The AI Slopocalypse

0:00Today on the AI Daily Brief, we are discussing the AI Slopocalypse. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

0:36smartest AI doers and practitioners in the land, shoot us a note at sponsors at AIDailyBrief.ai. Now, like I said, we have an interesting one today. It is the weekend, which means, of course, it's a long read slash big thing type episode. And this week, like many of you out there, my attention was caught by the story in The Hollywood Reporter about a new podcast studio that is, depending on your perspective, a fascinating evolution of the content creation model in a new agentic era, or alternatively, a harbinger of the end times where all of us die drowning in AI slop. And despite the fact that this is nominally AI coming for my industry, I probably have a bit more at least open to it kind of perspective than you might think.

1:20So let's start by talking about what this new studio, Inception Point AI, is doing and get the first reactions. The title of the piece in The Hollywood Reporter was 5 ,000 podcasts, 3 ,000 episodes a week,$1 cost per episode. And the basic idea of this is to use AI end-to-end to create new podcasts. That means AI for scripting, AI for voices, AI to do the production, etc, etc. Yes, they are actually inventing AI people, including this handsome gent over here, Nigel Fistledown. The company's podcast network is called Quiet Please and already produces more than 3 ,000 episodes a week across its 5 ,000 shows.

1:57Since September of 2023, when it started, they've seen over 10 million downloads cumulatively. The company says it takes about an hour to create an episode, from idea to actual release. In terms of what these podcasts are, the company produces different levels of podcasts. The lowest level involves weather reports for various geographic areas or simple biographies, and higher levels involve subject area podcasts hosted by one of about 50 AI personalities they've created, including food expert Claire Delish, gardener and nature expert Nigel Thistledown, and Ollie Bennett, who covers offbeat sports.

2:30ESPN's The Ocho, anyone? If you go over to their website, they talk about not just making content, but crafting characters. Each AI personality, they say, is built from the ground up, styled, voiced, and fine-tuned to engage and connect audiences in real time. One of their big provinces, they call from episodes to universes. We don't just create shows, we build entire branded ecosystems. Each vertical holds a voice-driven world powered by our AI talent. And importantly, this is not some 21-year-old coming out of Y Combinator with an idea to disrupt an industry that they're not a part of. The company is being led by CEO Janine Wright, who was previously the COO of Wondery, which is the big podcasting conglomerate that is now owned by Amazon.

3:10Now, a big part of the idea here is the economics of it all. The company says that they're able to produce each episode for a dollar or less, and then they attach it to programmatic advertising. They say then that before factoring in for overhead, if only about 20 people listen to an episode, the company actually turns a profit on that particular episode. Said, right, we might make a pollen episode that maybe only 50 people listen to, but I'm already at unit profitability on that. And so then maybe I make 500 pollen report podcasts. There are a ton of tools that go into this. They say they're using something like 184 different custom AI platforms or agents.

3:43And that's the basic idea. Now, to get a sense of how people feel about this, one only has to copy the URL into the search bar on X, and you will very quickly get a feel for the tenor of the conversation. Michael Sokolow writes about what the company does and then says, It's all AI slop, though. 100%. No real hosts. No real guests. What are we doing? David Ryan writes, Hard to pick out which part of this I hate the most. Spoiler, it's all of it. He also then copied another part of the article where they talk about how the titles were optimized for SEO and added, Please launch me into the sun. Nate DeFort writes, a new AI supervillain has entered the arena.

4:17And Georgia Cohn writes, as an audio producer, the idea of an AI podcast startup is depressing in itself, but then you see it's being led by a former exec at Wondery and you feel despair. Any creative with an ounce of self-respect would be embarrassed to put their name to it. Lori Kilmartin said simply, death, please take me. And Augustine LeBron wins the prize for naming the title of the episode with his tweet, the AI sloppocalypse is here. So let's now talk about what I think about this and try to move a little bit beyond the knee-jerk reactions, which are quite clear, as you can see. Let's discuss first, outside of the knee-jerk, just antagonism towards this, what is actually potentially bad?

4:54First of all, let's talk about the idea of out-competing human podcasts. This is the question that I sometimes get if I'm worried about AI taking my job as a podcaster. And the short answer is, frankly, absolutely not. Podcasts are not a medium that exists simply to disseminate information. They're a medium that people turn to for all the nuance and discourse and exploration and interesting opinion and analysis and synthesis around their interests and other important information that compels them to actually take the time to listen. For most mainstream topics and interests, there are already dozens, hundreds, or even thousands of podcasts about that exact topic.

5:33The ones that make it to the top have something special, something unique, something that the listener responds to in those podcasts that is really unique to them. I do not believe that just because there are more podcasts, all of a sudden the AI is going to outcompete the human podcasts. And to the extent that people figure out how to make super compelling podcasts using AI, it'll just be part of the landscape that people choose between. Except insofar as people have ultimately limited attention, podcasting is not really a zero-sum game. It's one of the reasons that podcasters tend to be friendly with one another and drop each other's shows on their feeds.

6:07It just tends not to be the case that if someone finds a new podcast that they like, that they stop listening to all the others. And so when it comes to things that are potentially bad about this, I really don't think that out-competing human podcasts is anywhere near the top of the list. There are, however, some things that I think are potentially problematic about this. First of all, I do believe that this is very likely to make an already bad discovery problem potentially much worse. Finding new podcast content is extremely difficult. It basically only works by either A, having a different distribution channel somewhere else that you've built that you can drop your podcast into, B, advertising dollars that put it in front of people, or C, the host companies, specifically Spotify and Apple, because no one else is even close to big enough to make a dent, deciding that they like a particular show and featuring it and making it more discoverable for their listeners.

6:59This means that it is very hard to break through with podcasts. There is an extreme power law where the vast majority of podcasts are heard by a tiny, tiny number of people and never really make it past their first few episodes. I do think that all of a sudden having a flood of 5 ,000, 10 ,000, 20 ,000 ultimately new podcasts into those systems could potentially make the discovery problem worse. But I will also say that that's kind of not Inception Point's problem. That is a challenge for Spotify and for Apple and for the companies who are the core discovery platforms. It's an opportunity for entrepreneurs who want to try to make podcast discovery better, although there have been dozens and dozens of those startups and they've never really worked.

7:39Still, to the extent that we are keeping track of good and bad, I do think that the mass influx of AI podcasts makes a problem that already exists even more challenging going forward. Could this confuse people when it comes to what is real and what is AI? Once again, it seems like, at least from what the reporter writes, that the company is trying to be really ethical about this. They write, The team is in the midst of navigating the ethics around creating these AI personalities as the technology advances. Each host now identifies themselves as being AI at the top of the episode, and they've stayed away from having the hosts invent their own backstories for now, but that could come.

8:13Said William Corbin, co-founder and CTO, I'm not going to create a personality that somebody has a deep relationship with. They've also decided not to do hard news at the time, but of course all of that is subject to change. Point being is that they're very clearly not trying to trick anyone into thinking that these are real humans versus AIs, but I still think that even with that identification, it is confusing. I listened to a few of the episodes just to get a feel for them, and the scripting talks about and shares experiences as though they had actually had them, when of course they hadn't.

8:43I listened to the first part of a knitting podcast, and they were talking about feels and emotions and sentiments associated with a particular type of weather or timing. And of course, an AI is just a model. It's never had those actual emotions. So it's just parroting what a human host would be. Meaning again, that even though they are, yes, identifying themselves as AI, it still gets blurry. There's also the problem that the discovery platforms right now don't really make clear differentiation between real and AI. At least I don't think right now you can click a filter to say, I don't want any podcasts from AI hosts.

9:14Although you gotta think that that's the sort of UX change that will be coming sooner rather than later. Could it get so bad that these things make people stop listening to podcasts entirely? Again, I don't really think so. The reality is, is that there is already so much noise for very little signal in the podcast space. The barriers to entry are still really hard to discover good content, as we just talked about. That if people have made it through this sea of human-created podcasts to find the ones that they like, I don't think the presence of these AI shows is going to turn them off to the field entirely.

9:45Lastly, on the what's potentially bad about this, could these things increase the tyranny of the algorithm, specifically the discovery algorithm? The answer is a big fat yes, but we will come back to that towards the end of the show. What if AI wasn't just a buzzword, but a business imperative? On You Can With AI, we take you inside the boardrooms and strategy sessions of the world's most forward-thinking enterprises. Hosted by me, Nathaniel Whittemore, and powered by KPMG, this seven-part series delivers real-world insights from leaders who are scaling AI with purpose. From aligning culture and leadership to building trust, data readiness, and deploying AI agents.

10:22Whether you're a C-suite executive, strategist, or innovator, this podcast is your front-row seat to the future of enterprise AI. So go check it out at www.kpmg.us slash AI podcasts, or search YouPen with AI on Spotify, Apple Podcasts, or wherever you get your podcasts. If you are a regular listener, you will have heard about Superintelligent's agent readiness audits at this point. But I wanted to tell you today about the full suite of agent readiness products that go beyond just the initial readiness report. Over the last six months, Superintelligent has built out an entire agent planning suite.

10:58We help you move from discovery to planning to implementation. After you've completed your agent readiness audits, we help you double-click on your most important use cases with what we call our use case planning reports. These reports are going to help you understand what sort of technical preparation you need to do to be ready for a use case, what challenges you might face in implementation, and whether you should be thinking about building, buying, partnering, or some combination. After that, you can even get a spec document in what we call our technical blueprint that gives either your developers or the developers of the partner you work with what they need to build exactly the agent that you're looking for.

11:31If you want to learn more about Superintelligence Agent Planning Suite, we've built a custom GPT to answer your questions. Just go to bit.ly slash super super agent. That's bit.ly slash super super agent, all one word. And if you have any questions, the agent can even help you book an appointment with our team. AI isn't a one-off project. It's a partnership that has to evolve as the technology does. Robots and pencils work side by side with clients to bring practical AI into every phase. Automation, personalization, decision support, and optimization. They prove what works through applied experimentation and build systems that amplify human potential.

12:11As an AWS-certified partner with Global Delivery Centers, Robots & Pencils combines reach with high-touch service. Where others hand off, they stay engaged. Because partnership isn't a project plan. It's a commitment. As AI advances, so will their solutions. That's long-term value. Progress starts with the right partner. Start with Robots & Pencils at robotsandpencils.com slash AI Daily Brief. Now for the sake of completeness, let's talk about what is good about this, or at least interesting. I think the very, very clear thing that stands out is that this potentially supports incredibly niche interests.

12:47This was already something that was very cool about the internet, that in the era of Web 2.0 and platforms where people could create their own communities, you could find other people who really like the thing that you do, even if some mainstream content producer hadn't decided that it was worth producing content about. The entirety of Reddit is basically this, right? Now, I do think that already podcasts have a really long tail of content that they cover. For example, this time of year, I am always interested in Lovecraftian audio drama type content. And there are, believe it or not, a lot of podcasts that build off of those Lovecraftian horror themes.

13:25And I'm not talking about one or two, I'm talking about dozens. But there are niches within niches, man. So to go a little bit farther on that interest, there's a series of books that place Sherlock Holmes in a Lovecraftian context that I read a couple of years ago, and have become perpetual rereads this time of year. I would love podcasts that got even deeper down that rabbit hole of the combination of Sherlock Holmes and Lovecraft. And right now, to the best of my knowledge, there are not any podcasts that have that. So the idea that there's this thing that supports incredibly niche interests, I think, could be really good.

13:56The pollen example that they gave also is really interesting from a hyper specialized news perspective. They're not doing hard news now, but really hyper localized news could be something really cool that this podcast model supports that traditional production methods just wouldn't. Local news, for example, already really struggles to survive, and it's incredibly valuable when it's done well. We have an amazing local news production around me that covers my town and the town next to it, but it is constantly on the verge of going under. So maybe this is another alternative path for that sort of thing.

14:26To me, this is the main value proposition and why I can see this being worth pursuing outside of any particular economic interest, just from the standpoint of why this might be valuable to the listener base and society as a whole. But now let's talk about what is completely inevitable about this. This is, to use an idea I've brought up a lot on this show, basically the Dr. Strange theory as a business model. So what's the Dr. Strange theory? I have a whole episode about this. If you actually just Google for Dr. Strange theory and agents, you'll find it. But basically the idea references the part in the culmination of the Infinity Saga in the Marvel Cinematic Universe, where the Avengers are all rocked back on their heels and they're trying to figure out if there's any chance that they can beat Thanos, who is the big bad of the series.

15:08He's acquired the Infinity Stones, and it seems like things are impossible. Doctor Strange starts to meditate and looks across the multiverse, examining all of these different parallel universes where versions of this same story are playing out at the same time. He comes back and Tony Stark, Iron Man, asks him, how many do we win in? Dr. Strange had gone and looked at over 14 million parallel universes, and there was only one in which the Avengers succeeded. Now, taking this to the realm of AI, the reference point is the idea that in a world of unlimited intelligence, too cheap to meter, we can theoretically run parallel processing at anything that we are trying to accomplish, any problem that we're trying to solve.

15:48So to take a very easy example from the realm of business and content production, when we think about how AI and agents are impacting and will impact content, it's sort of something like this. Right now in the early stages, we are already using AI to be more efficient in how we write content. We use ChatGPT to help us research and to draft it. And I would guess that a pretty meaningful percentage of business tweets that you see on X or other platforms like it were written in whole or in part by ChatGPT or Claude, etc. Now, as we move into the agentic era, people are kind of imagining an even more complete one-to-one replacement for currently non-agentic functions with agentic functions.

16:25Instead of asking ChatGPT for help writing those tweets, you're just going to have the ChatGPT agent do it entirely for you. The agent is going to be empowered to go do the planning, figure out what the topic should be, test the tweet, and actually send it. And that's kind of where people's imagination ends. But again, in a world of unlimited intelligence, too cheap to meter. I don't think you're just going to have one agent writing a tweet. I think you're going to have a hundred agents writing tweets, all with totally different constraints and things that they're trying to accomplish. 20 agents will be writing in the style of famous authors.

16:5520 will be imitating the previous brand voice of your brand. Another 20 might be imitating the brand voice of other brands in your space. And then the last set might be toggled to just random. Meanwhile, there will be a set of agents whose job is to impersonate different audiences in ICPs. They will react to and rate those different tweets based on the audience that they are the synthetic equivalent of. Then you will likely have aggregator or analyzer agents who take all of that information, sum it all up, make a determination of what the most high potential tweets are. And maybe they gave the one human who's in the loop on this, the choice of the top three, with a whole index of information around why those three have been selected.

17:31Now, obviously, this might be overkill for your average tweet. And right now, this would probably be more expensive and complex than you would care about for most social media content. But that's not going to be the case for long. And the point more than the specifics is that we have barely begun to scratch the surface of what it's going to look like when you can just create thousands, if not millions of times more of everything in an instant at basically no cost. This, my friends, is one of the first and fullest examples of that as a business model that we've seen. The entire premise is that in the land of programmatic advertising, if you can reduce the cost of production to a buck a show, it takes barely any listenership at all for it to be actually net profitable.

18:12The point is that I think that at least experiments like this are completely inevitable. I don't think that they will all work, but I think in every case where some version of this can be tried, it will be tried. I certainly think you're going to see this in basically every single content medium. A recent Atlantic article shows how this is happening in the world of YouTube. And of course, if you've been on TikTok at any time in the post VO3 release, you've seen a huge increase in AI content. And that is just in the realm of content production. Now, I think this is an obvious place for all of this to start, but by no means do I think that that's where it's going to end.

18:45So here's a big question, though. Will it work? Just because a model is inevitable from a test perspective doesn't mean that it's going to actually play out in practice. My bet is that in this particular case, the answer is no. The reason is I sort of think the form factor for what they're trying to achieve is wrong. Here's how I'll try to describe it. We've got a four-quadrant graph here. And apologies if you were listening and not watching, but basically it's a four-quadrant graph where the y-axis moves from my interest at the top to not my interest at the bottom. And on the x-axis, the left side is bad, the right side is great.

19:18Most people are going to spend as much of their content consumption time as they can in the upper right-hand quadrant, the one where the content is great and is my interests. Now, one of the things that has happened with podcasts is that sometimes the content is so great that even if it's not your interest, you'll listen. Another thing that happens with podcasts is that sometimes your interests matter so much to you that you'll listen even to objectively kind of bad content about it. Hence the long tail of my Lovecraftian podcasts. And then of course, there's the bottom left quadrant, the bad and not my interests where you're not gonna spend any time at all.

19:49I think that effectively, what they are going for with this, not to besmirch their aspirations, maybe they really think that this AI content can be great, But my guess is that if they're talking about an hour of production per show and keeping it under a dollar, greatness is not really the key metric. I think that the opportunity they're going for is the sort of top left quadrant where they're looking for something that is sufficiently of my interest that I'll deal with middling quality. Of course, they're not trying to have bad quality. They're trying for these things to be good. But really what they're trying to do is find interests that are too niche for there to be any content, good, bad, or otherwise.

20:22And frankly, for this to work, I really think that there does need to be almost nothing else for that interest. I think that a person who shares an interest and a passion is almost always going to outperform these AIs, at least in current state, although I'm always reticent to assume that that will be the case forever. But why I say the form factor is wrong is that I do think that there is an opportunity for this idea of sufficiently my interest that I'll deal with middling quality. But I think that the form that this wants to take is content I spun up for myself, not mass-produced content. In other words, I think that what will make sense for this is generative platforms where you can personally customize with a click or a couple of clicks exactly the content you want based on your exact niche as narrowly as you want to define it.

21:07Basically, I think that even with 5 ,000 podcasts, they're going to have a hard time getting a sufficient breadth of narrow niche interests for it to really work. Like my guess is that they're not going to figure out a Holmes Lovecraft crossover podcast idea. And if they do, I like that so much that if there's an option between listening to their version or working with a generative platform that's creating something equivalent, but exactly to my tastes, I think that that other platform, which to be fair doesn't exist yet, is a likely better home for this type of effort. Also, indicators currently suggest that there is a huge gap between people who are great at AI creation and those that aren't.

21:44TikTok videos, I think, are a perfect example of this. There have been infinite talking Bigfoot vlogs and Harry Potter vlogs and all this sort of stuff, but there is a massive difference between the people who are really good at it and who are ultimately going viral in a sustained way and everyone else who's just doing the same thing. I could totally be wrong. Maybe they can be sufficiently broad that they find enough people to make this work. And like I said, I think that the underlying idea of catering to really niche interests is something that AI is going to create a lot of really exciting opportunities for.

22:15I just don't think it's going to be in this exact format. Now, one of the biggest implications, and something I wanted to come back to from the what's potentially bad, is that there is no way in the world that this entire shift, both with Inception Point and all the things that come after it, don't increase the tyranny of the discovery algorithms. The platforms that distribute content are going to get even more powerful. In a world where the amount of stuff that is available for you to consume goes up a million fold, which sounds crazy but I don't think is actually out of range, you are going to have to rely on some intermediary to suggest things for you because it's going to be impossible to sample it all for yourself.

22:55We are already dealing with the challenging consequences of a world that is arbitrated by social media algorithms, and I think that that challenge gets nothing but bigger in this Doctor Strange content modality. I do also think that it creates opportunities. I would be shocked if we don't get some new native platforms for whatever types of content come next, but they are going to come with enormous responsibility and power. Now, one of the big questions that I have is whether their business model and economics make it so that they basically can't fail, or if there is at some point a backlash that makes it not work.

23:28On the one hand, if they really only need 20 people to listen to a$1 produced podcast, for it to be profitable, it seems like there is quite a powerful money printing loop right there. Now, of course, that doesn't factor in how much overhead you need to figure out how to nimbly change what podcasts are being produced. But then again, maybe AI can do all that. So on the one hand, if those really are the economics, it's totally possible for them to build a model where people don't really like the content all that much, at least not initially, but they're still economically successful. Then again, it's pretty easy to see how the backlash could happen.