Prog AI Live: AI's Slippery Slop

Summary

Podcast Summary

The Big Story – Prog AI Live: AI's Slippery Slop
Host: AdExchanger Editorial Team, Moderator Alison Schiff (Managing Editor)
Date: May 21, 2026

---

Episode Overview

This special live-recorded roundtable from the Prog AI conference dives into urgent and pressing issues at the intersection of AI, programmatic advertising, media, and digital marketing. The AdExchanger editorial team discusses the momentous Publicis–Liveramp acquisition, the industry’s muddled relationship with AI (“slop” and premium content), evolving definitions of brand safety, and the unintended societal and environmental consequences of AI's rapid adoption.

---

Key Discussion Points & Insights

1. The Publicis-Liveramp Acquisition: End of Neutrality?

[03:00–09:15]

• Deal Details: Publicis acquires identity vendor Liveramp for $2.2 billion, ending its “Switzerland” status as an industry-neutral provider.

• Impact on Rival Agencies:

  • Others (e.g., Omnicom) swiftly moving to exit Liveramp’s ecosystem due to conflicts of interest with Publicis.
  • Raises questions about interoperability and value—does Liveramp lose market power as a non-neutral intermediary?
  • James Hercher: “It can't really be recreated. The power is the network...The RampID...is in a lot of ways the default id.” [06:11]

• Industry Ramifications:

  • Potential for competition to “find their own daddy holdco.”
  • Uncertainty about future competitors and the potential for a “new Switzerland.”

• Is this about AI?

  • Alison Schiff: “Arthur Sidoon, the CEO of Publicis, made a comment [...] that this is about AI. And that struck me as a little bit odd because I don’t see that.” [07:10]
  • James Hercher: “Everything is AI just, like, stuck on it.” [07:32]
  • Insight: Acquisition is less about advanced AI and more about owning the “sinews of programmatic,” i.e., identity middleware and data interoperability.

2. “Content” in the AI Era: Redefining Premium

[09:19–16:26]

• Debating "Content":
  • Discomfort with the catch-all term; what counts as “premium” is increasingly up for debate.
  • Alyssa Boyle: “Apparently premium is engagement. I don’t know if I agree with that.” [11:11]
• Rise of AI Slop:
  • Explosion of low-quality, AI-generated content ("slop") that, nevertheless, garners engagement.
  • Examples include YouTube compilations, poorly narrated "Karen" videos, and entirely AI-fabricated viral phenomena.
  • Alyssa Boyle: “AI slop, for example—it does garner engagement...it’s probably making me dumber.” [12:07]
  • Victoria McNally: “So people being mad at stuff counts as engagement.” [12:55]
• Fragmentation of Audiences & “Lean-in” vs “Lean-back”:
  • “Premium” now varies by audience, not universal standards.
  • Victoria McNally: “No one’s watching the same things anymore. There isn't as much of a monoculture...You have to reach them where they are.” [13:38]
• Industry Definitions Diverge:
  • Ad tech firms define premium by ad experience, not legacy media reputation.
  • James Hercher: “You could have a well known news brand that has a terrible ad experience…and they wouldn’t call premium.” [15:01]

3. Trust, Engagement, and the AI Content Backlash

[16:26–19:04]

• Consumer Distrust vs Tech Industry Optimism:
  • Alyssa Boyle: “There is a disparity in consumer trust toward AI content vs. AI chatbots...” [16:42]
  • Stats: Gen Z/millennials with negative feelings toward AI rose from 18% (2024) to 30%.
  • Personal anecdotes: Older generations embracing ChatGPT uncritically.
  • Alison Schiff: “They are asking health related questions and they seem very comfortable with it.” [18:05]
• AI’s Marketing Problem:
  • Over-hyping, mystical language ("wizard" and "oracle") exacerbates public cynicism and misunderstanding.
  • James Hercher: “We should just be treating [AI] as...a really good version of Clippy.” [19:21]
• Anthropomorphizing AI:
  • Overuse of terms like “hallucinations” perpetuate misunderstandings of AI’s real capabilities and limitations.

4. The Coal, Horses, and Locomotives Analogy: AI’s Paradoxical Impact

[21:54–26:43]

• Jevons Paradox Applied to AI:
  • As AI tools become more efficient and accessible, their total societal consumption increases—not less.
  • Joanna Gerber: “As it gets more efficient, it also gets cheaper. And because it’s cheaper, people are more inclined to use it...that’s what’s happening with AI.” [22:12]
• Environmental & Societal Cost:
  • Like coal, AI’s growth brings mounting environmental impacts (water use, electricity, localized pollution).
  • Victoria McNally: “There is that environmental cost...the data centers that are needed to power these take up a lot of space and end up impacting the environment around them.” [24:04]
  • Alyssa Boyle: “Sustainability stopped being a marketing priority...as AI got more hyped up and adopted, we started to hear less about sustainability.” [26:14]
• Industry’s Opportunity Lost:
  • Companies could have proactively addressed social trade-offs—offering benefits back to impacted communities—rather than scrambling with marketing and lobbying after the fact.

5. Brand Safety, Engagement, and the Future of AI Content

[29:40–32:50]

• AI Slop vs Brand Safety:

  • The tension between “engagement at all costs” and maintaining brand-safe environments intensifies.
  • Alison Schiff: “If engagement or AI Slop is the new premium content, then that seems to be at variance with, quote, unquote, brand safe...are they diverging and should brands rethink brand safe?” [29:40]
  • DoubleVerify’s Mark Zagorski (quoted): AI slop represents a huge opportunity for verification tools to “get rid of all the AI slop out there.”

• The Move from “Brand Safe” to “Brand Suitable”:

  • Alyssa Boyle: “I think brand suitable means...that should encompass consumer opinions a little bit more. [...] People do not want to see AI content...if all engagement is engagement and that’s what matters, over time, consumer trust is going to dissipate.” [30:51]

• Disintermediation 2.0:

  • Alyssa Boyle: “Disintermediation” is now about AI potentially replacing not just people but entire ad tech platforms.

---

Notable Quotes & Memorable Moments

• “The sinews of programmatic are often Liveramp.”
  — James Hercher [09:10]
• “All news is not good news. But AI can, like AI slop...does garner engagement.”
  — Alyssa Boyle [12:07]
• “No one’s definition of slop is the same.”
  — Joanna Gerber [13:32]
• “There is a disparity in consumer trust toward AI content versus AI chatbots...consumers don’t like it.”
  — Alyssa Boyle [16:42]
• “Clippy...I have nostalgia for Clippy now.”
  — Alison Schiff [20:00]
• “As something becomes more efficient, we use up more of it as a society. That’s what’s happening with AI.”
  — Joanna Gerber [22:12]
• “Sustainability stopped being a marketing priority...as AI got more hyped up.”
  — Alyssa Boyle [26:14]
• “LLMs are more narcissistic than the most narcissistic man you’ll ever meet in a bar.”
  — Albert Thompson, as quoted by Alison Schiff [27:11]
• “Human in the lead.”
  — Takeaway from JPMorgan Chase session, cited by Joanna Gerber [28:53]

---

Timestamps for Significant Segments

| Timestamp | Topic / Quote | |----------------|--------------------------| | 03:00–09:15 | Publicis–Liveramp acquisition, neutrality, market effects | | 11:07–16:21 | AI “slop” and changing definitions of premium content | | 16:42–19:04 | Generational attitudes, trust in AI, and content backlash | | 21:54–26:43 | Efficiency paradox, coal/locomotive analogies, sustainability | | 29:40–32:50 | Brand safety, verification, disintermediation, industry future |

---

Takeaways & Final Thoughts

• Shift in “Premium”: Definitions are in flux; engagement now sometimes trumps traditional content quality, but this fosters “AI slop.”
• Brand Safety Challenges: Tools aiming to block AI-generated “slop” face an arms race against low-value but high-engagement content, raising the bar for “brand suitability.”
• Sustainability and Social Impact: As AI gets cheaper, use and environmental stress soar; the industry has yet to fully reckon with these externalities.
• Cultural Pushback: While tech insiders press forward, the broader public (especially the young) are increasingly skeptical or even hostile toward AI-generated content.
• Human Oversight: The industry mantra is shifting from “human in the loop” to “human in the lead”—AI is a tool, not the driver.

---

Closing Sentiment

The panel underscores a period of major flux in the digital advertising ecosystem, as AI-driven automation collides with questions of trust, environmental impact, and the basic nature of content. The debate continues: is programmatic AI advancement a slippery slope, or a necessary next step—assuming humans remain "in the lead"?

Transcript

[A]

Welcome to the Big Story, a roundtable featuring members of the Ad Exchanger editorial team. Every week, we bring you an in depth discussion of key developments in digital marketing and media. Today's episode is sponsored by Verve.

[B]

Verve captures over a billion daily search

[A]

AI chat and zero party signals, giving brands and publishers a real time understanding of intent. All right, this is the. This is the last session of the day, guys. But it's not just a session. It's a live recording of our Big Story podcast. So let's all take our seats and do this thing.

[C]

Scream obscenities.

[A]

James, did you say scream obscenities? Okay.

[D]

Oh, is Alyssa not.

[A]

No, no. Come on, come on, join us.

[E]

What?

[A]

Alyssa, you want to be on the podcast?

[B]

I don't know.

[A]

Do I have to write?

[E]

You're kidding.

[D]

I'm kidding.

[A]

Come on up. All right, so this is a full pan, a full panel. Really?

[D]

Full panel.

[A]

Chairs left.

[B]

When we're not on the plane, you don't have to avoid the middle scene. I was told to sit stage left.

[A]

All right, guys, enough banter. Got to do the intro. So, I'm Alison Schiff, the managing editor of Ad Exchanger, and I'll introduce my podcast guests in a moment. But it's been a packed couple of days at Prog AI, so thanks for joining us. And we're going to distill a bunch of the takeaways from the past couple of days with these most excellent colleagues of mine. But first, I think we have to talk about some huge news that broke over the weekend. James Hercher, our senior editor, covered it. Publicis bought. I was going to say Publicis bought Switzerland.

[E]

That would have been really.

[A]

But that's because Liveramp loved to refer to itself as Switzerland. Switzerland. A neutral party. And they are arguably no longer a neutral party now that they will be part of Publicis. Publicis spent $2.2 billion to buy them. So I'm going to put a pin in that for one second. That's the first thing that we'll tackle. But before I do, just want to go down the line, we have Joanna Gerber, our associate editor.

[E]

Hi.

[A]

James Hercher, senior editor. Alyssa Boyle, senior editor, and Victoria McNally, associate editor. And we're not used to taking questions. We're usually the ones asking questions, but what the hell? Use slido. If you have a question for us, pop it in there and we'll try and get to it.

[E]

But, yeah, take this opportunity because it won't happen again.

[A]

It won't happen again. Okay, so, James, just give us a Quick rundown on the Publicis Liveramp deal. It was just massive news. It broke on a Sunday, and I wish it had broken on a weekday, but that's not how it happened. You jumped right on it. So tell us what happened.

[C]

Yeah, that was an interesting one. I had a call being lined up with Scott Howe because Liveramp was supposed to report earnings, like, tomorrow or something like that, so that got canceled. Yeah, an interesting side, but. But, yeah, a very. A very big, like, for the price tag. I think it's gotten a lot of ink and it is one that everyone wants to share their opinion about because there are, you know, this was sort of what I talked about. My story is like, there's so many question marks with this deal. Like, okay, how is this going to work? Like, how is, you know, the bouncing ball going to play out here? And, I mean, Liveramp has. It's almost been. It's been rumored Publis acquiring Live Ramp for a while. Like, yeah, that rumor mill has been kicking around for, yeah, a year or more. But I don't know if anyone really ever thought, like, okay, like, Liveramp, would it make sense? Like, could Liveramp be acquired like this?

[A]

Can it remain neutral? I mean, that's the main question. And does it lose its value if it's part of a holding company?

[C]

I mean, we've already definitely seen some announcements, and I talked to Hold Go folks, and they're planning to get out of Live Ramp. Omnicom took this very gleeful, like, we were already getting way out of Omnicom, out of Liveramp. And, yeah, we're just going to speed that up. So I think it definitely does. From that perspective, it hurts. But on the other hand, this is one of the questions. Are they supposed to be continuing as this neutral, standalone business, or do they become, like, a lot of the publicist tech acquisitions? The point is, this becomes a part of the Publicis offering for Publicis client base, which is much larger than Liveramp's client base. So it's kind of like, okay, who cares? Publis has bought you. We have lots of business for you. Is the best.

[A]

They have enough data to feed into Liveramp because Liveramp is only as good as the data that that's in it. But, Alyssa, I think you had a point to make. You were waving at me.

[B]

It's the zoom equivalent of the hand emoji. So I can definitely. My first thought was, of course, other hold codes are not going to use Liveramp because they wouldn't want any possibility of their data touching publicises that Sounds weird no matter how much Liveramp grammatically

[A]

correct and sounds very weird anyway.

[B]

But what does this mean for the other alt IDs out there? Because there's too many of them. So I can imagine a lot of them are like, woo, move on over, it is my turn. What's the perspective on the other halt IDs?

[D]

Yeah, similarly, like, what do Liveramp's competitors do in this situation? Do they all try to find their own daddy hole co as well and they just all become divided up? Or is there a new Switzerland that's gonna be there?

[C]

I mean, that's how it rolls out. There certainly are smaller vendors, but the thing about Liveramp is that it can't really be recreated. Cause the power is the network. Like you can't, you know, it isn't even really the software that's like, are you really gonna go build all these connections? And the Ramp ID is in a lot of ways the default id. When people talk about interoperability or something being deterministic, even other identity graphs, I think people probably don't even realize how much in the background there is. Often there are other identity graphs that use Liveramp. These competitors that do onboarding and matching, they discreetly in the background, love to use their others to extend their reach, make their numbers look better. It's like, okay, what happens there? The UID program is in a way built on Ramp id.

[A]

It's quite incestuous. Right?

[E]

Yeah.

[A]

I want to bring it back to AI though, and then we'll move on from Publicis. But Arthur Sidoon, the CEO of Publicis, made a comment when they were announcing the deal that this is about AI. And that struck me as a little bit odd because I don't see that. But is this an AI deal at its heart?

[C]

I mean, I guess only in the sense of like, everything. Everything is AI just like stuck on it.

[A]

Now we're programmatic AI.

[B]

Now this deal's about. Insert buzzword here. You know, I don't know if this

[E]

is particularly worth mentioning, but there was also news last week that Arthur Seydun just got a 20% increase in his raise. Or, sorry, in his salary. Because, yeah, there was some evidence dug up that he was being underpaid.

[A]

So he was being underpaid.

[B]

Relative.

[E]

I mean, the word is relative. But I wonder how that fits into everything.

[B]

Compared to who?

[A]

I don't know. But I'd love a 20% raise.

[E]

I would love 20% of his salary.

[A]

Fair. All right.

[C]

So, I mean, I don't think Liveramp does like, I Don't think of Liveramp as, like, you know, a gentic. They did talk about, like, oh, like Liveramp is going to, like, build our agents. And it's like, I don't know if that's what Liveramp does. Everything is AI. But no, I think this is really about the core Liveramp business. It is about interoperability, which is like, sort of amusingly, Arthur Sudhun tripped over the word and he's like, I'm not going to say this again, because he couldn't pronounce it. Like, as a French person, he can't get his mouth around the word interoperable, which is just sort of funny. So he can't use this buzzword.

[D]

Trying to picture it a French accent

[E]

for French people in the industry just to reduce.

[C]

Well, no, that's how you end up with just agent being 24 times on the same call. But, yeah, I think it really is more about interoperability and owning this kind of just middleware layer that people don't even think of, like, don't appreciate how much of just like, the sinews of programmatic are often. Liveram.

[A]

Oh, that's beautiful. The sinews of programmatic.

[C]

Yeah. All right.

[A]

What a turn of phrase.

[B]

Visceral.

[C]

I should get in writing.

[D]

It's so French.

[B]

You should be a writer.

[A]

You should be a writer, James. That actually provides me with a nice segue. I want to talk about content and the definition, and we are content producers. Although I kind of hate the word content. It just feels a little bit too. Like we're making widgets or something.

[E]

Makes me feel a little bit discontent.

[A]

It makes me feel a little.

[C]

Content is just meant. There's ads. It's meant. It is there to. It is something that was produced to put ads in. In a sense, we'll just.

[B]

We'll refer.

[C]

Movie is not content, but a TV show is kind of.

[B]

Every time someone says content, I think content is king. And then I cry a little bit. And that's why my face looks like this. Like I just had a line or something.

[D]

I would push back on that, James, though, the idea of movies not being content because people put ads in movies, too. And it's all changed at TV upfronts. Last week, there were a lot of movie trailers because they're all promoting their streaming service and they all have ads in them and, like, they're going to go in the movies, too.

[E]

Yeah.

[B]

There's a nice little handful of companies that'll pitch you on the fact that they can do, you know, program. I mean, you can ask a million questions on is it one to one though. But point is, if it's programmatic, I mean, hey, it's in the theaters now.

[A]

Content 1 trailers are ads that I watch voluntarily. I'll fall down a YouTube rabbit hole happily.

[B]

And super bowl commercials, trailers.

[A]

But I want to talk about the definition of premium content in energentic worlds. And I always feel like the movie trailer guy when I say in an agentic world. But does the definition of premium change in an agentic world? You also need to live in reality because people are changing the way they find you. So you have to roll with the changes and produce the sort of content or whatever that LLMs like. And I don't know what that means for premium. Alyssa is jumping out of her seat. Yes.

[B]

Yes, I have something. Pick me. Okay, so, so Victoria moderated an amazing session on this yesterday and I have a lot of thoughts because it inspired a newsletter topic where I put in a bunch of my own thoughts. Woo. So here are those thoughts. So premium has been a debate for many years, ever since I joined Ad Exchanger because a couple months after I joined it became 2022 and then YouTube entered the upfronts and then it became this whole debate of oh God, what is premium? Does that mean user generated content is premium? And that has been a years long debate. But now I feel like AI is introducing more variables into that debate or at least giving more credence into this idea that apparently premium is engagement. I don't know if I agree with that. I wouldn't call it premium, but I would say that AI creates more user engagement that advertisers are interested in. For example, I mean, you know the phrase all news is good news? So I feel like this is the.

[E]

The phrase is no news is good news.

[B]

Oh, never mind.

[E]

I'm not good at analogies.

[B]

That's what I'm looking for. That's what I'm looking for. So in this sense it's like whole engagement world.

[A]

All news is not good news.

[B]

Not at all. But AI can like AI slop, for example. It does garner engagement. For example, I watch a lot of YouTube garbage. What do I mean by that? Karen Compilations and Public freakouts and Karen's trying to summit Mount Everest. Like weird, weird stuff. But years ago, this stuff used to be narrated by a human being. Now it's narrated by AI and it makes grammatical errors and it's probably making me dumber. And I shouldn't really say that where it's getting published, but hey, it's Happening, but people will engage with this in the comments, you know, and for example, here's another AI generated piece of content that I saw. It was drag queens chasing ice agents down the streets in New York City. Totally AI generated, not real. But people were engaging with it in the comments, like, whoa, this is. This is the only acceptable use of AI. Debate aside, it's garnering engagement. It's making people interact with it.

[D]

So people being mad at stuff counts as engagement.

[B]

Exactly.

[D]

A lot of tiktokers have found out and they get themselves to one of those airplane sets in LA and pretend to film freakouts and everybody comments on them.

[E]

I mean, that's kind of the premise of rage bait, right?

[D]

Exactly.

[E]

That people dislike to garner engagement. And nothing baits my rage more than seeing people engage with rage bait. You're just feeding the algorithm that it's a cycle. Yeah, it makes me very angry.

[A]

It's also like an eye of beholder, eye of the beholder kind of thing going on.

[B]

Right.

[A]

I mean, one person's AI slop, and I'm not an advocate of AI slop, is another person's lean back and watch this garbage thing and turn my mind.

[E]

Well, I think the thing is, no one's definition of slop is the same. Like I. Victoria, I mean, yeah, that's true.

[D]

You can. You find people who just call anything that even looks a little uncanny slop. And then you see people outside of AI sort of calling things slop just because they don't like it. Like, there was a thing in video games a couple months ago where people were calling certain co op games friend slop. And it's like, because you have friends, but to sort of go back to the idea of premium content. I forget who it was on the panel yesterday. I think it might have been Sheila Marman who made the point that fragmentation also has a lot to do with it as well, because no one's watching the same things anymore. There isn't as much of a monoculture. And so something that might have been considered premium 10 years ago might not be anymore. Because if you have a specific audience that you're trying to hit and they're not watching that thing, then that's not premium to you. You have to reach them where they are.

[C]

I feel like there's a premium definition thing too, where, like, there's, you know, people have like their, like a colloquial definition of premium. I think of premium as, like, high production quality, like something that I watch on my tv, or like the types of news companies that get paid like tens of millions of dollars by like LLMs to license their whatever. Associated Press, New York Times, Wall Street Journal. But like the trade desk defines premium based on like how many ads are on the page and like the quality of those. You know, like I've gotten into this with like the, you know, Mike Sincero folks like Mike from Sincera and those

[A]

folks at Trade Desk, last name is not Sincera.

[C]

Yeah, but you know, it like they don't. It's almost like it's irrelevant. Like it could be some blog that nobody's ever heard of with a great ad experience and like a real user and you could have a well known news brand that has a terrible ad experience on the page and they wouldn't call premium. So you do get into these debates, like certainly in like the YouTube and TV world, it's like, how does an advertiser, how does the ad ecosystem in the sort of like programmatic world define premium? And that's, you know, usually in conflict with how people think of premium.

[D]

Right. People thinking differently than ad tech. What a concept.

[B]

Speaking of which, segue, because here's something very interesting that came up related to this topic. So I had, I talked to Deborah AHA Williamson from Sonata Insights. She delivered our opening keynote and she had some really interesting stats that she discussed. And the point here is there is a disparity in consumer trust toward AI content versus AI chatbots. And my whole point in that is that AI generated content is something that people just do not like. AI chatbots and other tools in efficiency and other buzzwords, fine, consumer trust is growing. But AI content, whether or not you want to call it slop, consumers don't like it.

[D]

They hate Sora. They're all right with ChatGPT, apparently.

[B]

Apparently.

[D]

That's interesting.

[B]

Here's the stat that I remember. Ready? So Gen z and millennials, 30% feel negatively toward AI. And apparently that number was only 18% in 2024. And I mentioned this when Victoria and I gave our opener this morning. But I kind of compare that to consumers are not becoming less trusting of AI content. They don't like it. They're just being more vocal about it because we're inundated with it. Similar to how Gen Z millennials are not getting gayer, they're just becoming more comfortable talking about it. I think that's a great analogy. I think the consumer sentiment is the same. It's just the volume of how much of it exists, how much we're talking about it.

[D]

Well, I do think it is worth pointing out that there is this huge discrepancy between how people in the tech sphere talk about AI and how every day people talk about AI. Because there's been this big trend in the last couple of weeks as all these different graduation sessions have been happening of university commencement speakers bringing up AI and immediately getting booed, no matter what they're actually saying about AI. And you know, nine times out of 10 they're saying, oh, it's the future. We're so excited. So like, you know, your mileage may vary there but like, you know, you got these 22 year old kids like very loudly proclaiming that they don't want this. And I think that is something that the industry kind of has to contend with, you know, because if everybody hates your thing, I mean obviously like you could say the same thing about advertising more broadly, but like if everybody hates your thing, that is going to affect how you're able to implement that thing.

[E]

I think that segs into something else. Alison, you wanted to bring up earlier on the topic of coal. I believe on the topic of coal.

[A]

I do want to put, let's put a pin in coal for one second.

[E]

You can come back to coal.

[A]

I just, I wanted to react to what Victoria was saying. There are a lot of people in my life, a lot of them are older people actually and they have started to rely quite heavily on ChatGPT and they're replacing the word like Google or search with let me ask ChatGPT. And there's a very high level of trust. It's not necessarily a great thing because they're not interrogating the results, they're asking health related questions and they seem very comfortable with it. So there is this backlash and we're talking about young people too entering the workforce and being quite nervous about finding a job. And then you also have people that are falling right into ChatGPT, like a warm embrace and they're using it for more and more things.

[C]

I think AI has also just has a really poor marketing problem. Like part of this whole issue is is the AI companies doing a really poor job around the show. I've, I've heard people talk about like, oh, like in San Francisco people are hiring like wizards. You're hearing like wizard as a title and Oracle is like not fair.

[D]

That's Oracle.

[C]

Oracle is like a, not the software company but is like a term in machine learning. And AI is like the data set that you consult to get an answer back for like a couple people use that term. It's just as always, like there's all this leaning into, you know, like, this is software. Like, as opposed to being a person with a name and like sometimes even an avatar and some sort of like vague ethnic background. It's like we should just be treating this as. This is more like a really good version of Clippy than us, like the next evolutionary step forward.

[B]

I miss Clippy, but it's like we.

[D]

Can you imagine missing Clippy though? Remember when we all used to hate Clipp?

[A]

I have nostalgia for Clippy now.

[D]

That's wild.

[B]

Me and Joanna are like, we're Gen Z. What's going on?

[C]

Like, the desktop world was just so, like there was all these efforts, like the word desktop, which maybe you don't know, but like, was because, like to make a computer not scary to someone. It was just saying, this is your desktop, here's your folders. The trash button, like the trash icon in your computer used to be called eject. And nobody used it because obviously no normal person touches the word eject. And so they just was like, no, this is your trash. Right? And it's just as like all these things are like metaphors. And by being leaning into all this, like mysticism and spirituality and like metaphysical stuff, like, they just make technology that instead of just being like, this is totally normal. This is like the work that you've been doing like a million years. Like, just think of this as just being like so normal. Like they just know, like, this is like weird and mystical and like evolutionary. And it's like, oh, why can't you just be normal? And there wouldn't. There would just be so much less hate, so much more adoption.

[D]

All right, my hot take. We should be hiring fewer wizards and more clerics. We need more wisdom based classes in these fields.

[B]

I think we need a bard or two. I'm always the bard in Dungeons and Dragons.

[D]

That's charisma based. That works.

[C]

Yeah.

[B]

Also, is this a good segue to say that we really need to stop saying hallucinations at work? It just feels very strange.

[D]

I don't know.

[B]

I think I've heard. I've heard. Although other people say that's the buzzword they would like to die. I've also heard agendic people are tired of that word because people just use it to like.

[E]

I don't think we can be intelligentic yet, though.

[A]

I mean, that seems premature. And I don't know that hallucination is really a buzzword. It's just overused.

[C]

We need something there, though.

[D]

It also has the connotation of something with a conscious mind. Which tends to be going back to dd my bugbear about talking about AI because we tend to personalize it and anthropomorphize it and ascribe feelings to it that it doesn't have. And that tends to be my issue personally.

[A]

Well, now I want to talk about coal.

[E]

I was waiting for this part.

[A]

So my question, and this is for you, Joanna, is why did coal and why did locomotives and why did horses come up so often during the past couple of days at this AI focused

[C]

conference about steam engines?

[E]

I'm so glad you asked. So, yeah, not one, but two different people brought up Javon's paradox, which was created in response to energy efficiency. Basically it was a paradox saying, oh, energy gets, you know, energy efficiency is developed. You'd expect that to mean people will end up using less energy because it's more efficient. But actually as it gets more efficient, it also gets cheaper. And because it's cheaper, people are more inclined to use it. So as something becomes more efficient, it actually becomes. We use up more of it as a society. And everyone was then suddenly saying that, like that's what's happening with AI. As it gets cheaper to use, we're actually going to be using more AI because it'll be so easily accessible, which

[C]

is because I don't think it's going to end up getting cheaper. I think we have a big price raise in our.

[A]

People have also been saying that more expensive.

[E]

Even the giant companies like Claude and ChatGPT, it's so expensive that people will have to be using basically like insurmountable quantities of tokens in order for these companies to still be afloat in 10 years. They're also extremely wealthy companies, so I feel like that could really go either way.

[C]

And they sell tokens.

[B]

So Cloud's going to have ads then.

[E]

I mean, Chachi Pinky has ads. As Sam Altman said, it was a

[B]

last such a stance against heads. Everyone saw that super bowl commercial with anthropic versus OpenAI.

[E]

Victoria also made a really good point with the coal analogy, if you want to take that away.

[D]

Yeah, well, because, you know, obviously, yeah, you have more access to the thing, it becomes cheaper, you start using it more. But that's just the financial cost. And there is also, in the coal metaphor, a huge environmental cost. And then ultimately it does start to become more expensive because you're losing that resource. You're, you know, damaging. I think the ozone layer. Don't quote me on that. I don't remember.

[E]

Fixed the ozone layer and now we're causing other fossil fuels.

[D]

Listen, it's the end of a very long day. But, you know, like, there is that environmental cost that is sort of attributed to coal. And I think AI that is one of the reasons that people get booed at commencement speeches because the general perception of AI is that it uses a lot of water. And the data centers that are needed to power these take up a lot of space and end up impacting the environment around them. And I think.

[E]

And also raising the electric bills of those.

[D]

Raising the electric bills of people around them. And there are some localities, municipalities that are sort of trying to force the AI companies to pay for that or. Which makes it more expensive rather than less. So, yeah, I think that's something that, you know, it feels crunchy and kind of funky to talk about in a, in an industry setting, especially because I'm the one doing it and I don't actually know what I'm talking about with regard to the environmental impact. Clearly I just said fossil fuels.

[C]

It's part of my, like, bet, like, why I'm angry at their marketing too, is, you know, I feel like these, like, they should have just like, leaned in instead of, you know, now they're dealing with this huge pushback to data centers. And it's almost like this existential issue for the AI industry. And, like, you should have just sort of foreseen the problem. And, you know, if you wanted to, like, build data centers in Wyoming, you should have said, like, you should have made some commitment that, like, this is going to be cheaper for your electric bill. Like, you are going to get, like, we are going to pay you back for this. Just like, if you live in Alaska, like, you get paid for, like, oil, you know, and it's like you get like a fat check or, you know, it should have just been like, literally like, okay, like, we're going to do this in your state. We are going to make it worth your while for, like, citizens in your state. We are going to commit to lowering the electric bill. And it would have been a tenth of what they're going to end up spending in, like, marketing, lobbying, just dealing with all of these issues that they've made for themselves.

[D]

But to hear some people talk about living near these data centers, like, it does have an impact. They do tend to get built near more marginalized communities and affect the, the health outcomes of people in those areas. It's like living next to, like, a fracking system or like a mine, you know, like the level of output and the level of power and all of that. That it takes to create these institutions can hurt people.

[B]

So do we think it's a coincidence that all of a sudden sustainability stopped being a marketing priority? At least that's what I don't feel like I'm seeing it discussed as much. So I feel like as AI got more and more and more hyped up and adopted, we started to hear less about sustainability.

[D]

And I just feel like, yeah, there's less greenwashing.

[A]

There was a certain sustainability startup that suddenly became an AI company.

[E]

So there was also a footwear company that suddenly became an AI platform. Everybody and their mother's got to be an AI company now.

[A]

So I think we're going to wrap it up.

[E]

But slidos.

[A]

I don't know if we have any questions, but feel free to pepper us if you desire. I just want to go down the line. Give me your one takeaway. Just one line that struck you. I'll start Albert Thompson from Walton isaacson saying that LLMs are more narcissistic than the most narcissistic man that you'll ever meet in a bar.

[E]

Well, okay, I actually kind of want to follow up on that. I don't know if this is my one takeaway, but I thought that was a really interesting point to make because as a mid-20s person who's been in the New York dating scene, if I meet a man who's that narcissistic, I'm not going to engage with him. I'm going to walk away. So to me, that almost felt like a call to walk away from AI

[D]

altogether rather than try to wrangle it.

[A]

Yeah, that's your hot take.

[E]

Okay, that's my hot take. It's not my single takeaway from this event.

[C]

To clarify, I feel like if there was any sort of consistent things that I heard that maybe I. I did know that I sort of felt like I would come out of this with more of a sense of, like, practical, really, like digestible, like, first step, someone could take like, you know, this sort of idea of like, oh, AI, get it? As an undertaking. Just this very intimidating thing, like, I gotta go, you know, learn about this. Are we gonna, like, how are we gonna adopt this as a company? As opposed to sort of thinking like, okay, I could very easily go sort of set up an instance of something and like, test and see how this works and take first steps. And also I think a very consistent through line has been this real concern. Like, okay, actually, sometimes this can end up just making the wrong decision really quickly and scale up. We had Google and Horizon just on here before, and they mentioned that sometimes what AI can do is just it has the wrong incentive, the wrong data point, and they can just go and like, oh, run with that. And you could end up with just a terrible outcome that scales out and tells you it was great.

[E]

Can I take a more optimistic frame?

[A]

Please do, Joanna. I was going to say that during

[E]

the session with JPMorgan Chase, the phrase not just human in the loop, but human in the lead came up. And I think that was something else that I noticed as a big through line, keeping kind of heart and humans at the forefront of everything and using AI as a tool, but not like as, you know, a sole controller. And I thought that was a really nice way of framing that.

[A]

I actually think that's a wonderful way to end. So I'm going to not allow Alyssa or Victoria to give takeaways because we're going to leave off now. Alyssa looks devastated. The podcast is over. The event is over, and thanks to everyone who stayed. I saw someone out there, had a beer and. Oh, there's a question. Oh, my God. All right, more Encore.

[B]

Encore.

[D]

Now's our chance, Alyssa.

[A]

If engagement or AI Slop is the new premium content, then that seems to be at variance with, quote, unquote, brand safe. Are they diverging and should brands rethink brand safe? It's an excellent question. I think it was mark Zagorski from DoubleVerify on the most recent earnings call. He said something along the lines of how AI slop is one of DoubleVerify's biggest opportunities because they can go and try and get rid of all the AI slop out there and they're developing tools to do it.

[D]

I don't envy that challenge.

[A]

And I think then there are companies, too, that take the opposite stance, right? And really lean into the idea that engagement is engagement and AI slot for one person is a valuable medium for somebody else. So, Alyssa, why don't you react to that and also share your takeaway? I feel bad for taking away your opportunity.

[B]

Oh, it's okay. I shouldn't have looked at you like that. I wear my heart on my sleeve

[A]

like a puppy dog.

[B]

Okay. So I would say, yeah, I guess when it comes to brand safe, I honestly think not to. Not to have a cop out response, but I think brand suitable is where I'm thinking about more because I think brand suitable means I think that should encompass consumer opinions a little bit more. As far as, like, really, we got to acknowledge that people do not want to see AI content. They shun it. They are totally you know, not into it the second they find out something is AI. So I kind of feel like if, if we're going to care about that as an industry, then, yes, I think we should reconsider that because eventually, you know, if all engagement is engagement and that's what matters, over time, consumer trust is going to dissipate and consumers are going to shop somewhere else if you keep, you know, showing them AI ads. So I think, you know, I'm going all over the place. But in the short term, yeah, engagement's cool. And if you, if you need to prove performance to the C suite like everybody else, okay. But over time, you know, you need to actually engage consumers and make them want to buy from you over time. Which is also why, you know, at the upfronts, reach came up again. But isn't that interesting? You know, it's not all about just a race to the bottom. So that's important. My other takeaway is about the word disintermediation. I think that came up a lot, too. It used to be used for programmatic means like DSPs and SSPs replacing each other. Now it's, you know, disintermediation is coming up as, oh, is AI going to disintermediate and disintermediate humans?

[A]

Well, publishers, commerce companies, I don't know they're going to disintermediate humans necessarily, but that's exactly.

[B]

I agree with that. And I think that's why I think we're getting more comfortable as an industry, like pontificating on facts that are kind of obvious like, oh, of course AI is not going to replace humans or disintermediate them. But I feel like it's almost a cop out from actually addressing what disintermediation really means in the sense, which is, is AI going to disintermediate programmatic platforms? Because that's possible. But I don't think we're talking about that.

[A]

These humans in the loop are gonna leave the stage.

[E]

Humans in the lead.

[A]

These humans in the lead. So let's give ourselves a round of applause because why not? And I can't really see who's left in the audience, but clap for us, please. We're tapped out. And that's a wrap.

[E]

Thanks for sticking with us till the very end.

[A]

That's a wrap. All right. It's a wrap. Yes. Today's episode was sponsored by Verve. Find out [e-mail] that's V E R-V E dot com.