Thoughts on AI

Stu Eddins:  And right now? Don’t bet the farm. Because whatever you’re doing today is gonna be different in a week and a half. 

Did I say that out loud? Welcome to did I say that out loud a podcast where Stu Eddins and Mariah Tang reflect on agency life and answer questions from our higher ed and healthcare clients about the latest in digital marketing, content and SEO? 

Stu Eddins:  Well, we’re going to talk a little bit about AI. Definitely not for the last time. There’s a whole lot going on with the topic. Most of what we’re going to give you his opinions right now. Because while there’s a lot of solid see a lot of procedure, and a lot of a lot of thinking going on right now about how to apply it and where people stand individually, or organizations is still pretty formative. And we get asked a lot of questions about AI. And for the most part, those questions come down to what the heck, everybody, it’s the thing everybody wants, but nobody knows what the heck to do with it right now. 

Mariah Tang:  Yeah, I saw a really funny meme the other day that said, AI is kind of going backwards, I want it to do my laundry and my dishes and things so I can have time to work on my art instead of doing my art for me. Oh, yeah. 

Stu Eddins:  Yeah. You know, I, I think that has a lot to do with it. What the expectations people have, are going to be as varied as there are people as it is right now. And this big amorphous thing that we see being applied in specific ways, but you know, search for right now in our, in our industry, where he we’re hearing about AI powered television sets coming out. So I wonder what that really means. I mean, does that mean that if I want to watch the old Magnum P I, it’s going to find it for me all the time and bring it up with the right one I’m touring? Money Mustache? Yeah, sure, everybody needs an 80s mustache. But the but the, the applications of AI are gonna be very, just to lead off with an opinion. I think AI right now is in the state that things like internet search was in the early 2000s. I think it’s at that point where everything is exciting. There’s so many different options out there. Nobody knows exactly what this can do for you, and how it’s going to work. Everybody wants a slice of it. But it’s very, very formative. It’s very early on. 

Mariah Tang:  Yeah. You’ve said this before to sue that. Back in the 2000s. You know, when they were building up Google and it was this new shiny thing. And even then they were starting to build their their databases, they were starting to build their large language models and all those things. So how, I guess how, in your mind are the opinions you’ve heard from clients and and people out in the world? How does how is this different? Like, how is this new? 

Stu Eddins:  One way it’s new? Well, first off, I think one thing we can expect is a growth trajectory. That’s going to be astounding. I mean, we look at that search from Bing, Yahoo, Google, whoever today. And we think we’re looking at a mature model. And we’re not, it can’t be a mature model AI is about to be added into it. There’s growth right there. But it took search engines a long time to get where they’re at now. But I believe that what’s going to happen is AI will also mature but the curve to maturity is gonna be much shorter. Every time we have a new technological thing come out. It’s ramp up time from from being a concept to reality to functional, keeps getting ever shorter. One thing that that may be contributing to this is, some degree AI can instruct itself on how to get better AI, the machine learning part of but I don’t know, we’ve been working with AI, quite frankly, since I think about 2007 When Google put quality score into Google ads, it AI is not new. When it comes to digital interfaces and how things function. It’s been around for quite a while. Really what’s happened is it’s just stepped into the forefront of conversation in the community, among consumers and so on. Because of things like chat GPT. And I think several things have been lost along the way. You notice how I have haven’t answered your question yet. Give me a few minutes. I’ll completely forget what it is and go on a different tangent, but I think that people have a hard time wrapping their arms around what exactly AI is at the moment what it’s doing, I mean, in search, well, everywhere. 

Stu Eddins:  I was reading an ad from somebody who was suggesting that they were using chat GPT to help people forecast their purchases. For stocks, what? Well, the it’ll completely overlook the fact that at this moment in time, all the information that is in chat GPT, and most of those search engines is from 2021. Before the large language model was built at that time, the only reason that tool like chat GPT. And please understand where we’re at, at several watershed moments, and by the time this gets published was meant to be true. But at the moment, the only reason that the AI tool like chat GPT knows that one plus one equals two is because it read it somewhere, it doesn’t do math. The large language model is just that it’s about language, it’s about words, putting things together, your side of the world content. That’s right now where it has the potential for the greatest impact. But the question is, if I asked you to write an article on XYZ, and you ask it to write an article, using the exact same prompt the exact same everything? Will we get to uniquely unique and different articles out of it? Or will we just find that we have the same thoughts perhaps rearranged in different paragraphs and a different rhythm to the article. It’s all tapping into the same model. It’s massive model, huge model. Consider for one thing that Google has been digitizing both on a mission to digitize every book ever written as part of their length, large language model. If you go to books with Google, you can use it for research with interesting recent example that reading a book about generational differences and what they are now. They rely on the Google Book tool, find out when certain themes come into people’s consciousness, because it starts appearing in articles and in books and so on. SafePlace was really not a thing before the mid 2010s. Yeah, it didn’t appear much. And then since then, it just ramped up. Again, an example of how language helps create analysis, but it’s not doing the analysis, right. language helps create an article, but it’s not really writing the article in most respects. So it’s that expectation side of it. One plus one equals two. Yes, it does. I see it, Chet GPT, comes up with it. Well, my expectation that it’s gonna help you buy stocks. No, not now. And we’re starting to get some hybrid, going on hybridisation fill in the blank with that word. But what’s happening is, they know what they have chat. GPT is now using Bing to do the computation inside. But the thing is, you’re not getting an answer on AI. The AI says, I don’t know that lets us band come up with an answer. It’s just simply doing the search for you. Right. So that asks the question, when it comes to your stock portfolio, would you ask the question and being and do what it says? Right now, kind of the same thing. There are people out there building AI models who handle stock purchases, they’re doing that. But that’s not chat GPT. That’s not Google’s barred. That’s not those things. So really, kind of reeling this back in from this flight of fancy I’ve taken, but expectations are wildly all over the place. For myself, I think that looking at the state of AI right now, as it affects my working world and my clients, I think that we are at a moment where the expectation is that AI is going to make life simple. For advertising. I go into my campaign, I set up my account set up a new campaign, I click all the AI buttons. And that’s it. I don’t have to have the skills anymore for keyword research. I don’t have to have the the knowledge about how to write a good ad. AI is going to handle that for me. And there’s a great big goal, yes, but yes, AI can do those things for you. But they’re gonna give you the same results as everybody else. In that respect, AI just creates a new level playing field. That if you simply turn it on, everybody’s on the same level. So how are you going to stand out? Yes, you’re unique, just like everybody else. 

Mariah Tang: That’s how I feel about not to knock on any of the programs out there. But that’s how I feel about things like Grammarly when I have clients that say, oh, I’ll just run it through Grammarly. It’s fine. And I’ve done editing for people who have self published books and things like that and I’ll just run it through Grammarly. I don’t need you to review it. And then I’ll just review it because I’m spiteful and find all kinds of errors it’s because Grammarly is trained on specific rules you know, it’s not necessarily colloquialisms and and you know Have analogies and little stories that people tell. So while yes, it may be exactly pointing out specific things, it doesn’t always generate content that a human would say, or a human would read. It feels like that’s the same point we’re at with AI, you type in your question or your or your prompt and search, it’s gonna still pull up that list of URLs and relevant sites, but it’s also right now, giving you this negative, here’s all the stuff we piled together into the summary. And just working with healthcare clients, I’m not a clinician, but boy, have I seen some crazy things in those summaries that are absolutely not correct. Same thing was just about anything, though. I mean, you have certain sources that you would use when you say go out to YouTube, and I needed to learn how to do the oil on my car change label my car, am I going to listen to somebody from AutoZone, or I’m going to listen to some guy on Tik Tok that’s like, Hey, dude, I’ll tell you how to do goes right. You have to still look at your sources, you have to still look at, where’s this information coming from? And I don’t know that we’re quite there yet. At some point, we will be, but it’ll still be up to you as the user to make those decisions. Yes. 

Stu Eddins:  Okay. And I think that that is probably one of the key takeaways from what we’ve talked about over the last several weeks. AI is there, it can make life easier. It can do a lot of things, you know, it can it can make your hair shinier, and your teeth brighter, or whatever, we can do all these things for you, maybe. But the real, the real application of AI isn’t enabling it, it’s controlling it, you do have levers to pull in almost every situation. Or you can guide the response you’re going to get by applying the knowledge you have on the topic. The way we see this, again, relating it toward SEO or toward advertising. You can turn on AI and Google ads. Beautiful example. Performance Max is a campaign type lots of AI built in baked into this thing. And most advertisers will simply turn it on. And they’ll supply all the images they need, they’ll supply the short and long headlines at a body copy and let it go. That’s what most people will do. That’s the level playing field. It’s understanding what the tool can do that’s going to be beneficial. That’s gonna be the separator between the average and the not so average, the slightly higher low bar. Yes, yeah, well, but frankly, if AI sets the low bar, it’s up to the skilled advertiser to set the high bar, right. If you want to be a better promoter, through advertising, if you want to be a better writer, you have to apply yourself your knowledge, what you know. And the way you do that with the ads, you give it more information, but you give it the information, you know, is going to shape it in the correct direction. If I have this performance Max campaign out there that’s mixing and mashing together stuff to make the ads that it finds gets clicked more often that get converted more often on or on, great, but what if it’s good sending all those ads to people who are already science customers, or students or patients or whatever? Preaching to the choir. All I’m doing is I’ve just scavenged all my organic or brand equity, and paying for it. The better idea is to take that tool, and then give it the list of the people who’ve been successful. Give it your list of conversion events that happened, let it say, Okay, this is what success looks like. Let’s go find more people who look like that. Yeah, that is a first step in making AI work for you in that digital marketing space. That’s Google ads. 

Mariah Tang:  Teach it what you want it to do, right, more of that thing, in 

Stu Eddins: Teaching is a good word guidance is another good word. Eventually, AI may come around and bend itself to your will, if you just do the basics and set it up and let it go. But how much are you going to be willing to pay in time, effort, capital, whatever, until it finally catches on and does the right thing? Are you gonna shorten that? You could do that right up front. You could do all along by inputting your knowledge. AI is terrific at parsing things out, but sometimes in most cases, is not terrific at making connections. That’s where the human comes in and does that. It says, Thank you, Google, you give me this AI tool. But it needs to know this. It needs to know that it needs to not spend more money than this. You need to set the guardrails up and you need to feed the information in. 

Mariah Tang:  Yeah, it starts with that really good data, which is the problem for a lot of organizations, I think, yeah, if you have messy data, you’re going to have a messy. Yeah, you’re gonna have a messy data set that comes out. 

Stu Eddins:   Yeah, it will be ages ago, computers first came in garbage in garbage out, it goes back to that. We’re no different today, you’re only going to get out of it what you put out. And if you put in just enough effort to enable things, you’re going to get that level of output. Yeah. That I guess that was my Captain Obvious moment for this particular episode. But it, it’s something that’s easy to overlook. When you’re in the middle of a very busy day, when you’ve got nine different people chirping at you during the day about things they need right now. And you’re sitting down to do something that could be an AI enabled task, writing that blog article, whatever it is, it can be tempting to turn to a tool, like chat GPT and say, give me a blog article on patient acquisition in the Detroit area x y in printers, and go and get back what you get back. And then your effort goes into editing what it gave you, instead of applying what you know, to the suggestions that have reading the output as a suggestion, looking for gaps that you didn’t know existed, but still using that information, it to augment what you’re already planning to talk about, as opposed to say, Good luck getting published. 

Mariah Tang:  That’s how all of the leaders in content right now are using the tools and are creating their own personas with their own data, not relying on the master data that’s out there and creating prompts that generate useful outlines, like you just said, shooting the gaps, finding the opportunities. It really, it makes me think that back when I was in college, because I’m an old lady now in my in my early 40s, here, when the internet really took off and you you were no longer going and poring over card catalogs and things at school did write your thesis paper. And how the librarians were a little I wouldn’t say off plate. But they were a little nervous, just like all of us are now what’s going to happen to my job am I going to be useful? It didn’t get rid of any librarians, it turned them into supercharged data scientist, right now, the librarians, as they always have been, are some of the smartest people on the planet. They know how to find everything, the tools, they use changed. And I feel like that’s where we’re going as an agency as colleges is healthcare systems. The tools are changing, the way we use them are changing, but it’s still that expertise. And that years of experience and the knowledge that goes into using those tools that will generate the outcomes.   

Stu Eddins:   What is this library of which you speak? Anyway, with the cookies with the books and the smart? Yeah, the I don’t know, we could go on this tangent forever. And we have multiple episodes in our future that will extend this conversation. I think the thing that I would get across at this point, it’s young, it’s new, it’s fresh, there’s still a lot of growth ahead. Don’t expect? Because you can get a response back, don’t expect it expect it to be the best response to the question being asked. And I think the further extension of that is going to be not only we get, get out what you put in, but you need to do the research to find additional inputs that you need to supply. Find the ways that you can take the data you already know, the information, the style, the whatever it is you already know, and feed it into the system to get back the correct unique and somewhat personalized response that it may be able to give you. And right now? Don’t bet the farm. Because whatever you’re doing today is gonna be different in a week and a half. Yep, I referenced Google, Bing, Yahoo, Amazon, Bing and Yahoo. Search before. Let’s also add in dogpile Lycos crawler, think about all the search engines that were available in the early 2000s. They’re gone for they’re using Bing or Google algorithms to power their search and they’re just you know, also RANS. The same thing is going to happen with AI. We will probably see an expansion, huge explosion of AI tools and companies and there’s going to be consolidation around the best ideas. Yes, yes. Vote with your attention. If it’s a subscription vote with your with your money and your and your time. But understand that whatever you whatever you start off with today, will most likely not be the solution five years me. 

Thanks for listening to Did I Say That Out Loud? with Stu Eddins and Mariah Tang. Check out the show notes for more information about today’s episode. And if you have any questions, concerns or comments, hit us up anytime at stamats.com.