and stuff like that. But that's not AGI. So AGI is still farther away, think. Elon Musk says it's going to happen next year. I don't think so. Brett House (01:01:56.282) Yeah. Karsten Weide (01:02:14.268) I mean, yeah. They are. are. By the way, I'm not saying that these models are not super useful. I so I'm an independent contractor, right? So this increased my productivity by a factor of at least 50%. It's very significant. yeah. So. Rio (01:02:14.734) think so either. they're getting models getting much better, right? But I think you're right. It probably needs something that's like not LL. Like LLMs would be part of it, I think, but. Brett House (01:02:16.26) Yeah, just- just listen. Yeah. Brett House (01:02:35.78) Yeah. And yeah, we're getting more automated by the day. Rio (01:02:37.026) We use it for everything here in the pod, everything. We couldn't do this podcast with, we couldn't do this podcast with the small crew we have, like with the volume of content, there's no way in a million years we'd get it with that AI. Karsten Weide (01:02:42.114) But maybe AI can have you should go on a chat button ask where your cough is coming from what you can do about it. That's right and then add for Robitussin that's Brett House (01:02:44.357) you Brett House (01:02:52.006) I think you're gonna have to end because I'm getting you across. Losing my voice here. So Rio, you do the ending. Rio (01:02:55.043) Yeah This is golfing fit. Rio (01:03:03.926) and it'll recommend some Robitussin or something like that. Brett House (01:03:10.15) Yeah, exactly. Exactly. A little bit better. Rio (01:03:10.21) Yeah, an all-time display ad. Karsten Weide (01:03:11.836) We'll little time to play out. So there. Well, if people want to get a hold of you, let's say people want to look into your research, chat with you about things, how do they get hold of you? Probably best on LinkedIn. Just look up my name and text me. It's always happy to chat. Brett House (01:03:18.04) There we are. Rio (01:03:18.446) All Karsten, well, if people want to get a hold of you, let's say people want to look into your research, chat with you about things, how do they get hold of you? Brett House (01:03:35.654) Well, Carson is great, great, haven't you? Rio (01:03:35.874) And you're usually at the industry conferences and you're based in San Francisco, right? Karsten Weide (01:03:36.157) You're usually at the industry conferences and you're based in San Francisco, right? Based in San Francisco. I'm at industry conference. Next one is advertising week. Then it's CES. Then it's possible. Then it's Cannes, which I missed this year, unfortunately. What can you You're going to Cannes? OK. I'm not going to be in Cannes. I planned on going, but I just can't pull it It's just too busy. Yeah, we're doing some big stuff in Cannes that you'll need Are you guys going? Brett House (01:03:40.582) Mm. Rio (01:03:52.792) You're going to be a Candace Brett House (01:03:52.998) You're going to miss counting this year. Got it. Yeah, well we're doing some big stuff at con that you'll the audience will hear more yes, we're doing a big activation For signal and noise the news is coming out very soon. It's probably gonna come out post the publication of this podcast Karsten Weide (01:04:06.972) Okay, so here's some practical advice. Put your wallet and your phone in your front pockets. Lots of pickpocketing. Last time I was in Cannes, well that's also bad, but Cannes is even worse. So I, last time I went to Cannes I was like... Brett House (01:04:18.914) Yes, well, I live in New York City. You live in San Francisco. You live in San Francisco. God's worse than San Francisco in the Tenderloin District. Rio (01:04:32.984) minus the zombies, Karsten Weide (01:04:35.068) Last time was in Cannes, somebody pickpocketed my airline tickets. And I go to the police and they're like, really? It was a physical ticket at the time. Brett House (01:04:39.993) Wow. Brett House (01:04:44.504) Wait, Carson, you're aging yourself. How many years of experience did you say? Airline tickets or did they pickpocket your mobile phone? Who carries physical airline tickets anymore? Rio (01:04:59.118) Did they get on your flight? I'm kidding. Karsten Weide (01:05:00.976) No, no, it was no problem. I got to replace. Exactly, Not just my seat on the plane, but on my couch, yeah, I know. Brett House (01:05:04.198) They stole your complete identity. Rio (01:05:06.86) You went home and some dude is on your couch. Brett House (01:05:13.336) Watch for pickpockets. Yeah, yeah, definitely. There's definitely like a group of people that, you know, that's kind of, they're like the parasites that feed on those festivals and you can kind of see them when you're going home at night. You're like looking around the corner and. Karsten Weide (01:05:19.26) It's kind of... doing its favorite thing. Rio (01:05:26.158) No, but there's a lot of like robbery of different sorts. I remember last time I was there, like I met someone from Amazon and I think it was that, I think it was the Carlton, right? And we, went in, I was, you know, I was just ordered a couple of Negronis and like, I think it was 70 euros for two Negronis. So I'm like, this is crazy. so not a cheap place, but most people, your companies are paying. So, so it'll be fun. Karsten Weide (01:05:26.972) There's a lot of robbery of different sorts. I remember the last time I was there, I met someone from Amazon. I think it was at Carlton. I went in and ordered a couple of macaronis. It was 70 euros. Some were like, this is crazy. Not a cheap place, but most people in companies are paying. Just because the show is in the video, we bought like... Brett House (01:05:49.966) Yeah. yeah, we had a we had just just because the show is independent. We did have a we bought like a three hundred dollar bottle of Rose when I was at Nielsen. So so my my colleague did he came over and he put it at the table and and and no and a prostitute who was sitting at the table who we didn't know was a prostitute at this point grabbed the bottle, poured it into her glass until it overflowed all over the table. And then we like looked over at her and her eyes were just beard dog. Karsten Weide (01:05:56.924) I So like, how is it? came over and he it in the table. It's a magnum, at least? And a prostitute, who was sitting at the table, who didn't know what the prostitute was doing, grabbed the bottle, poured it into her glass until it overflowed all over the table, and then they like, over her and her eyes were just beer goggles. And we're like, yeah, and then the security sort of came over and... Rio (01:06:02.136) Was it a magnum at least? Brett House (01:06:18.982) like spinning and we're like this yeah and then the security sort of came over and yeah so you do have to be careful with with those type of folk that kind of you know Rio (01:06:25.4) Was this at the gutter bar, Brett, like three in the morning? Karsten Weide (01:06:35.276) Either way, France or Southern France, is lovely at this time of the year. yeah, good luck. Brett House (01:06:42.008) And we'll see you next time. was good chat and thanks for joining us. And for everybody. Karsten Weide (01:06:44.605) No, it was fun. I enjoyed it a lot and and safe travels, okay All right, thank you Rio (01:06:44.792) This is fun. Good. Good conversation. Brett House (01:06:50.309) Yeah, and for everybody that joined, you can visit us at www.signalnoise.ai. Tons of editorial content, video content as well as podcasts, obviously. You can also find us on YouTube, Spotify, and Apple Podcasts, and we will see you next time. Thank you. Rio (01:07:07.416) Thank you.
Comments