Mike and Paul are back for another episode of The Marketing AI Show.
Paul ran a head-to-head test of AI writing tools and posted about it on Twitter and LinkedIn. On both platforms, the posts gained traction and good feedback from our community and from technology companies. In the experiment, Paul ran the same three prompts/use cases through a handful of different tools. The tools were OpenAI’s ChatGPT and their Playground functionality; Cohere; Jasper; Writer; and HyperWrite. It’s an interesting case study showing AI in action.
Next up, we’ve all seen plenty of examples online, many impressive, of what ChatGPT can do. In addition to writing text, ChatGPT is generating outlines, doing research, answering complex questions, and surfacing information. This has led to some serious online conversations asking the following question: Will ChatGPT replace Google search? Today, plenty of ChatGPT’s responses are inaccurate and don’t contain citations, but some people are already imagining a future where we no longer use a Google search to find information.
The third topic of this week’s podcast is MyHeritage. MyHeritage is a discovery platform that helps you find people related to you. The company just released a free tool called AI Time Machine. You upload photos of yourself, then the AI creates hyper-realistic avatars of you in different historical time periods, from prehistoric times to Ancient Egypt to modern times. It’s a timely twist on hyper-popular photo apps like Lensa that are getting a lot of buzz.
Paul and Mike have a great conversation on this week’s episode. Listen to the podcast below or in your favorite podcast player.
00:02:35 Head-to-Head AI Writing Tools Test
00:21:16 Will Chat GPT Replace Google?
00:30:49 MyHeritage AI Time Machine
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: Did it just open the door for us to create way more value for our audience and community because we can now create the things we couldn't create before
[00:00:08] Paul Roetzer: Welcome to the Marketing AI Show, the podcast that helps your business grow smarter by making artificial intelligence approachable and actionable. You'll hear from top authors, entrepreneurs, researchers, and executives as they share case studies, strategies, and technologies that have the power to transform your business and your career.
[00:00:28] Paul Roetzer: My name is Paul Roetzer. I'm the founder of Marketing AI Institute, and I'm your host.
[00:00:38] Paul Roetzer: welcome to episode 27 of the Marketing AI Show. I'm your host, Paul Roetzer, along with my co-host Mike Kaput, who is our Chief Content Officer at Marketing AI Institute and the co-author of our book, marketing Artificial Intelligence, AI Marketing in the Future of Business.
[00:00:52] Paul Roetzer: What's up, Mike? How's it going? Good. We haven't been in the office this week, so. I haven't seen you for a few days, . Just virtually. Yeah. Yeah. Um, so today's episode is brought to you again by our piloting Air for Marketer series, which is now live and on demand. Mike and I spent a good portion of the last three months of our lives creating that series, and I guess probably the 10 years prior to that, I feel like.
[00:01:17] Paul Roetzer: So, um, yeah, I check out that series. It's, so, it's designed as a step-by-step learning journey for marketers that are trying to understand and adopt ai. Uh, really, you know, starts from the beginning of an intro to, it goes through state of the industry, walks you through how to identify pilot projects, prioritize problems to solve with ai, and then like Mike takes you through this great journey of like advertising and communications and SEO and analytics and all these different areas of marketing where ai.
[00:01:43] Paul Roetzer: Is gonna be prevalent and looks at, you know, use cases technologies. So it's 17 courses. You can take it in a day, it's eight hours, put it on one and a half times speed. You'll get it done in whatever that math comes out to. Five hours, six hours. So you can literally learn everything you need to know about AI and, uh, drive that transformation in your organization career, um, starting now.
[00:02:05] Paul Roetzer: So piloting ai.com, you can learn more about that. It is, like I said, available on demand. You can binge it in a. Uh, so with that, I'm gonna turn it over to Mike and he's gonna hit our three topics. So again, if you've never joined us for our weekly, Mike kind of curates the big news of the week picks three things we're gonna talk about.
[00:02:24] Paul Roetzer: Uh, we chat about those. Sometimes we do a rapid fire at the end. Today, I think we're just gonna probably stick it to the three topics, um, and, uh, and kind of go from there. All right, Mike, it's all. Great.
[00:02:34] Mike Kaput: Thank you. So first up you are again in the news, um, Paul, you, you ran a head-to-head test of. AI writing tools and you posted about it on Twitter and LinkedIn, and what's really cool is not only what we'll get to in the post, but on both platforms, um, your post kind of blew up getting a ton of attention from a lot of AI companies, um, business leaders, executives, and in your experiment you ran.
[00:03:04] Mike Kaput: The same three prompts or use cases through a handful of different tools. And these tools were open AI Chat, G P T, uh, open AI, playground functionality, cohere Jasper, writer and Hyper. Right. So do you wanna tell us a little bit more about this experiment, the results of it, what you were trying to learn
[00:03:26] Paul Roetzer: from it?
[00:03:27] Paul Roetzer: Yeah, so it was sort of one of those things that just happened organically. So I, I don't remember the exact sequence, but I would say like Friday night, you know, I'm laying in bed thinking about. The, the institute for starting point, like, Hey, we're not fully utilizing these tools yet. Like we obviously know all about what's going on in the I writing tool space.
[00:03:46] Paul Roetzer: We talk about it all the time, but we haven't necessarily operationalized what is our plan to either scale up content to enrich the content we're creating. Like we didn't, we don't really have a strategy yet. Per se, we're just experimenting with the tools like many other people, but also understanding the power that they have and that it could be transformative to our business.
[00:04:04] Paul Roetzer: Because largely the institute is a media company. First and foremost, we publish content to educate people about artificial intelligence. So it started with, okay, I have to more deeply understand how these different tools, What's the difference between them? Which ones do we wanna scale our institute with as a starting point, I guess.
[00:04:27] Paul Roetzer: Um, but to do that I had to really understand, you know, what, how they worked and what the differentiation was between them. Um, the other things that we started looking at was more around like, Consulting, you know, people come to us now and ask us for guidance on this. And it's kind of a practice area that's emerging for us is starting to do more work at helping people figure out how to pilot and scale language tools, um, you know, specific area within their organizations.
[00:04:54] Paul Roetzer: Uh, and then it's just an area like. Again, people come to us all the time asking opinions on this stuff, and I felt like we just gotta go deeper on it. And then there was a, there was actually a blog post from Cohere. So if people aren't familiar with Cohere, it's actually like a company that builds language models, so writer, hyper write, Jasper.
[00:05:11] Paul Roetzer: A lot of these other AI writing tools are built on the backbone of G P T three from Open ai. So they use a language model that isn't theirs to basically build an interface and. build the capabilities on top of who here builds their own language model. And so like, Ja, I wouldn't say open AI and coherent necessarily competitors.
[00:05:28] Paul Roetzer: I, I'm not sure that I would consider them that, but they do, they're more like the platform company that's building these capabilities and then people build on top of it. So I thought, well, if I can go into Open AI and play around with it in the playground and then chat G P T, if I could go into CO here and play around with it, then I, I can maybe understand it.
[00:05:44] Paul Roetzer: So I was reading this article from CO here about how the language model. How you can set these different parameters to determine how creative the outputs are gonna be. And I was like, okay, this is kind of cool. Like I, I'm starting to really get a deeper understanding of it. Let me just go in and experiment.
[00:05:58] Paul Roetzer: So I think I messaged, I don't remember at what point I messaged you, Mike, but it was, I think it was Saturday morning. . And you know, going back to 2015, Mike and I did this project at my agency called Project Copy Scale. And it was an internal initiative to say like, basically how do we scale up content creation?
[00:06:15] Paul Roetzer: And back then the question was like, can AI help us do that? Can it write content for us? And so we learned back in 2015, 16 was no, it, it actually can't. We were not at the point where AI could really write content for us unless we programmed it specifically to do certain types of things like analytics reports.
[00:06:33] Paul Roetzer: I messaged Mike and I said, I, I think it's time for copy scale 2.0, like large language model edition. Like we, we need to revisit this idea from seven years ago and say, where are we at now? We know the answer that it can actually create this content. Now what, what does that mean? What's kind of the bigger picture?
[00:06:51] Paul Roetzer: So, yeah, so I, I, I don't know, it just like a Saturday morning, afternoon dive in a few hours. I, I just randomly come up with three prompts. One of them I needed to create an, um, ad copy for Amazon for our book. So I was r I was launching a campaign for our book on Amazon, uh, under their sponsored products.
[00:07:06] Paul Roetzer: And I was like, well, I could write this, or I could see if they, you know, these tools can write it. So that was one of them. I did, uh, an outline for an ebook. So we're working on a series of AI blueprints. And so I picked one of those and said, create me an outline for AI for this. And gave a few other parameters on the prompt and saw what it did.
[00:07:24] Paul Roetzer: And the other one I was playing with was, uh, um, create an agenda for marketing AI conference. So I was like very different prompts. Like it wasn't just write me a blog post on this. I was trying to like, see what it's actually capable of and how it output. And so that was really it. It, and I didn't know what I was gonna do with it.
[00:07:39] Paul Roetzer: I was honestly, Experimenting and then I did it and then I like took a break and stepped away and I went upstairs and was hanging out with my family and stuff. And I was like, what did I just learn? Like, I don't even know, I'm not even sure what I just experienced, but I just started jotting it down what my initial reactions were to what I had learned.
[00:07:57] Paul Roetzer: And that was what led to the LinkedIn post, which maybe we could, you know, we could kind of cover a few of the key findings I, I noted in the Twitter feed and the LinkedIn post. Um, but that was really it. It was just more of the spur of the moment. I, I had the urge to like figure this out and so I dove in and started playing around with it and in that process arrived.
[00:08:17] Paul Roetzer: There's like five or six quick points that I wouldn't say are overly profound, but they obviously sort, sort of hit, hit, hit with home with a lot of people because it did lead to a lot of responses and comments and engagement, which was great. And it kind of like helped me form my own thoughts a little bit more.
[00:08:37] Mike Kaput: Yeah, I thought it was really interesting. It really did seem to resonate with a lot of people in a lot of different areas of business, and I think that's just because everyone is trying to figure this out as well. And I think that, you know, in, we may be somewhat close to it sometimes and think certain points are obvious, but I don't know how obvious they are sometimes too.
[00:08:59] Mike Kaput: Everyone just learning about this stuff, you know? A week or two ago when, when we, when Chad j p t came out. Came out. Well
[00:09:08] Paul Roetzer: it is a lot of, you know, I saw, I think one of my friends said something like, you know, there's a lot of kind of like AI amateurs stating opinions about what's happening in ai. And I don't necessarily think that's a bad thing.
[00:09:22] Paul Roetzer: Like I do think we just, it's important for all anyone who has an experience or an insight to share it. Um, but I think the point he was trying to make is like, there's a lot of people. Commenting in an authoritative way about a topic that they didn't know anything about till two weeks ago, . And so you have all this like mass race for everyone to have a point of view all of a sudden on AI when they may have no fundamental knowledge of what a large language model is, how it works, the history of them.
[00:09:52] Paul Roetzer: and I think it's important, and maybe that's part of the, hopefully the value we're creating through these podcast episodes is to maybe give a little more context as to what is going on right now and why it matters and what could be coming next. And I think that's what a lot of people are missing in the commentary they're probably seeing online.
[00:10:09] Paul Roetzer: And what I was kind of trying to like restate in the LinkedIn post that got turned into a Twitter thread.
[00:10:16] Mike Kaput: So one thing that really jumped out to me is, Uh, the comment you or takeaway you had about the chat G p T style interface is going to become the default interface for all writing solutions. Having to click around and test a bunch of templates already feels outdated.
[00:10:33] Mike Kaput: I just wanna ask a question or prompt in action and get an output. Can you tell me a little bit more about that
[00:10:40] Paul Roetzer: thought? Yeah, I think I've used the analogy of like bion search versus Google search as like an example. Like with bullion, you had to know what you were asking. There were certain inputs that needed like quotation marks and symbols and things to like, so to get the results you wanted, it took effort.
[00:10:59] Paul Roetzer: Like there was just work that had to go into it. And if you've used any of these AI writing tools, and again, many of them are great. Like we, we use 'em in our business. Some of them are partners of the institute and they're, it, it's awesome. Well, what happened I think though, is like a lot of them. Were built, almost all of them that I can think of actually are built where there's like these templates or pre-trained models and it's like, okay, I want my case.
[00:11:19] Paul Roetzer: I want an Amazon ad. So the first thing I do is I go into their template library and say, is there a template for Amazon ads? No. Okay, there's a Facebook ad template. All right, maybe that'll get me where I want to go. So I go in, I click into the Facebook ad. This is the exact thing that happened. By the way, on Saturday I was doing this
[00:11:38] Paul Roetzer: I go into the Facebook ad thing, then it wants like, what's the product? It's like, okay, fine. Like I put that information in and then it would ask for like a description or something of what you wanted. And it's like, okay, now I'm having to think like, I don't know, like that, that's kind of why I'm doing this is I don't want to have to think about what I'm gonna say.
[00:11:54] Paul Roetzer: So like there was effort to, to put these inputs in to get the output I wanted. Mm-hmm. . And it felt like there was friction at that point. Like I had used chat G B T many times already. I knew. That in chat G P T. Once I got of the, you know, again, six, I was evaluating. Once I got to testing chat G p T for this function I was simply going to put into it, I would like an Amazon ad, 200 characters or less, whatever that is about the marketing artificial intelligence book done, go generate.
[00:12:26] Paul Roetzer: There was no other inputs. I didn't have to like go search for information and write the draft for it and then have it re write. So in that moment I was like, wow. Yeah, I can't imagine a world in the future where I'm gonna have to like filter through libraries and templates and like, cuz eventually you're gonna have dozens or hundreds of templates.
[00:12:46] Paul Roetzer: Like they all have dozens already. It's gonna be in the hundreds. And when I know that Chad g p t already has the capability for me to just tell it, write a blog post, write an ad for Facebook, write, I just tell it what I want. Mm-hmm. and then I give it as little information as I want. I can expand that information if I'd.
[00:13:02] Paul Roetzer: But it's still going to produce an output for me if I don't give it a lot. Right. Amazon ad about the marketing artificial intelligence book by Mike and Paul, whatever. Yep. And it's gonna come back with something and then it's like, well that was, that's okay. All right. Let me just finish. Right. And I may take it and tweak it and maybe like 50% of it I'm gonna write myself, but at least I got to the output in three seconds instead of three minutes.
[00:13:25] Paul Roetzer: And it might not seem like a massive thing to like have to spend three minutes putting a bunch of inputs into the prompt, but it feels. Like archaic after you've used, just give it the prompt, like chat, G b T style. So that was my first takeaway is like, wow, all of these tools are gonna have to eliminate the friction completely.
[00:13:45] Paul Roetzer: Like people aren't going to be willing to do that much work to get an output when they know you can just give a voice or text prompt and get exactly what you want.
[00:13:55] Mike Kaput: Right. So you're saying that. Likely going to see all of these AI writing solutions gravitate towards a more chat G P T, like conversational interface versus templates, doing outlining description and prompting it more strictly in the way you
[00:14:11] Paul Roetzer: described.
[00:14:12] Paul Roetzer: Yeah. And it's already happening like hyper right? Came out with theirs. Yeah. I think last week Jasper came out with theirs today, so we're recording this on December 20th. Jasper announced this morning Jasper chat, so they know it. They felt. They feel it too. They live in the product. As soon as you had chat at G P T as like an interface, it's like, oh, okay, this is different.
[00:14:31] Paul Roetzer: This is simpler. So I mean, again, I'm not saying templates are obs deleted completely and like what anybody's built doesn't matter anymore. It's just one person's opinion. But the difference seems so dramatic that I can't imagine the user experience going back to that, like having to search. It's almost like, it's almost like.
[00:14:53] Paul Roetzer: like we use Google Docs like Google Drive. Yeah. Maybe like going in and having to click through seven folders to find the doc I wanted, instead of just typing in a keyword from the doc and having the doc show up like those seven clicks drive me insane. It may only take me 12 seconds, but I don't want to have to click through folders.
[00:15:11] Paul Roetzer: I just want to tell. I just want this doc, and that's the doc that shows up. And I think that's the difference is like people are, are used to immediate gratification when you can give something a prompt or tell it what you want. And I think that's just inevitable where this is gonna go very quickly.
[00:15:28] Mike Kaput: That's interesting. And you know, that kind of begs the question as you talked about in, uh, your post and in the thread. You know, you said if you lead a content team, a content agency, or a media company, these AI writing tools will transform your staffing operations and production. You have to prioritize defining your roadmap in early 2023.
[00:15:51] Mike Kaput: What kind of questions do we. Heads of content, leaders of content teams, businesses that rely heavily on content should be asking,
[00:16:02] Paul Roetzer: well, I know what they are asking cause I've had like three companies tell me this, this week already. Uh, do we need all the writers we have on staff? Hmm. Like to the immediate assumption when you see it, either hear about what chap g p d does see a demo of it, or see like a clip on social or try it yourself.
[00:16:19] Paul Roetzer: Mm. I totally understand, sympathize with the immediate reaction being, wow, we don't need our 20 writers anymore. C couldn't we use five and get the same level of output or better or more? And it's a valid question. Like I don't, I don't, um, challenge anyone for, for having that thought in their head. Um, I don't think it's the right thought, but I, I understand why it would be the first thought.
[00:16:44] Paul Roetzer: So I think the. The organizations that develop a deeper understanding of how this stuff works, what the limitations of it are, what it's capable of outputting what like, like we were doing with our own project copy scale internally. We only have like the two of us are the main writers. Um, and largely you're the writer.
[00:17:06] Paul Roetzer: Um, but we're the content creators. Mm-hmm. , I'm not sitting here saying, well, how do we reduce the need for both of us to write content? Right. I'm looking at saying, wow, like what are all the things we've wanted to create? Over the last six years that we've never found the time and energy to create, how much more efficiently could we do that now?
[00:17:26] Paul Roetzer: Like, is did it just open the door for us to create way more value for our audience and community because we can now create the things we couldn't create before we had, we had this, um, recently you and I were collaborating on a internal project for, for a blueprint, like an ebook, a digital book, and we went through our usual process.
[00:17:45] Paul Roetzer: Build our own outline, did our own reason, whatever. And then at the end it's like, all right, let's throw this into Jasper. Let's throw it into writer. Let's throw it into, you know, open ai, play, whatever, and let's see what it would have created. So we did our thing first. Then we went through and said, are we missing anything?
[00:18:01] Paul Roetzer: And you go through, you create these output, it's like maybe there was something in there, maybe there wasn't. But they were, they were interesting. They were good and they were really, really fast. And so I could definitely see for us, you know, the other point I I make in here is, you know, I think good to great writers are just gonna find ways to let AI enhance the process.
[00:18:20] Paul Roetzer: Outlines ideation, inspiration. Average writers are not good writers are gonna use it to write the drafts. Yep. So I would put you and I probably in the good to to great writer category, like certainly more advanced writers than the average marketer. It's what we do for a living. Both came out with that background.
[00:18:38] Paul Roetzer: I don't see AI replacing my process at all for writing. Mm-hmm. , but ideating more and more I could totally see that. Like, I wanna write a post about this, draft me an outline, or Right. Gimme a list of the key points I should address in this post I'm writing, or a script for a podcast, or like whatever it is.
[00:18:57] Paul Roetzer: I could absolutely see using it in that sense. And then maybe have it. Assess my language, like simplify it now, or, or like make this a seventh grade level. Cause I tend to write at a more like, higher level and I, I shouldn't, like, I should simplify my, my language a lot and I'm not good at it. So it's like, okay, let AI do that cause it's good at it.
[00:19:16] Paul Roetzer: Um, so I think what you have to do is you have to look at your team, your organization. You have to look at the capabilities of the people on that team and say, how are we going to use AI writing tools for our team? Uh, do. 50% of them are average writers, and this is going to make them good to great writers.
[00:19:32] Paul Roetzer: Then the adoption for them is gonna be different. Do we. Five amazing editors, they're gonna use the tool totally differently than a writer would. So what all I'm saying is people need a plan. They need to understand what the tech is, how it works, what it's capable of today, what it probably is gonna be capable of by the middle of 2023.
[00:19:51] Paul Roetzer: And they need to seriously think about their operations, their staffing, and their production, just like we're doing, um, at a smaller scale. So that was, that was my main point, is like, Don't wait, like start thinking about this now, I'm not saying you're gonna change your whole thing around next year and you know, switch up your staff and have to reskill everybody in three months, but you better be at least thinking about it in that time period.
[00:20:14] Mike Kaput: Yeah, and I think we've seen in the last month with the developments of tools like chat, G B T, whereas in the past, you and I might have said, Same advice, you know, a year or two ago, this coming year is when your executive team is going to start asking about that
[00:20:32] Paul Roetzer: a hundred percent. I've gotten that call bunch by, Hey, our board said this.
[00:20:36] Paul Roetzer: My c e o wants that, um, from people who didn't three weeks ago care about ai. Now, now it's like, it's the thing they don to solve for and they don't know where to go and they don't even have, maybe don't even have someone internally to turn to, like they're not even sure who on their. Has any knowledge of this or can figure it out.
[00:20:53] Paul Roetzer: But yeah, I agree. I think, and we, I think we talked about this on a previous episode, chat, G P T made AI awareness a mainstream thing. Now comprehension we're a ways off from adoption. Were even further off from, but it made it top of mind for a lot of organizations, stations. So
[00:21:12] Mike Kaput: pivoting slightly still chat, G P T related, but I wanted to unpack maybe some of.
[00:21:19] Mike Kaput: Bigger ripple effects of this technology. So, for example, there's a lot of examples online of what Chat G P T can do, and a lot of it that we've seen is really impressive. You know, there's writing of all, people are writing all different types of text and generating all types of content, but in addition, they're also doing things like generating outlines, like you mentioned, doing research, answering complex questions, surfacing information.
[00:21:49] Mike Kaput: While it's not always all accurate today, some of it's been really, really impressive in my opinion of what it's been able to produce, and this has led to some serious online conversation asking the following question, will chat G p t or something like it replace Google search because. Today, like, you know, you're not getting accurate information with citations, but you look at what this tool can do and you can start imagining a future where instead of me going and hunting down information via search engine, why can't chat G P T or something similar?
[00:22:25] Mike Kaput: Serve me up the answer to my question immediately, and if I wanted to give me citations links to go verify the information's correct, I mean, you really could imagine a potential future that comes to pass where you are using a chat G p t like interface to do everything you would for a search engine and in my mind, Gaming that out.
[00:22:47] Mike Kaput: It seems like the implications to search would be pretty immense. Like if that future came to pass, you wouldn't really be spending a bunch of time clicking through to content on websites. You wouldn't have that much reason to say, go to a blog, go consume a piece of content designed to attract U V s search.
[00:23:07] Mike Kaput: Um, It would've a huge impact on content marketing and seo. I mean organic traffic to your website based on keywords, back links that would no longer be nearly as relevant in a world where this stuff was integrated into a search product or replaced one. So I kind of wanted to get your take given some of the chatter I've seen in the space.
[00:23:25] Mike Kaput: I mean, do you think that's a possibility? What do you think the odds are that a chat G P T or something like it replaces or fundamentally
[00:23:34] Paul Roetzer: changes surge? I mean, I, I don't see how it doesn't have an effect on the future of search, but I think we touched a little bit on this concept when, you know, episode or two ago when we first talked about chat, G P T I I, I, I wouldn't bet against Google.
[00:23:52] Paul Roetzer: It was like my main feeling right now. I don't know, like, I, I mean, I've talked to some people and tried to gather some opinions I've followed, like Margaret Mitchell was, you know, had a, a good thread on it. I saw another one this morning related to. You know how it's unrealistic that search moves to this.
[00:24:09] Paul Roetzer: Um, beyond the reasons Margaret gives, which we can touch on, I is the, the, um, the energy consumption and the cost. Like, so the cost, I, I don't know if they've publicly shared yet. I've only seen like guesses and Sam Altman himself said it was, I think close to like, Under 10 cents, but per output. Um, but still a significant cost for each query you put in or each prompt you give Chad g p t.
[00:24:34] Paul Roetzer: It costs a lot of money, relatively speaking, to generate those responses. Yep. So I've seen some estimates that it's like $3 million a day that open eyes burning on, on these, you know, free searches basically, or free prompts. So because the way the model works at the moment, it seems very cost prohibitive that something like chat, G P T would just all of a sudden replace search and then the points Margaret makes around page rank.
[00:25:03] Paul Roetzer: So hers is specifically around, you know, the history of search used to just be. And history searched like the late nineties, um, when Google was formed mm-hmm. , um, that it was just like I searched for this, these words, this page appears to have those words in the closest sequence or order. So it seems like it's a match to what I was searching.
[00:25:23] Paul Roetzer: And then Larry and SEI come along and, and they basically build Google. And we tell this story actually in our book, um, based on citations being the key thing. And so if, if all of these pages linked to that page and that page probably has more authority. And so her point was like, it's really hard to replicate.
[00:25:42] Paul Roetzer: That, that val value proposition of citations and paid rank. Right, because that was my, I read her thread three times and I still wasn't a hundred percent sure what, what exactly the argument was, so I'm like, I probably should put that in context, but it, it seemed like that was the argument she was primarily making is that pay, rank and citation still matters for you to trust the result and chat.
[00:26:06] Paul Roetzer: G p T doesn't have that. There is no citation. There is no. Verification of fact through any transparent means. It's just a prediction of what the words in the paragraph or sentence should be based on its learning data, which we don't know. So yeah, I think there's technical obstacles. I think there's energy and cost obstacles, and I think there's an obstacle of Google isn't stupid and they likely have more powerful technology.
[00:26:37] Paul Roetzer: Sitting in their research labs or maybe sitting in within Google search already. Like we just don't even see it. But I, I just can't fathom that Google doesn't have an action plan right now.
[00:26:50] Mike Kaput: Yeah, you would have to think. And what you were referencing, a thread from AI researcher Margaret Mitchell, who is breaking down kind of why.
[00:26:58] Mike Kaput: Today chat, G P t can't really replace Google search. And yeah, to your point, I mean I think there's a lot to unpack with what she said. From my understanding, she basically is saying that G P T Chat, G P t is in a stage that's similar to the early days of web search. Yes, it can give a lot of information.
[00:27:15] Mike Kaput: No, there is not a great match between what you want and useful or reliable results. Now she thinks eventually we will get there, but it's going to take a fundamentally different approach to. Chat, G P T or something like it is trained. So presumably either something like that or incorporating that idea into potentially search is what Google is working on.
[00:27:39] Mike Kaput: But it would be interesting to see. They can't have gotten, I would imagine caught blindsided. Yeah. By
[00:27:45] Paul Roetzer: this. Well, and, and we've talked about this before, but I also go back to. Like, so chat, G P T learns from the corpus of knowledge on the internet. Mm-hmm. corpus of knowledge on the internet was created in large part because there's incentives for individuals, individual writers and publishers and brands to create content that people find organically to arrive at their site.
[00:28:07] Paul Roetzer: Tried to write site traffic. So if Google doesn't exist, search doesn't exist in its current form. The motivation. To create content that drives people to click on a link to land on your website is almost gone. Mm. Which then removes the training data that enables chat G P two to exist , so like chat, G P T exists because the internet exists in the format does because search exists in the form it has for the last two decades.
[00:28:38] Paul Roetzer: So that's, I think I just can't wrap my brain yet around. All the dominoes here, right? That there's just, but at the same time, I also. Look at what we've now know after the last few weeks and say the future looks to anything like the past. . . I just don't know what it looks like.
[00:28:56] Mike Kaput: Yeah. . I, I think that's the wisest take here, right?
[00:29:00] Mike Kaput: We always joke that if someone's telling you what it's gonna look like five years from now, they're making it up because Oh, totally crazy.
[00:29:08] Paul Roetzer: the industry a hundred percent. Yeah. And that's, I think that's the real key for people. And again, if you're, if you're early in your career and you haven't learned this, Nobody knows anything.
[00:29:17] Paul Roetzer: like the people who are the experts, they, they have more experience and knowledge, but like even look at AI in this space, like we've been studying it for 10 years. Looked at, you know, probably over a thousand vendors. We've talked with entrepreneurs, researchers, you know, people within the research labs, like about as much knowledge as I could imagine having gained over the last 10 years.
[00:29:38] Paul Roetzer: And yet there are many times where I still like, I have no idea. I can maybe make a better educated guess than most marketers at what's gonna happen, but that's all it is. And, and I take some solace in, like if you go back and look at major predictions around AI over the last, I mean, you can go back 70 years, but take the last 10 in particular.
[00:29:59] Paul Roetzer: Mm-hmm. , the, some of the, the, the brightest minds in AI today. Like the, the. The absolute leaders in this space have been wrong over and over and over again around when they thought major AI milestones would happen. Yeah. They've overestimated, they've underestimated. So that's why I say like nobody actually knows what's gonna happen.
[00:30:20] Paul Roetzer: Um, I just think some people have more insights based on how close they are in proximity to what's being created, but even then, they're just trying to project out what could happen as a result of this innovation. But, It is, it's a great time to be trying to figure it out because we just need more people thinking about it related to their own specific knowledge set and domain.
[00:30:42] Paul Roetzer: Yep. Because that's where the opportunities lies, to figure out what it means to you and to your company. Absolutely.
[00:30:49] Mike Kaput: So for our third and final topic here today, this is a fun one, I think. Mm-hmm. , um, A company called My Heritage. It's a discovery platform that helps you find people related to you. So think like a 23andme, I think you take a D N A test, use it to do genealogy family research.
[00:31:09] Mike Kaput: They just released a free tool called AI Time Machine and. You upload a photo of yourself and an AI creates hyperrealistic avatars of you in different historical time periods, so you can show yourself in prehistoric times, ancient Egypt, modern times. There's all sort of black and white photos that look like they're from like the 1940s.
[00:31:32] Mike Kaput: It's just a cool little twist on this hyper popular photo app concept, like a lens, which we've talked about. and the reason I mention it is for a couple reasons. One, it's just getting a ton of buzz. I think it's another interesting example of AI becoming really accessible to show people exactly what uh, it is capable of and kind of blow them away.
[00:31:56] Mike Kaput: Now there's a lot of people using these, testing these out. I saw a lot of journalists trying it out, uh, recently, but also I think it's an interesting marketing strategy because my heritage has also. Release multiple AI tools that have gotten a ton of buzz, and obviously we're talking about it so it's working too.
[00:32:13] Mike Kaput: So I guess first I wanted to kind of ask, could you just anticipate these types of like consumer image AI apps, just like really accessible consumer apps to become kind of super popular moving
[00:32:24] Paul Roetzer: forward? Yeah, I mean that definitely, I mean, we're a marketing show and an AI show, I guess a marketing AI show.
[00:32:33] Paul Roetzer: But the marketer in me first and foremost is just. Applaud, like just brilliant marketing. So, you know, I, if people don't know my background, I started HubSpot's first agency, so like PR 2020. My agency I started in 2005 became HubSpot's first partner in 2007. So we were there in the early days when Darash Shock created the website Greater and launched website greater.
[00:32:57] Paul Roetzer: Mm-hmm as a lead gen vehicle for HubSpot, which have been used millions of times and generated millions of leads for the years. All they did was created a tool, like in that case it was a value-based tool. Um, that's all my heritage has done here, is like they looked at AI and said, cool, how do we create some like viral tool?
[00:33:18] Paul Roetzer: That people either find fun, interesting, maybe there's some emotion tied to it. And so good on them as marketers to just look and say, what are we capable of doing to market our brand that we couldn't do before? And so, yeah, I mean they had the, what is the, um, the one where you could animate? So they, yeah, deep nostalgia, I think they called it Yes.
[00:33:38] Paul Roetzer: Where you could animate. And then they did live story where you could actually synthesize voices if you. A deceased relative's voice. You could train a synthetic voice on it and then animate the photo. So if you like your kids never met the grandparent, yeah, you could actually bring a photo to life and have it in the voices like creepy as hell.
[00:33:58] Paul Roetzer: I don't know that I would do it, like it kind of like gives me the heebie-jeebies a little bit, but. Brilliant, and I'm sure it's wildly successful. So I think that that's my, my first take is it's genius marketing. Kudos to them for doing it. They obviously have someone on that team, which we should probably try and get on the show.
[00:34:18] Paul Roetzer: Yeah. Who must be the brains behind this who's looking at AI and saying, and I think that is maybe the greatest takeaway here, is what did they do? To, uh, infuse AI into their marketing. Mm-hmm. . And what lessons can we learn as brands and as marketers to, to do the same? What is the opportunity for us to create something that's a hook or a lead gen magnet or a value creation that can draw people in and build an audience?
[00:34:47] Paul Roetzer: Um, but in their case, it's just brilliantly connected to their actual brand and value proposition. Like it's, it's so well done. I, I don't know who the marketer is behind this, but I would like, Yeah, I
[00:34:58] Mike Kaput: thought what was really cool, even going one level further on the marketing genius here is if you think about it, I mean, I don't know anything about the genealogy space, but you gotta figure one of the bigger barriers as people.
[00:35:12] Mike Kaput: Sometimes just don't care enough to go Yeah. Do that kind of research. Right. There's obviously people very passionate about it, but if you sit someone down who's not super into that space, it's just like, okay, like, you know, look at a bunch of old documents about maybe a photo or two if you're lucky. Yeah.
[00:35:27] Mike Kaput: Of like very old relatives. This lady makes this experience something completely mind. Yep. It changes the game and just really connects in a way to their off, it probably transforms in some way their space itself and how you might even be doing certain genealogical research or investigation moving forward.
[00:35:48] Paul Roetzer: Yeah. And there's a lot of interesting potential there. I agree. And I think, again, it's like once you understand what AI is capable of, you need to connect the dots in your own career and business. Mm-hmm. . So I don't know, in whatever field you're in, like look at this and say, okay, what's the parallel to me?
[00:36:06] Paul Roetzer: Like what, what could I be doing that would be like so cool and out of the box that someone who doesn't understand AI would never think to do, cuz you wouldn't even know it's possible. And that was one of the themes of our book is like you just look at problems differently. Once you understand ai, like once you know what you can do with language AI and vision AI and predictive modeling and stuff, you just start to think about things very differently.
[00:36:27] Paul Roetzer: So if. If you were on the demand gen team at My Heritage and you're trying to like fill the funnel with leads or whatever, and you were doing it through e-books and webinars and direct mail, whatever you were doing before. Yeah. And then someone comes along and like, Hey, why don't we make, why don't we animate dead people, , you listen to me like what dude, are you talking about?
[00:36:51] Paul Roetzer: And then we'll like create synthetic voices. We'll train it on. Have people upload, uh, recordings of their deceased relatives, and we'll train synthetic voices and everybody in the rooms can be like, I think you need to leave. Like, this is . But that's, that's where we're at in marketing right now is like, be the one in the room who has that idea where everybody else is looking at you.
[00:37:09] Paul Roetzer: Like, what are you talking about? And you know, it's doable because you've seen it, you know what AI can do. So yeah, I I love that you picked this one because I think it is a very, it's more of like a strategic creative. Avenue. Yep. Um, where people need to connect some dots on their own, but they're there to be done.
[00:37:26] Paul Roetzer: It's so cool. It really is.
[00:37:29] Mike Kaput: It's, uh, it's pretty stunning what we should Yeah. If, if, uh, by any chance the My Heritage Team
[00:37:35] Paul Roetzer: listens to Yeah. If you're the cmo, we'd love to. I'm gonna look it up right as we get done here. But yeah. If you're the CMO or the brains behind this, this initiative, Hit us up. We'd love to have you at the conference and get you on the podcast.
[00:37:46] Paul Roetzer: Absolutely.
[00:37:48] Mike Kaput: All right, well, I think that's all we got for today, Paul. Um, thank you again for all your thoughts and good stuff. Thank you for your analysis. I think we learned a lot from, uh, from your analysis of writing tools and also just talking through these other really exciting topics and ai.
[00:38:02] Paul Roetzer: Another week in ai, man, it.
[00:38:05] Paul Roetzer: It's not true. It's a fascinating space right now.
[00:38:08] Mike Kaput: one of these days, we'll do the episode and at the end tell everyone it was our deep fakes or
[00:38:13] Paul Roetzer: something like that. Yes, we We could do that. We should do that. We could do that. 2023, we'll create synthetic versions of ourselves and we'll do an entire episode with just our synthetic versions.
[00:38:22] Paul Roetzer: Nice. Write the script with chat. G p T. Yeah. Synthia, dystasia, whatever it is to create our synthetic version. All right. We got, we got a plan. Good project . All right. Well, thanks as always everyone for joining us. We will, uh, hopefully be back again next week with three more interesting topics that, uh, see what the world brings us in the next seven days,
[00:38:46] Paul Roetzer: All right, Mike. Thanks again. All right. Thanks Paul. Talk to y'all soon. Bye.
[00:38:50] Paul Roetzer: Thanks for listening to the Marketing AI Show. If you like what you heard, you can subscribe on your favorite podcast app, and if you're ready to continue your learning, head over to marketing ai institute.com. Be sure to subscribe to our weekly newsletter, check out our free monthly webinars, and explore dozens of online courses and professional certifications.
[00:39:12] Paul Roetzer: Until next time, stay curious and explore ai.