AI summaries are raising the bar for content quality... with Pete Pachal

Summary

In this conversation, Jeffro and Pete Pachal discuss the transformative impact of AI on media and marketing. They explore how AI is changing the way information is consumed, the implications for content creation, and the ethical concerns surrounding AI-generated imagery. The discussion also touches on the future of AI and its potential to solve significant problems while addressing the challenges of intent and general intelligence.

Takeaways

• AI is fundamentally changing how we consume media and information.
• The role of AI as a disintermediary raises existential questions about information accuracy and bias.
• Content creation costs are decreasing due to AI, but quality must remain a priority.
• AI can filter out low-quality content, promoting unique and valuable contributions.
• Ethical considerations around AI-generated imagery are becoming increasingly important.
• Businesses need to establish clear AI policies to navigate new challenges.
• AI can enhance content creation processes but should not replace human involvement.
• The future of AI may not include super intelligence but can still address significant issues.
• AI-powered search is evolving, providing more accurate and reliable information.
• The balance between using AI effectively and avoiding its misuse is crucial.

Chapters

00:00 The Transformative Power of AI in Media
03:02 AI as a Disintermediary in Information Consumption
06:02 The Impact of AI on Content Creation and Marketing
08:50 Navigating Copyright and Ethical Concerns in AI Imagery
12:09 The Future of AI: Intent and General Intelligence

Links

https://mediacopilot.substack.com/
https://mediacopilot.ai/

Free Website Evaluation: FroBro.com/Dominate

Transcript

Jeffro (00:00.972)
Welcome back to Digital Dominance. Today, we're tackling what has become one of the most transformative forces in media and marketing, AI. My guest is Pete Paschal, the founder and CEO of the Media Copilot, where he explores how generative AI is reshaping the media landscape. has years of experience in journalism and senior roles at CoinDesk and Mashable. And Pete now helps professionals and organizations harness AI to upgrade their work and navigate all of the implications that come with it.

So in this episode, we'll dive into the rise of AI generated summaries, the blurring line between real and artificial images, and how these changes are impacting small business marketing. Pete, thanks for joining me today.

Pete Pachal (00:42.143)
Hey, my pleasure, Jafro. Glad to be here.

Jeffro (00:44.716)
Yeah, know, Pete, AI continues to change so many industries at an impressive pace. And I'm excited to dig into how these shifts are also creating new challenges and opportunities for small businesses in particular. So that's what we'll talk about today.

Pete Pachal (01:00.875)
Cool. Yeah, I'm totally down for it.

Jeffro (01:01.869)
How do you see AI reshaping the way people consume media and information?

Pete Pachal (01:10.497)
Right. That's a, that's a big, start with the small questions, right? so AI is rapidly becoming this layer between, you know, information and the way people get it. It's, it's, it's basically a disintermediary as they sort of talk about, right? So there's all this info like articles, whatever it is. And previous to AI, how you would sort that out is Google, right? You Google it. Okay. You get a bunch of links. You click on a few.

Jeffro (01:13.686)
I know, right? We gotta get right into it.

Pete Pachal (01:40.417)
that look interesting, you read bits of them, figure out the thing that is unique to each one, and you form in your head some version of what you were looking for, whether that was news or whether it was like the best recommendations for a peppermint vodka, martini, or what have you, right? And so now AI has come in and said, I'm gonna do that for you. I'm gonna do the Googling and check it out, and I'm gonna summarize all that information and just...

kind of give it to you and give you like the short version of that. And that's no doubt that is helpful, right, for a lot of queries. it's essentially removing friction, which is what technology does. Of course, it invites a whole host of questions, which is, know, which of those links is it looking at? Is it biased in its selections? Is it accurate? And, you know, what does it mean when we start to get this, you know, average?

of everything out there as opposed to sort of looking for for ourselves. These are all sort of, honestly, existential questions to the information ecosystem. And the technology is getting better, but it still has all of these sort of implications for the people creating the information and for the people consuming it. it's a big deal, probably the biggest sea change since the printing press, honestly, that's not really much of an exaggeration.

Jeffro (03:02.765)
Well, and I heard recently about how many searches now are not leaving the search results page because as you mentioned, those summaries from AI are answering the question. So they don't need to click on anything. But now, now is where the gray area comes in, right? Cause like you said, are there biases? How is the algorithm picking which ones to look at? Is it assessing AI written content from these sources? Is it human written? Do we know now it's AI interpreting AI? So what do we do from here? Yeah.

Pete Pachal (03:25.451)
Right.

Pete Pachal (03:29.781)
Yeah, all of that.

Yeah, that's all happening and it's all a concern and it's still early days. So I feel like a lot of using AI and whether it's to create content, well, particularly what to create, let's look at that zero in on like people creating content with AI. I think that is something that is essentially taking this stuff

this AI written stuff and inserting it into a web two world where the laws are all about social reach and search functionality. And it's going to become and it already has become really, really easy to gain that. know, like the cost of content is going to zero. You can get AIs to summarize things. You can publish them and then think you've done your job. The problem is, you know, the filters are getting better.

Essentially, like if you all you're doing is putting out AI generated, you know, swap, so to speak, you're going to get downranked, whether it's your site or yourself or whatever. And the volume of that content is getting so large that it's a lot of it's just kind of like getting filtered out. It doesn't make any difference. So you got to do something better if you're going to try to stand out and market your stuff and get in these sort of AI summaries, because what these

AI engines are looking for is stuff that's unique, is stuff that's differentiated, is stuff that's definitive. And if you want to sort of play this new game of SEO, which is like the AI overview, the summarization, you really have to offer real value. And then so in this sense, I think AI as kind of a filtering mechanism is actually somewhat healthy for the ecosystem, right? Even though on the other end, I was just describing the slop, but I think it's pretty good at finding out what is slop, you know, on the other end.

Pete Pachal (05:28.073)
So while it's a helpful tool in parts of the content production process, if you're not bringing anything human and unique to the content in the first place, I don't think it's gonna help you much.

Jeffro (05:40.437)
You always have to have a human in there to at least review it before posting and yeah, add something original because if it is just pulling from other sources, then we also need to talk about the copyright issues and privacy. And even let's say you do create your own content. Do you need to protect your own content from the AI kind of summarizing it and putting it out there as its own?

Pete Pachal (05:52.501)
Totally.

Pete Pachal (06:02.217)
Right. Yeah, it's a huge deal. the thing about what we were just talking about, if your content is summarized and it's put in one of these sort of overviews or whatever, people interacting with that generally look at it and leave, right? They're not clicking through. They're not converting, so to speak. So what is really the value of it if that's happening now? There are it's early days again, but it looks like

when Google puts out some of these AI overviews, these summaries at the top of search pages, that the people that do click are really engaged, that they're obviously the people who want to go beyond the summary. So there's that. There's going to be some of that, but it's going to be a lot less than what you used to have. So I think one of the things that AI is disrupting via all this summarization is

how commerce has been done on the internet up until now. So you usually have sites like whether it's the wire cutter or what have you, talk about what products are good to buy. You go there, you click on a transaction link and then there's a commission that goes to the site and other people sort of in the process, right? Well now with AI and it's just giving you a recommendation, that's kind of severed to some extent. Now, what I think is gonna happen is like there's going to be

some standard version of that within AI summaries, which is just going to be a much narrower collection of links, right? So it's going to be a lot more competitive or really less competitive if you think about it. Like in other words, like I think only a couple of sites will end up surviving in terms of these recommendation engines. And then they'll sort of end up putting the link through, right? Cause you know, whatever there's money to be made, someone's going to figure out a technology to make it.

So I think that's going to happen, but again, it's early days right

Jeffro (08:00.468)
Yeah, that's interesting because the affiliate marketing industry has been really big. think it's been shrinking and stuff like this is going to make it shrink further to your point. Cause if you don't have the volume, you know, or the, history to compete against these top sites, then there's, you have no chance.

Pete Pachal (08:10.217)
Right, exactly.

Pete Pachal (08:20.417)
Well, also Google has been being, has been recently merciless about people abusing their SEO domain authority a little separate from AI, but they're trying to get at this idea of like, you're say people magazine and you have a really strong URL and a lot of SEO clout. Then you can just start reviewing air purifiers or putting up air purifier reviews and getting some of that commerce. it just, that doesn't seem right. They're like, so they've kind of tried to.

cut back on that. what that means is now there's fewer sites with all this actually having reviews of this stuff up. I think that's generally might be a good thing. It's like, let's make sure the people who are real specialists in this and have real unique and hands-on insights to offer are highlighted. so again, in some of these ways, think AI can be kind of healthy because arguably it's a little more discerning than just sort of a Google search.

Jeffro (09:21.138)
Yeah, it also increases the barrier to entry for people though. Like you said, you have to be more committed doing this for longer time and be an actual expert before you can go around touting stuff or selling people on things. I mean, you still can to a certain extent, but at least not through this mechanism, right? It's going to be harder for you to fake that.

Pete Pachal (09:40.351)
Yeah, you'd have to stick to TikTok probably. It's all about the incentives of the platform and the rewards of them. And I'm sure TikTok's going to have its own shakeout at some point because it's kind of become TikTok shop. yeah, we'll see how long that lasts.

Jeffro (09:43.211)
Yeah.

Jeffro (09:58.301)
Yeah, well do like how on the Google AI summary and now in chat GPT-2, I'm seeing more of the citations. When it is pulling from a direct source, it at least gives you a link. So still not perfect, but it's at least trying now. And I have a feeling that'll just get better.

Pete Pachal (10:15.049)
Right. yeah, AI-powered search is essentially what it's slowly evolving into, which is great. Because previously, well, it still is. So ChatGPT obviously goes into its knowledge base when you ask it questions and gives you summaries of things. And it could always browse the web in a limited way. But now, in the last, I think it's the last couple of months, they've released generally their search function. So it can actually actively search the web.

Now this is great for a couple of reasons. One, it's probably less hallucination prone. when it goes into its knowledge base to give you something about say like the French Revolution or whatever, it's looking at the entire world's knowledge on that. Some of this stuff's gonna be contradictory and it's mixing in a lot of other stuff too. this is where like an AI tends to be generally correct about something, but it gets things wrong like dates or names and you gotta go check those out yourself.

When it's a more general web search on a specific topic, particularly for news, there isn't usually as much stuff out there to get confused by. And the narratives tend to go in a certain direction. And also, it's a language task. It's not going into its knowledge base anymore. So it's dependent on the accuracy of the information that it's looking at. But generally, doesn't hallucinate as much. So that's helpful. The other part is that

then now new sources can be cited, like you said, so there's that, there's that, again, there's not really a lot of traffic because of the summarization nature of it, but what is slowly emerging now is this new economy around it. Cause open AI is pursuing licensing deals with all these publishers, perplexity, which also does pay a powered search. Obviously they're doing something a little different. It's a rev share. So there's these ways that people are trying to start the engine on making sure.

Jeffro (11:43.432)
Mm-hmm.

Pete Pachal (12:09.281)
content creators get paid. Because that's the big question here, right? Which is like, is it actually okay for AI engines to go out and crawl and index the web and ingest all this stuff? The courts haven't really firmly come down on one side or the other. Everyone's looking again at sort of web 2.0 precedent, which said it was okay, but that was back when you got traffic. So now that that bargain is broken, is it okay anymore? We don't know yet. We'll see. The New York Times is suing OpenAI over this.

News Corp is suing perplexity over it. It's a big, big deal. So one of the downsides of the licensing question, like OpenAI and others have proactively gone out and tried to license content, is that you only will get news from organizations of a certain size that are essentially worth going out and making a deal with. So if you're a smaller publisher, you're independent publisher, if you're an individual,

They're probably not going to answer your emails about having a deal with you. So I think that's unfortunate because one, you either won't get that information in the summary or the search because these publishers will block their content, or they will be there and they simply won't be compensated for it. Now there's a number of startups out there right now that are trying to create essentially an economy around this.

So what you would do, you'd put your content on a marketplace and then in a sort of large scalable self-serve fashion, like an AI company could just say, I want to buy all of that content or portions of it from this particular vendor, this middleman, right? And there's a few companies that are doing that now. I'm hopeful for that because as an independent individual contributor, I want to see that sort of thing succeed, but it's...

Jeffro (13:54.544)
Mm-hmm.

Pete Pachal (14:06.981)
I think they still have to sort of figure out how this is going to shake out on the high level before that really gets going.

Jeffro (14:13.607)
Right. Cause I mean, but that still sounds like it's a gatekeeper, right? It might be a different gatekeeper that you have to follow this marketplace's rules to get your stuff through. I'm wondering if there's a better, more scalable way to do that, that, you know, someone's going to come up with, and it could be as simple as dot org and adding additional schema so that you can pass through your affiliate, relationships or, you know, proprietary, stuff that you've, if you've got a trademark or a patent on this,

Pete Pachal (14:36.874)
Right.

Jeffro (14:42.576)
highlighting that so that AI doesn't just say, yeah, you can do this. Anybody can, right? And so there might be ways to do that that just, okay, I can update my website and now AI knows.

Pete Pachal (14:54.613)
Yeah, yeah, exactly. So that is kind of what some of these places are sort of trying to create, because it's less about the individual deals and more about something that's self-serve, that you can just go in and sort of mark your stuff, and you don't really even have to get on the phone with anybody or whatever. So yeah, so once you, one of these startups, for example, is, I think it's called Tolbit, and it's essentially a really quick upgrade to your site, whether it's a WordPress site or whatever.

And then it sort of marks the content in the sort of fashion that you are describing. And then like, if a crawler comes along, there's sort of this automatic thing that happens. And again, you'd have to sort of just make sure you're connected to the toll gatekeeper, so to speak. But it should be scalable. There's no reason it couldn't be.

Jeffro (15:46.723)
Yeah, and hopefully, know, SEO agencies will start to adopt some of these platforms and technologies and just kind of bring that under their wing so that a small business owner doesn't need to learn all this stuff, but they can just have their team do it and know that it's going to be taken care of.

Pete Pachal (16:03.263)
Yeah, exactly. And the whole idea of like SEO for chat bots is very new. You know, I don't think that anyone really knows what is the formula for getting in a chat bot summary. The best definition I've heard is that it's definitive. We get something that is unique that does, you basically I would define it as like if you were to write the ultimate summary, the encyclopedia Britannical on this topic.

Your information is something they couldn't leave out. So in other words, it's going to encourage more unique and hopefully human written takes on things. And with that, it gives me a little hope. It's like, OK, that was something we used to call good journalism when you have something new and fresh, whether it's new information or a scoop of insight or whatever. So if AI alters the incentives that that is what it values.

Hopefully we'll see less of what I would consider sort of the aggregated junk of the past, you know, couple of a class decade or so where, you know, I used to work at Mashable and I'm certainly guilty of this of writing my own version of a Twitter react story, you know, like somebody did something in the news and a bunch of people on Twitter said something about it. And I, if I ever told myself I was doing actual journalism there, I'll tell you this, I was wrong. That is not, that's what I would consider a virus.

in terms of a piece of content. So if AI disincentivizes that kind of quick hit, lazy aggregation style of content and all so much of that.

Jeffro (17:44.931)
Yeah. And I'm all for that. I've always been pretty much optimistic about new technologies and even AI. I know a lot of people can get freaked out like, no, AI is going to ruin everything. But I think in the, as long as we have smart people working on this, it's going to get better. Like, yes, it might be messy in the short term, but in the long term, like you said, it's actually going to drive higher quality content and force people to be better and not just, you know, race to the bottom.

this lowest common denominator where you can, who can spit out the cheapest, fastest stuff, it's gonna actually get better. So that's good.

Pete Pachal (18:19.743)
Yeah, yeah, I would say that the, you know, not to throw cold water on anything, but, you know, there's, there's the other side of it, which is that there's what I would say is a fine line between use of AI and abuse of AI, right? And that's kind of where things can go wrong. Cause you know, as a user of it, you know, in getting summaries, you may trust it too much and then rely on something that was either.

even if it's correct, maybe not the best information for whatever reason. And certainly if you're creating content, you know, it's, it can easily become a crutch, even though it can, has a lot to offer in terms of efficiencies, a content creation process. like whether it's giving you ideas for stories, whether it's, you know, enabling you to do a kind of content you couldn't do before, like great example, this is Google notebook LM, which essentially creates podcasts out of

documents or even just ideas. That's really interesting. And, you know, or just the distribution of it where you're promoting your stuff on social or optimizing it for SEO. AI can offer all of these stages of content, like a lot you can use, but once you, you aren't, once, once you get to just sort of an undisciplined use of it and you're just kind of like using it to write your stories and do all the, like, and you're kind of like not even really involved.

That can be, that's going to result, that's going to get you to the slop place, right? Which hopefully, like I said before, the filters and the incentives will discourage.

Jeffro (19:56.942)
Well, that's kind of a good point is we can move into like looking at imagery, right? Because that line between real and AI generated stuff is getting harder to detect. But there's always been complaints around like copyrights and how it's just stealing people's art. But what are your thoughts on the implications of this when it comes to small business marketing and just the internet in general?

Pete Pachal (20:19.977)
Yeah, the AI image question is probably illustrates the economic problem more than any other content. Because once you have these AI image engines, which are super common now, have mid-journey, Dolly, Meta even now, and they can create images to represent your story or your concept immediately, like within seconds, what hope is there for stock imagery?

anymore, you know, like, I mean, people who, like, as soon as I did my first dolly image is like, why would I ever use Getty again? Well, obviously, for news purposes, there's a lot of reasons to use Getty. But for a lot of time, like a lot of my journalism career, you're often hunting for something to illustrate an abstract concept. Like you mentioned at the beginning, I worked at CoinDesk. One of the big problems with doing markets coverage is there's a lot, you know, to never go up, never go down. Like how many graphs can you put out there?

Jeffro (21:16.588)
Yeah.

Pete Pachal (21:17.249)
And you just need images. like, hey, AI is great for generating these sort of abstract images. And yet you sort of go like, well, wait a minute, we used to use stock imagery for that. What happens to them? What about using images in the style of an artist? That doesn't seem quite right, but you can easily do that. I think they're starting to put safety safeguards on things. using imagery, it's a bit of a dilemma still now. And I think every business needs to come down with some kind of AI policy.

that fulfills that need for imagery, but also has an eye to the long-term and to being fair, right? So there's various AI resources, AI-powered resources that do think about this and try to come out and be ethical about it. So I'm hesitant to recommend them, there's a couple of, Adobe has famously sort of come out and-

which said that it's going to respect content rights and make sure it's not going to train on its own users' images, but also that the rights that it's purchasing from stock imagery are going to include this. they're trying, right? and also its Firefly models are not trained on public domain data. That's the big thing. That's what sort of a lot of these image companies

Jeffro (22:16.482)
Mm-hmm.

Pete Pachal (22:43.409)
whether it's mid-journey or whatever, have clearly done. They clearly just harvested tons of images that are out there publicly available and created their models on that. So as you think about what you're going to use, you might think about which service you're going to use, but also how you're going to use it. Use it in these kinds of articles, but maybe not these. I know places, publications that allow some limited use of generative imagery.

Jeffro (22:49.334)
Mm-hmm.

Pete Pachal (23:12.971)
forbid asking for the style of a particular artist or photographer. That's off limits. And that seems kind of like a reasonable compromise to me. Because again, the big factor you need to also take into account is if you're going to take a stand from any kind of ethical standpoint, you're going have to understand there are going to be competitors who don't. And that's not an excuse for you to do whatever you want, but it is sort of a factor you have to sort of think about.

as you're doing the stuff. Just full confession, I use generative imagery in my newsletter and various other places. Again, I'm a one man show mostly, I'm a solopreneur and again, that's not an excuse, but it is kind of a, if I were to use imagery to illustrate a news event, I would probably wanna go to some legit source like Getty or something like that, right? Cause you kinda don't wanna get into this zone where you don't.

Jeffro (24:06.07)
Right.

Pete Pachal (24:11.509)
know, like as you said at the beginning, whether it's real or not. that's a very dicey line to cross. And this started to happen with some wartime coverage. I think it was with the Israel-Gaza war early on. And some people were publishing stories, but they would use, I don't think the people publishing the stories even knew that the image was fully generative. I don't know. I might be misremembering. But it was sort of like more generic wartime imagery. And that's a really tough one.

Jeffro (24:37.014)
Mm-hmm.

Pete Pachal (24:39.649)
I think there's a strong case to be made that if you're doing wartime reporting, unless you're, there's some, I can't think of what it would be, but some big extemporaneous circumstance, I think you'd want to stick with real imagery or just don't use anything.

Jeffro (24:57.953)
Yeah. Well, this is something that, you know, people didn't have to think about before, right? You used to just slap something on there, go to one of the free stock photo sites. But now, yeah, there's all these implications that you got to worry about. So hopefully, yeah.

Pete Pachal (25:10.421)
Well, I'm old enough to remember the early days of the internet where a lot of people would just slap anything on copyright or not. And that quickly became like, you you can't just grab a photo from Google image search. Like that's somebody's photo. that obviously that's the early days of the internet. that did, you know, it takes, it takes a while to get sort of the publicly educated about this. then I think.

Jeffro (25:16.769)
Yeah.

Jeffro (25:32.065)
And that still happens, even at large organizations. They might have an intern or somebody in there who's supposed to publish an article and they, you know, grabbed a photo from Associated Press or something. then a couple of months later, they see an email with like a lawsuit or something about the copyright. They take it down. Yeah.

Pete Pachal (25:46.881)
Yeah, well, it can be quite expensive for each instance. I think I used to be chief of staff at CoinDesk. I would get those emails. usually, if it goes to court, it could be as much as $10,000, for instance.

Jeffro (26:00.565)
Yeah, some of those law firms, it depends on the way they handle it. Some of them are just like trolling, I think, how to get people, which is not fun. But obviously they have legitimate stuff too.

Pete Pachal (26:11.333)
agreed. Yeah. I had to deal with those. There are legit problems and like issues and then you have to sort of figure out the right solution. But there are definitely folks who just troll. I've had emails about stuff that was clearly creative comments, for example. And someone would try to shake us down for it. There's a lot of automated services out there. You got to be discerning. some of them are legit.

Jeffro (26:36.82)
Well, thanks for joining me today, Pete. We're near the end of our time. AI is always fascinating. I appreciate your insights. For those of you listening, if you need an AI training consultant, Pete's links will be in the show notes and he will give you a discount if you use the code THINKAI. I have one last question for you, Pete. You know, most of what we talk about is all the generative stuff as large language models. At some point, inevitably, we're going to actually have real artificial intelligence.

What are your thoughts on either how far away that is or what's the change that's going to come as a result of that?

Pete Pachal (27:14.699)
Good question. Okay, so I am not an insider in Silicon Valley or OpenAI or any of these places. So I don't know what they got in their Skunk Works. That said, from what I have seen and where things are going, I do not see what people are fearing as either general intelligence or super intelligence happening anytime soon. I feel like there is no spark there. There is no thing that is going to instigate

an intent, you know, that's the thing about AI. There was a brilliant essay in the New Yorker a couple of months ago by guy named Ted Chang, an artist. And he talked about how AI doesn't have intent. And that's one of the reasons it's not ever going to be valued as much as human content. I thought that was a very insightful comment. so, so my view is that, that that's really far out. Like we're talking like sentient computers. It's just, I just don't see it happening. That said,

these things, these agents and the autonomous systems that they're slowly becoming can still be dangerous and can still be quite capable of a lot of things. So it's not a reason to sort of not be concerned about that kind of thing. so I put probably more faith than most, I know this might sound naive in sort of the tech companies and how they're thinking about this. And this is mostly based on

interviews with folks like, you know, Sam Altman and the CEO of Anthropic. Like they are really looking closely at these levels of automation and what they need to do to ensure safety. But I also think there's a good sort of utopian view of this. It's like, as long as we can make sure the proper safeguards in place, they could solve a lot of big problems, gently by putting them to work on those problems and

enlisting all the help they can possibly get from other AIs mostly to work on them. So I don't think we're going to get super intelligence for probably a century or more. That said, think AI is going to help us solve some really big problems in the next 20 or 30 years.

Jeffro (29:28.703)
Totally agree. Well, thank you again for being here, Pete. Thanks to all you guys for listening. If you thought this was valuable, please leave a review for the show on Apple or Spotify. I look forward to seeing you guys all again next time. So take care and have a good one.