Eli Pariser on the end of the feed and how AI will change your morning routine

0:10Welcome to another episode of the Mixed Signals podcast from us here at Semaphore Media, where we are talking to all of the most important and interesting people shaping our new media age. I'm Max Tani. I'm the media editor here at Semaphore. And with me, as always, is our editor-in-chief, Semaphore's editor-in-chief, that is, Ben Smith. Ben, do you use AI in your media consumption? Do you ever have it summarize articles for you? Are you having it generate some personalized podcasts to give you the news? Has AI infected your media consumption yet? Probably the way that I use it most, and I don't know if we sometimes use it this way at Semaphore, is if I read something that I'm like, and there's an argument, and I'm sort of shaky on it, I'll say, can you find me something that disagrees with this?

0:59Can you like test, find something else in some language that will kind of like challenge this idea? Well, you did build, we actually internally built a tool that would basically, what, steel man some arguments essentially. It would feed some headline or some story into this AI tool and it would respond to us as, for example, and these are real examples. Sorry if I'm revealing some trade secrets here. On the left, Mehdi Hassan. And on the right, Ben Shapiro. It would insult us in the styles of either Mehdi or— Steelman is a strong word. But I did spend a very proud Sunday vibe-coding with, I think, some combination of all of Gemini and chat, this app that predicts mean tweets about stories.

1:41It didn't get wide adoption inside Semaphore, although it is in our leg in the sidebar of Google Docs. But I was very proud of it. I think it's because we don't really need it. If we put something out onto the internet, like essentially— We know what J.D. Vance is going to say about it if he has a shot. Exactly. Well, the reason why I'm asking this is because our guest on the show today is somebody who envisions a world where much of our media consumption is going to be channeled through artificial intelligence. We're talking to Eli Pariser. He is the co-director of New Public. It's a media and technology organization.

2:13He was also the co-founder of Upworthy, and he's the author of a remarkable new report called After the Feed. Ben, you were actually interviewed. You're cited as part of this report. Tell the audience why I found it to be so interesting and so slightly disturbing. Yeah, I mean, I'd start by saying Eli is a really consequential figure in sort of the history of the internet. Right after 9-11, he ran Move On, the anti-war organization that was in some ways a huge email list. It was like the first big mass digital politics in America, I think among the first in the world. And then wrote this book coining the term the filter bubble to mean the kind of political bubble that people get into became very commonly used.

2:52Then started this company called Upworthy, which in some ways tried to take advantage of those bubbles in his view for good. And so has been sort of on this ride of what is the internet like forever. And in this new report, I think he's sort of saying, kind of calling the end of an era, the end of this huge mass social era, and thinking really hard about how the very intense personalization possible with AI is going to totally change the way we all consume news, consume entertainment, connect to other people. And he's running a nonprofit. I think they're trying to think about what is the most positive outcome here.

3:28How could you go from this basically insane, wide-open moment that we're in into something that feels more human and more connected? And the report kind of takes you on that journey. Yeah, one of the really interesting things that we're going to hear Eli talk about, and this is how the report is structured, is just essentially how our news media consumption, if you're a news media consumer, is going to change really, really radically over the next few years from a system where we all kind of open up our phones and maybe some people start with Instagram or some people start with TikTok. the more responsible of us might wake up and read the New York Times or the Wall Street Journal app.

4:07And instead of doing something like that, we're going to open up essentially our agents who are going to summarize and tell us essentially what we need to know, essentially mediating the experience between the information that's coming into our phones and what we're seeing. I thought it was a really remarkable and interesting report. And I'm curious to hear more from Eli about what that's exactly going to mean for how we get information and the types of information that's reaching us. And for whether we will remain employed, of course. Of course, right. Exactly. And if that's going to mean the destruction of the media business, which includes Semaphore.

4:42Well, we'll get to our conversation with Eli Pariser, but we have to take a short break. So we'll be right back after this.

4:58It's called Frontier CMO, hosted by our friend Josh Spanier, Google's VP of Marketing. It's about the forces driving innovation and reshaping the marketing landscape today. Whether it's AI's impact on the industry or creator's role in culture, Josh is getting into the weeds with the people who are actually shaping what's next. These are your notes from the frontier. You'll hear from CMOs, agency legends, brand visionaries, and brilliant thinkers. Search for Frontier CMO from Think with Google wherever you get your podcasts or watch on YouTube.

5:34Eli, it's good to see you again. Thanks so much for joining us. Thanks for having me on. And you and I last talked when you were rolling out this deck, which is, I think, like the new way to release an idea. You don't need a book. You just have a deck. We'll see. I guess I wonder if you could just start in plain English. Paint a bit of a picture of what you think news and information consumption look like in an AI future. I guess the starting point for me, and this is the way that we kind of start the deck, is just imagining a day and a couple years. How do people get news and information? Where is it coming to them?

6:11and our premise is that AI is going to be a kind of central intermediating force for any kind of like personally relevant information acquisition that people do in their lives. And we talk about this idea of kind of an agentic interface which for different people is going to look to this part. Please explain this without using the words agentic interface. Okay, sorry. Yes. We have these words like medium right? But even medium to me suggests like there's a particular form that you're receiving information through. The medium is video. The medium is text. And what's really weird about this era that we're moving into and why it's kind of hard to even come up with the right words is that it can be any medium.

7:02If what you want is to be hearing someone who sounds like Taylor or Swift deliver a song that tells you everything you need to know about your kid's school schedule, like that is going to be available to you. And if what you want is a video, you know, newscast version, or what you want is a written briefing that looks like the president's daily briefing, like all of those things are available to you and we'll be able to kind of seamlessly synthesize the most relevant information for you in the way that you prefer to get it. And so, I don't know, we call that agentic interface. I kind of hate yet more jargon.

7:43But the core idea is that you'll have AI doing a number of jobs, doing this job of consuming a ton of information that you don't need to consume and figuring out which parts of it are relevant, bringing in your personal context, translating that into compelling, like, quote, unquote content and making that available to you in some kind of interactive way. And that's really different, I think, as a paradigm than the one that we've sort of become accustomed to. And so part of what I wanted to do with the deck a little bit was just like remind us, you know, sort of like the David Foster Wallace, like this is water.

8:18Like we've been swimming in the water of big centralized algorithm land for a while. And it's easy to imagine that's like a permanent state of affairs. And I actually think we're moving into something that's different. Let's break this down kind of in a similar way to what you do in the deck, which I thought was really fascinating, a little bit chilling, a little bit exciting, a whole lot of emotions kind of all at once. So right now what I do, unfortunately, every day I wake up, the first thing I do is I check my phone. Usually I've got some texts from Ben because he wakes up a little bit earlier than me.

8:52I scroll through the various emails. That's okay, Ben. I scroll through various emails that I've received, newsletters. I read the New York Times app. Maybe I'll listen to a podcast or something like that when I'm kind of getting ready. You envision my day will start a lot differently in either a probably not that long of an amount of time because of the way that these things are rapidly changing. Walk us through what my day or the days of our listeners are going to look like in this new world that you envision. If you think about that set of tasks that you're doing, you're checking text messages, you're looking at some group chats, you're checking some news sites, you're checking some emails.

9:32All of these are like information sources that right now you're having to kind of manually like seek out. There's a little bit of like you're doing some self-direction, but they're living these different kinds of information are living in these different silos. And I guess the premise here, and if you listen to, by the way, Google I.O. happened, I think, like a few days after we put the deck out and it was like, they're launching a lot of stuff that like pushes toward this exact same kind of place. That all becomes one seamless experience where it's like, we're going to make the personal, you know, semaphore newsletter just for you, Max.

10:05And it's going to have a ranked, you know, kind of like what's most relevant and most important for you to know. And probably the Ben texts are still at the top, but maybe Ben is sending you a link to a document saying, what do you think about this? It's ingesting the document. It's saying Ben sent you a document that was about the new strategy that has these three main points. You should take a look at it. Doing a bunch of synthesis so that that surface becomes the primary place that you're paying attention. And part of the reason that we wanted to have Ben in dialogue during the deck launch was that, you know, there's a few of us who have lived through this last era of kind of attention mechanics, attention mongering, and who's paying attention to what is one of the coins of the realm we now all realize.

10:54And like, this interface, I think, is going to own a lot of our attention and intermediate a lot of those other sources and mix across kind of personal and what we think of as public media contexts. When is this transition going to happen? Obviously, this is going to happen in pieces and kind of slowly. But when do you think my morning routine will look substantially different? I think even in a year, there will be substantial parts of this that we will be starting to rely on. Like, just if you think about Google alone, like I see Google trying to do this coming from multiple different angles where in Gmail, they're starting to like think past the sort of ranked list of emails as the way that people interact with their email.

11:44Right. That's like one dimension. At the same time, they're like slowly trying to creep Gemini toward having more and more access to more and more of your own personal context. and all of that's going to be like a place where you begin to be able to cross-reference different pieces of data. Gemini just yesterday asked me to be able to see what's going on in all of my tabs and like to get a sense of like, what are my media sources? What am I like consuming? And so I think all of that's going to start to be really productive and useful if also creepy and surveillance. And all of us are going to have to make some decisions which are going to be easy to say yes to in the moment because they will be useful about what kinds of information we want to give them.

12:26But yeah, I would say in a year, we're going to start to really see new behaviors in terms of how people are kind of doing their basic, like, morning information routine. If our attention, as you say, is sort of shifting to this new surface that we're all starting to see, but obviously is emerging, what is going away? As you say, we've been swimming in the water of, for me, X, mostly, but for other people, other social platforms, for the last 10, 15 years, what is this going to replace? What's going to go away? And when we look back, say, oh my God, remember back in 2025, when in 2026, when we were still looking at whatever.

13:03So big parts of the web as we know it are going to kind of get, are being eroded already. And by the way, the part of this conversation that we haven't talked about yet, but it's partly because I just feel like it's a little more obvious is, meanwhile, you have this wave of like AI generated content also. So it's sort of like the joke where like one person takes three bullet points and makes it into a research report, another person takes the research report and makes it into three bullet points. There's going to be a lot of generation and summarization happening across the ecosystem. But I think the other piece that I feel like is under real stress and threat in this context is the feed-based information economy.

13:45We already know that people are kind of taking their personal lives and moving to the the underground, the group chats and the discords and these other places that are sort of like a little more safe and personal. But I think, you know, the risk versus reward of putting things out there publicly as a person on an internet where there are these weaponized agentic tools is going to continue to shift in the direction of people kind of like holding their information and their lives pretty close. You're saying essentially that the feeds that are so much of the part of of our lives on the internet are under threat by this.

14:24I guess, why would those become kind of less relevant? Isn't kind of part of the fun of the feeds right now that there is some level of discovery of something new and kind of unexpected, and that's the reason why people are coming to these feeds? And feeds meaning, I guess, Instagram, TikTok, Twitter? Yeah. What people do for sort of semi-passive entertainment is a really different set of behaviors from what they do to get to know what they need to know in the course of a day that's sort of useful information, right? And there's some bleed over between those two, but I would just say like, I feel like less clear about the future of that like entertainment and TikTok-y use case than I do.

15:04So we'll still have TikTok, that's a relief. We'll have something, but again, there's these questions of like, what really is defensible in an AI era and what is not from a kind of like video standpoint that I think you really have to scrutinize with some humility. Humility might not be the right word, but like that any assumption that AI cannot categorically make a certain kind of content, I feel like is increasingly like a bad bet. And so you have to kind of imagine that whatever amount of engagingness, whatever amount of like, quote unquote, authenticity. And I think Adam Masseri said something similar recently, you know, just that like the notion that humans will be able to out authentic AI, I think is like a questionable notion.

15:50So it really kind of then raises these questions about mass public communication. And there I feel like volume and the ability to optimize is really powerful. I wanted to actually dig in on your platforms part because one of the things that I thought was most interesting in the deck was you created a list of which platforms you thought were the most exposed to disruption by AI and which ones you thought were the most kind of protected. Can you walk through some of them and explain which ones you thought were the most exposed to disruption by AI and which ones were the most insulated? So the way that we approached this was that we looked at, you know, we sort of use this jobs to be done idea.

16:31Like people go to platforms because they want to be entertained or they want to be connected with other people who have a shared experience or they want to, you know, there are these particular things that people are looking for. So we kind of tried to break it down into all of the component parts of like, what makes people show up there? I'm bored and I want to be entertained. And then we kind of thought about in which cases might an AI built experience be just as good or better? Maybe because maybe it's sort of a classic disruption thing where it's like not quite as good, but it's twice as available, or it's twice as cheap or whatever.

17:07You know, Ethan Mullick has this great quote that, you know, people use AI when it's better than the best available human. And my therapist is better than Claude, but my therapist isn't available at midnight when I'm stressing out. So the question is like, when is AI going to be better able to do some of the jobs that social media does than the best available, easily findable social experience? And I think at the far end of that spectrum, where I think there's going to be a lot of displacement, like a lot of YouTube content is, I was just trying to replace the wheel on my dishwasher. And the best way to learn that was to get a little video that was some dude, you know, showing people how to replace every part in a dishwasher to make a few bucks.

17:52Like, I don't know that that's a viable, like putting the manual in my cloud and having cloud tell me was like actually way better. And there's a bunch of things like that where really what I'm seeking is a piece of specific information or to complete a specific task where I just think, and I'm not like cheerleading this, by the way, but like, I just think that there's going to be a lot of competitive pressure from AI tools that know more about my personal context, know how I like to learn and are able to like bring that information into my context in a useful way. On the other extreme, you know, so I do think there are these pieces of the big kind of content library and content experience apps that are going to be under threat.

18:34On the other extreme, we talk about when I text my group chat of friends, like I do not want my friends' agents replying to my texts. That's not the point of the group chat. And maybe our agents someday will have their separate group chat where they like gripe about us and figure out how to manage our calendars together so we can actually hang out. But like, I don't, I want to talk to my friends when I want to talk to my friends. And there are a bunch of places in the social web where humans want to talk to humans. And I think those places have a lot more durability in their kind of current form.

19:11So something like WhatsApp, which is a community of group chats, there's a lot of durability there because the core use case is not something that I think people are eager to replace with an AI substitute. You were sort of maybe most famous for coining the phrase the filter bubble, which I think really at times has just been the defining experience of particularly political media, just that you live in and all the Democrats and Republicans in the United States live in just totally alternate information environments. And some of the things that you are saying now sound very filter bubbly. Some sound almost the opposite.

19:42They sound like you're maybe in a bubble of one and everyone is in a bubble of one. And I'm curious how you, how this, and maybe particularly for these purposes, news environment and kind of information environment, how does that connect to that idea of the filter bubble? Which I think you rewrote a book in 2011 called The Filter Bubble. Yeah. The power dynamics are different. And there are some ways in which it's just, it will be kind of like a hyper bubble, bubble of one phenomenon where I think there are a lot of things that are not yet set about the way that we do this, right? And part of my interest, I'm like a public interest, public spirited guy.

20:18I want a conversation about this because I want us to like look at it and shape it. And it's still very shapeable because a whole bunch of things about how this all plays out are not yet set. So just to like name my intention, it would be very easy to make this the ultimate like self-reinforcing epistemic closure machine, right? that AI can convince you you're right better than any technology ever has been able to convince you of that and show you all of the evidence better than any technology ever has been. On the flip side, I guess one of the things that I won't miss is whether it's Elon Musk or Mark Zuckerberg, a singular monocultural logic that mediates cultural production.

21:04Like, I don't think that's been great. I guess I'm with Kyle, a filter world, Kyle Chayka. Kyle Chayka. Well, wait, explain what you mean by that, because I don't follow. Ultimately, when people learn what the TikTok algorithm, the Facebook algorithm like, and build content to support it, I did that for a while, I ran Upworthy. And it is not a recipe for a diverse and varied media ecosystem, right? it's built to uplift a handful of things in different kinds of combinations, but that have to do with surprise or have to do with conflict or have to do with a few other things. I think this is going to be a different kind of, a more varied world.

21:50And so you're just saying like, that's a world where every coffee shop looks the same because every coffee shop is chasing the same Instagram sort of juice. Yeah, exactly. So are we going to get like, this AI produces a world of like, more varied coffee shops? I think it could be because the filters become much more bespoke and weird and idiosyncratic the way that people are. It's a coffee shop where we're the only customer and sit alone. But it's really cool and weird. Well, we have a lot more that we want to get to with Eli, but we have to take a short break. So we'll be right back with more after this.

22:32In this week's Branded segment from Think with Google, I talked with Google's VP of Marketing, Josh Spanier, about the massive increase in search queries. So Josh, in Google's last earnings call, the CEO, Sundar Pichai, said that search queries are at an all-time high. What's behind that? What's going on there? So search has always been synonymous with 10 blue links. The reality is today, Google Search is unbelievably more competent, capable, and intelligent than it's ever been before. And what we see under the hood is when consumers use all the advanced features of Google Search, when they're planning a trip, when they're booking a meal, when they're investigating quantum tunneling using AI mode, they're actually searching more and more and more, more than they've ever searched before.

23:16And they're kind of transitioning from, I think, queries into quests. It used to be you wanted to know a specific piece of information, you just put in a query. Today, that query becomes a whole quest, a whole series of deeper and deeper explorations using Google's technology. We're way beyond 10 blue links. And what should marketers do about this? The great thing is that AI has transformed Google Search in that there's longer queries, more conversational queries happening. The great news is these quests that people are doing with Google Search are resulting in more insights, more signals, more ways to find customers as they're thinking about your product and as they're shopping.

23:57AI Max for Search is a year old. It's one of the most successful products launched in Google's history. AI tools, AI systems, Google's AI investment is actually helping marketers find customers for less money and drive more conversions. It's a great thing. Everyone's winning. Where can people, where can they quest for more? Head on over to thinkwithgoogle.com and just search for AI Max for search or just AI mode or really anything to do with AI. There's so much great content on thinkwithgoogle.com.

24:37you mentioned upworthy and i've always actually been curious because you you write this book the filter bubble about exactly these mechanisms that you're describing and then and i was at buzzfeed at the time so i'm not sort of i'm not condemning cynical traffic mongering but i've I've always wondered, it felt like you then like write this book discovering, you know, sort of this terrible technology that distorts the world. And then you're like, and I will use it to build a company. No, I wanted to build a company that specifically was focused on elevating what we thought was like socially valuable, socially, the stuff that people could should be paying attention to, but that was not likely to be optimized for algorithmic success.

25:15That was the story we told ourselves. Use the tools for good was the idea. Yeah, yeah, that was the idea. And then the short version of the story, as you have chronicled in others, is that it turns out that like when you get really good at a tool and you're using it for good, but other people are just interested in the tool, which is like a particular form of headline or a particular analysis of where spaces and the ecosystem, they can take that tool and do whatever they want with it. And indeed, that's what happened. So there are lessons there about sort of, I don't know, there's a whole a whole set of philosophical ethics, I think, that you could get into about means and ends.

25:51That reminds me a little bit of one of Ben's former employees at BuzzFeed. Maybe not your employees specifically, Ben, but Benny Johnson, who used to work at BuzzFeed and was just this. He was yours. I wasn't sure who brought him on, but of course, it was just this guy who was an expert at virality who, upon leaving BuzzFeed, entered on this journey of using those virality powers for something quite different, which is this kind of political project. And, you know, he never lost his touch for understanding what works well on different platforms. That was true throughout his entire trajectory. Sorry, Ben.