In this episode of Your AI Injection, host Deep Dhillon sits down with Greg Starling, the Head of Innovation Lab at Tailwind, a company intent on utilizing generative AI to optimize social media campaigns and small business marketing. Deep and Greg begin by discussing the early days of Tailwind, from their engagement with OpenAI's GPT-2 to the acquisition of Replier. The two also discuss how to leverage generative AI as a brainstorming tool, a market analyst, and a method of generating user stories. They then delve into the ethical considerations of generative AI for marketing— from preventing plagiarism to defining & eliminating hate speech. As the two look towards the future of AI in marketing, they discuss the potential shifts towards trust-driven, community-focused platforms, foreseeing a rise in micro-communities and autonomous AI agents that personalize experiences and handle much more complex tasks, such as personalized travel planning.
Learn more about Greg here: https://www.linkedin.com/in/gregstarling/
Learn more about Tailwind here: https://www.linkedin.com/company/tailwindapp/
Check out some of our related content:
xyonix solutions
Learn more about Xyonix's Virtual Concierge Solution, the best way to enhance customer satisfaction for industries like hospitality, retail, manufacturing, and more.
Listen on your preferred platform here.
[Automated Transcript]
Deep: Hey there, I'm Deep Dhillon, your host. And today on our show, we have Greg Starling, head of innovation lab at Tailwind. Greg focuses on leveraging AI to optimize marketing strategies for businesses. Greg's been leveraging generative AI and his products, and is currently working on a book called “Conversations with robots”. Today, we'll dive into the world of AI in social media campaigns and small business marketing.
So maybe we'll get started. At Tailwind, you decided to leverage generative AI as a product offering. Talk to us and our many product builders that are listening. Like, how did AI get on your radar? How did you go from that stage of kind of maybe just knowing about AI through to product development and sort of integrating it and leveraging it to kind of accelerate your business?
CHECK OUT SOME OF OUR POPULAR PODCAST EPISODES:
Greg: Sure. So probably at this point, uh, three ish years ago, um, my job really is just to kind of go out there and see what's coming, trying to figure out what's coming, what's going on and, um, opening I got on our radar, uh, around the GPT-2 timeframe and started playing with that. Trying to get in on some of the early betas, understand what we could could do there, um, really felt like that's where things were going to go really almost the first time that I saw what generative AI could do.
Uh, it's magic, right? It feels like magic. And it's like, how do we, how do we figure out how to get this into our product? So we shortcut it candidly. We acquired a company. So we, we had a, a company called, uh, replier, um, that was using, um, uh, GPT-3, um, they, they just early, early days of GPT-3. Um, it was still whenever GPT-3 was in beta.
So by acquiring them, we were able to kind of shortcut the line, um, start using it. Plus we were able to acquire a little bit of talent, um, that we needed. So some developers, uh, we had in house guys working on it and we're, we're learning, but it's always nice to try to shortcut that and people have been, been living in that world.
So this is, um, I guess early 21. So are these,
Deep: uh, this team that you acquired are like customizing models, um, through a service offering or are they.
Greg: The first version was just straight using GPT 3. And then we were, we were working through the outputs, um, and trying to get really good at, you know, whatever you want to call it, prompt engineering, prompt design.
Um, but, but, but utilizing, um, prompt, um, design to, to come up with, with the kind of results we wanted. Um, and they were doing it, uh, in the world of like, you get a customer view on your Google for your business and automatically replying back, um, based off some sentiment analysis. Um, looking at the stars, um, the star rating, looking at what the words are based off that sentiment coming up with or applying.
So that's, that's what we acquired. Um, candidly, we scrapped most of what they were doing because that's not really our business. Um, but. The concepts behind it were the same. And so for the last, at this point, uh, over two years, we've really just been kind of been working on, on really honing that in, um, and, and trying to better understand what models make sense when it makes sense to train versus when it makes sense just to use things out of the box, when it makes sense to use single shot or zero shot or multi shot.
Um, prompts and, and those, those types of things. And so that's, it's, it's, um, there's no books really to get out there and read, unfortunately. It's, it's so new. Uh, but it's, it's, uh, there are some good groups and some good people that were able to bang ideas off of. And so we've just kind of brute forced it, uh, to get to the point.
Xyonix customers:
Deep: maybe like. Talk us through your product a little bit. Like what, what's the products that you were trying to put some innovation into? Yeah. What are they trying to achieve? And then as you know, the leader of some of innovation in your company, once you've identified ChatGPT, like how, how do you go about.
Getting your company on board with the innovation and, uh, and, and leveraging the AI and actually getting it
Greg: into product. Probably the best way to start was getting people on board because that was so insanely easy. Uh, this technology is so kind of blow your mind out of the gate. And I think At this point, we look at it and it's still really nascent, but even in its kind of early stage, we look at it and we're kind of getting used to it or accustomed to it.
But if you go back, you know, even six months, like pre ChatGPT release, seeing you ask a question and then getting something generated blew people's minds, like, like people would have thought, you know, I was the greatest magician of all time. Being able to say, hey, generate a Pinterest pin description. So that's what we do.
We, we create social media posts, mostly around organic content. We do some of the ad space too. And so what we were using or what we're using it for today and what we continue to build on is around. Marketing, we work mostly with, um, very small businesses. Like people talk about like SMBs, uh, SMBs usually are way too big for us.
We, we work in like a whole new space. Like we got like VSBs, like very small business, solopreneurs, one, two, three person companies, people who don't have time and usually honestly don't know how to market themselves. And so we've created tools for scheduling, historically scheduling and being able to batch process, things like that.
What generative AI got us into was the ability to start to actually do some things for them to create these based off maybe a product they're selling a particular cell they're running, um, a new a new line they're releasing a new blog post they might have written and take that content create all of the organic.
Social media pieces, look at what's performing, look at some sentiment analysis, and then maybe you can push some through some ads. Um, but so that's kind of what we do in a nutshell.
Deep: So is that, is that like the entry point for them? They give you a blog post or they give you something, uh, you know, or like their product, their product catalog or something.
Greg: Yeah, that's that's that's a really good way to think about it. We do create things if you don't have that or you're just kind of getting started, we create marketing calendars like I'm sure everybody has seen, you know, it's a lot of like
Deep: HootSuite like functionality then. Yeah,
Greg: yeah, yeah, that's a really good way to imagine HootSuite for your neighborhood painter, you know, you're okay.
You're your mom pop sculptor who is is back at the back sculpting stuff, and he doesn't have a social media team. That's the week we um, we're like maybe maybe a good example would be like And, uh, is to Adobe, um, we would be that like a hoot suite. It's just a much simpler, much more streamlined, much more opinionated concept to make it much easier for somebody who doesn't live in this world to kind of get up and running.
So, so a lot of our,
Deep: our, a lot of our listeners are, you know, product managers, product builders, you know, maybe, um, executives at companies that, um, could be in a very different space than yours, but. Share this need to, to like, or this desire to leverage AI and as an innovation engine in their companies, like, how would you maybe talk to them about your experience if you take it up a level from your specific product?
But like, what kinds of things do you think they can do to, you know, maybe get started and, um, and, and try to figure out how to even move forward? Like, what would that even
Greg: look like in their product? I think for me, I live probably in that world. That's one level up. Um, so, so this is really kind of in my wheelhouse.
What, what I use, I think what I use. Generative AI for on a daily basis. Most anybody in any business would use on a daily basis. I love it for brainstorming for for saying, Hey, here's my idea. Here's what I'm thinking. Give me four or five counter intuitive concepts or challenge me on. This, um, I've actually written a few prompts, uh, just using like Chat GPT, that's some friends, some of you mentioned you have some product, uh, product manager folk they'll use to create cards, you know, here's my idea and here's my concept and here's what I want to build, go and create a whole bunch of, uh, user stories for me or, or cards or, or give me some market analysis, like anything a product person does, uh, I think, um, AI, uh, in a lot of ways.
Is is a brilliant, um, assistant and it has the ability to, um, scale expertise. And so there's a lot of stuff. I don't know a lot of stuff. I don't know. And so whenever I'm playing in spaces, I don't know, uh, I can get up to speed in a hurry, uh, asking the, the GPT, a lot of, a lot of questions, a lot of, um, uh, you know, what does this market look like?
What is, how big is the size? You know, what's the addressable market. And it may not be perfect, but it's going to give those a hell of a lot more than I know, and so it's going to give me a long ways there. And then I can use that to create, um, create user stories, create cards for my engineers. Um, so I utilize, um, uh, AI on a daily basis.
Um, so just act as like, for lack of a better word, four or five extra hands to kind of just help me, help me do, do, uh, things a lot quicker. And, and it really, I wouldn't say it's doubled my productivity, but I'm probably 30, 40% better and quicker. Yeah, I think
Deep: that's a, that's a really good point. Like, I think a lot of us.
Maybe pre ChatGPT, we're used to having certain boundaries around what we know about what we're an expert in, what we maybe know a little bit about, and maybe stuff we don't know anything about. But now, you know, you can, you can really just like very rapidly, like, I don't know pretty much anything about quantum physics, but if somebody starts talking to me about it, and I've got a little bit of time and ChatGPT, I can suddenly like have a way more interesting conversation with them than I could without any quantum physics in my back pocket, you know, absolutely.
And I feel like I, that's a really good point though, is before trying to figure out how to actually use AI in your product. Um, and some of these kind of new capabilities like LLMs, like generative machine learning, or even just kind of like more classical, uh, approaches or time series forecasting, all that stuff.
You GPT and say, Hey, I've got this scenario. How would AI make sense in here? And you'll, you'll start to get all kinds of ideas. Um, how you execute that, you know, could be a second secondary thing, but yeah, it's actually an amazing place to start. And
Greg: then you, you mentioned it's easy to take that and say, you know, this is what my company does.
Uh, what is, what, what are things that I, we could, we're missing or what are things that we could add on to? And then you go down and to your point and how, how can we leverage AI? And then all of a sudden you have a roadmap that you can build out in a day that it's probably better than something I would have put together in two weeks.
Deep: It's kind of mind boggling. I know. I mean, I've been using it flat out for like project proposals. I mean, stuff that, you know, requires a lot of in depth expertise. I mean, I don't just say, write this proposal and go, but like, you know, I get to, I don't have that blank page problem. And, you know, and a lot of times though, it has some really good ideas and then you kind of go back and forth and you're like, Oh, okay.
Yeah. I hadn't thought about that particular angle. It's
Greg: like whiteboarding with a team. Yeah. Yeah.
Deep: Yeah. And I don't know if you've seen some of the other tools out there. Like one of the ones I really like using is perplexity AI. Um, it's, it's, um, it's kind of more, it's got a lot more depth on the, on the academic content.
So one of my use cases I used to, you know, really. Be lazy on the paper reading front and I just have, you know, my engineers like read stuff and then I just throw it at the whiteboard and ask a million questions. Um, used to take a long time and but now it's like, I just go to perplexity and just dig in on a new concept that I don't understand.
And it's, it's like a whole other level of depth and then they because of the citations, you can go read the original source materials at any time, which I find. So,
Greg: yeah, I love being able to kind of rabbit trail that for sure.
Deep: You're listening to your AI injection brought to you by Xyonix.com. That's X-Y-O-N-I-X. com. Check out our website for more content, or if you need help injecting AI into your organization.
Deep: For the product, um, conversation, going back to that again, I think this idea of like going straight to ChatGPT to just ask stuff and get ideas is awesome. What's the next step after that? So now you kind of like. Um, you know, are you always building in house? Are you bringing in some outside expertise?
Are you, um, like, do you, for some of the larger companies, do you have... Are you seeing more folks set up like innovation labs, you know, where there's people in a role like yours that are tasked specifically to play with the new stuff coming out?
Greg: And I wish I wish there were, um, uh, it's kind of a dream job.
Honestly, I had a company I worked at previously. My previous roles is a little more kind of in your, your realm is more like CTO, CIO, techie kind of roles in my previous role. I got to go out to Cupertino and work with Apple's innovation lab. And then I went to San Francisco and worked with Airbnb's innovation lab.
And I was like, I love this idea. Why aren't there more of these out there? Um, but I get, I know why it's cause they're really expensive and probably this is this generative AI thing. We were on it early and it, it's put us in a really good position, but I've probably thrown out 10 times more ideas than we've ever used and, and I get there's a, there's an expense there, but if you ever hit one, it can be really.
Really beneficial. Yeah,
Deep: I mean, I think that's the key, right? Like, that's the hard part that it's like, you can have a great idea, you know, it has to be done. You've still got some institutional momentum that's going a different direction. In some cases, you know, it's so obvious that you have to pivot, but almost always there's costs involved, whether it's monetary, whether it's time, whether it's, we have to do this other thing, um, to get there.
And like a lot of traditional businesses just have a hard time. Leveraging innovation, even if they had their own Xerox park or they had their own, you know, Apple innovation lab. So
Greg: yeah, it's, it's just like classic innovators dilemma, right? Like we have this, we have this engine that's making our, all of our money.
And what you're saying is we need to throw out some or all or some portion of that and, and take a risk on something that we don't know is going to work. And the problem is right. If you don't. If you don't take that leap, you don't take that risk, you'll get out innovated and you'll become obsolete, but it's a but you can also and I think it's important to understand that there's a risk to doing it right like you can you can take that risk and it doesn't work right like you you go all in on some tech or some concept and it doesn't work and you're like, man, if we would have just stuck with what our cash cow was and hadn't diverted all these resources, we'd be in a much better position.
And that is the dilemma and I think I think it's a very fair struggle and it's it's fair conversation, um, to have entire side of any organization, but then I would hope that most companies could spare 1 person, you know, just just spare somebody to go put off in a room to the side and let them research and come to the executive meeting, you know, once a month or something once a quarter and say, these are all the cool things that are out here.
Um, that person should become really good friends with the head of product, you know, and say, this is what you need to be doing this. Um, but, but I think it's, I think it's got to be at least worth that in a company of just about any size.
Deep: So you mentioned you were at Apple's Innovation Lab. I mean, Apple strikes me at least during the Jobs era, maybe, maybe even still, but like more than most companies willing to, to innovate and even at the risk of cannibalizing their own business.
Like, I think, you know, they took a big risk when they released the iPad. Knowing full well that it was going to eat into MacBook sales. I'm curious, like, when you were there, did you, like, like, did you learn, like, like, what were your kind of main takeaways? If you can talk about them at some level? Yeah,
Greg: no, yeah, I definitely can.
So, so the company I was working at, we were in the hospitality space. And so Apple specifically, and why we got invited, they were looking to, to their, um, they bring their, their, their technology into hotels and things like that. Um, and so, so we got to work with their innovation team, and I think what they did.
Um, that was so interesting. Is this their innovation teams are actually because you have the big cool, you know, Apple Park. None of their innovation teams actually in Apple Park. They're all off campus. Not far. They're like they drive and eat on campus, but all their innovation teams are off campus. And I think what they had that was interesting and I don't know if this is the right way or the wrong way, but it's worked for them.
Um, those innovation teams. Didn't have a, um, a PNL, like in any real way, it was very much your job is to come up with the next new, cool thing. Uh, and, and when you're Apple, you can do this, you can say that, you know, cost to be damned because you're Apple and you have a giant war chest, but that was their approach.
It's like, we're in there, their teams are huge. They actually have five different innovation labs, um, that focus on different verticals that they're each innovation lab focuses on a different vertical. Um, and those. Those labs, uh, they, they have budgets, but they don't have any expectation of generating revenue because I think they're really good understanding that we need to be thinking 5 years down the road.
You need to be thinking 10 years down the road. These people aren't going to be the people that are going to hit our bottom line this year. And I think whenever you get short sighted like that, when you start thinking, like, What's the cool thing that you all are going to launch? It's going to move the number next year.
You start thinking about it in the wrong way. And I think that's something that Apple did really good. It's like, we don't have any expectation. Our, what we want to do is we want you to bring something to our meeting. They, they, they had to come present to senior execs once a quarter, bring something to our meeting.
That's going to blow our minds. That's going to make us go. Oh, wow. That's super cool. And and I thought there and one I think that was actually on the walls is present something cool. And that was kind of their their mission. I think doing that put them down a path that that allows them to continue to innovate and.
Um, come up with things that are just different, um, than what it's
Deep: kind of. It's kind of like the old almost the old school. Um, you know, pre pre, like, Google 20% idea, you know, approach to innovation where you had a. Somewhat segregated R&D lab where, you know, you hire a bunch of big wigs and you kind of can, can like contemplate out.
I remember my first job out of grad school was at a, at one of these kind of R and D labs was one of the first AI labs, um, uh, um, back then in a J it was kind of tucked inside of a giant telco. And because of a monopoly, you know, we were able to, to basically have this. mandate to just do whatever the heck we wanted.
I mean, we had some roles. This was, uh, the company's not around anymore. It's called GTE. They got bought out by Verizon, but GTE Laboratories. But back then, um, you know, our, our mandate was just to, I think we had a few roles. Like one of them was like no hardware, like, um, at no, like, cause before they had, you know, we had, like, there was like gigantic radar dishes and like physicists and like, so they'd moved everything to software like at one point.
And that was like a big, big decision for them. But other than that, I remember having the most bizarre mandate in meetings. Like when I first started, I was like, so wait, what am I working on? They're like, whatever you feel like, I'm like, what's my budget? And they're like, well, you get 80 grand to buy whatever you want, which, you know, it's like 1995 or something.
I said, Okay. But they said, but you can't spend more than 3000 on at one blow. Everyone was like incrementally constructing like these, you know, like their, their hardware devices and everything. And we came up with all kinds of cool stuff. We ended up specializing, you know, we, we spent a few years doing a bunch of like 3d, you know, world stuff, you know, that, that, that's cool.
Yeah. I mean, it was a lot of fun, but I remember part of it was that, you know, we also, this was really early days of the internet. We were also like utterly convinced the internet was this massive deal. Um, you know, and I remember having the most weird conversations with executives though. Like, you know, there was like an exec who was like a former ostrich farmer and You know, and we were trying to like, let them know, Hey, like the internet's a big deal.
Like everything's going to, this whole business is going to be radically transformed, like you gotta, you know, change onto this, that, and the other. And, and then meanwhile, we'd like give them access to, um, you know, we had this, um, you know, it was like this, uh, we had, we had just a ton of stuff. That's all, you know, become startups that are huge that we don't even think about as a big deal anymore, but you know, you can pick anything, pick search, whatever, and we'd put them in front of a browser.
And then we finally realized that. And we're like, God, why, why does this guy keep coming to web pages and just never clicks on anything? Like, never, never, he just never just didn't know what to do. And so, so then we kind of like, walked him through that. And then the takeaway was like, oh, we need to make a smart phone.
But like smartphone back then was like, you know, just like some clunky speakeasy looking thing. And, and that, and that to me, I like my takeaway, it was like, this is going to be a really tough place to actually drive innovation because the model's broken. Like, what needs to happen is anyone in this lab that wants to take this thing forward should be able to walk down, raise money, get a couple of million bucks, and then buy their freedom, go out.
Get outta the entity. Yeah. And I'm curious like what your take is on that, like, you know, like, um, you know, like, uh, like venture funds in that venture fund approach versus the kind of the big r and d innovation lab
Greg: model? I, I, I like it. Um, now I'm not a, I don't know how, how , I don't know how honest I should be about venture venture capital.
Um, I'm not a huge fan of venture capital. A lot of times I've just been on some bad, bad ends of, of venture capital, uh, a few times. Uh, but the idea of like, uh, the first time that I was playing with the lab, uh, was a similar type concept and, and I was actually trying to stand up a lab internally. But I think having a partnership with a big company is, is, is really beneficial.
Um, well having access to their resources is, is really beneficial because, you know, anybody who goes out and starts something, you need resources, whether that's people, resources, financial resources, whatever that might be. And, uh, I think it's a really interesting and, and I, I, I would love the idea of a model where the company has a little bit of an ownership and this.
Thing that they spent out, you know, whether that's, you know, 10%, 20%, 50%, who knows, you know, whatever's fair and makes sense. I think there's a lot to be said for that, especially in industries that, you know, we talked a little bit just about kind of the innovators dilemma where they can't kill the cash cow.
It just, it just, they can't, they're never going to get it past the board or even threaten
Deep: it. You know, it's like, it's just a non starter, right? You can, you can grow a new thing. But you can't kill the old thing. And even if the new thing starts eating into the old thing, that can also be problematic. So,
Greg: yeah.
And that's why I think it's a cool concept of what you're talking about. And I really liked the idea of if you come up with a new way or a new innovation that might kill the cash cow, to be able to spend that out and, and run it as an outside entity, as an outside organization without all those politics, but still having the company has some upside in.
And your success. I think that that makes a ton of sense. Or
Deep: maybe like for some time period, right. Of first acquisition or something like
Greg: that.
Deep: Yeah. So, okay. So maybe going back a little bit to, um, To the, the marketing, um, applications that you're seeing out. Like, what do you think, like with all this generative, um, AI stuff, are there any sort of ethical dilemmas that you're seeing, you know, where, you know, like, cause every like, whether it's, I mean, I know plagiarism is a, certainly a big deal to within the academic world, the educational world, I think less so in the business world, because I think everybody's just been copy pasting stuff forever, like with minimal citations that they've kind of moved on.
But, um, But like, I still think plagiarism is a big deal, you know, and, but beyond plagiarism, like, are there other ethical considerations that you think are going to make, play an increasingly big
Greg: role? I do. Um, I, I think, I think there's, there's enormous ethical considerations, um, and I don't. I think there's getting a little more.
Um, we're seeing it more brought up, but I don't think people realize how bad some things can be. Um, so there are multiple biases when you're looking at language models, right? You have the bias of the input, the data, um, unfortunately, the, the Internet is, is not always, um, a great place. And so whenever you're training it and you're just sucking up the Internet, you're bringing in a lot of those really terrible things.
Um, and I know ways like companies like like open a I know Sam Altman's talked about this one. You know, they have armies of people who are trying to go through and sanitize that I mean, they don't want people being able to rapidly generate hate speech or rapidly generate things like that. But then you introduce a whole nother set of biases and a whole nother set of problems, potentially, because now it's.
It's the person who has the control on what's going in and what it is able to be output. You have to have their biases. And so there is there significant biases. Um, I know we're, we're on a, an audio cast here, but, but I'm a middle aged white guy. If you go to mid journey and say. Generate a picture of a CEO.
It's going to look like me and there is real problems to that. Um, you say, you know, you can say go generate any X stereotype and it's going to generate somebody that looks like a stereotyped, um, ethnicity or, or person. There's problems with that. Those are things that we as a society have tried to kind of overcome.
And then there's real criminal type things. Like we can get real, real bad. I think. Uh, I know my grandfather who's passed, uh, now, but he got a call, um, several years ago about his grandson being in jail. Uh, and he called me on my phone and he's like, Greg, are you in jail? I'm going to, you know, there, I need to do, I need to wire you 500.
I was like, no, grandpa, that's a, somebody's just trying to take your money. But for somebody who's in their eighties, I mean, that was, that was a scary thing for him. Now imagine it's not somebody calling him. It's me. Or at least it sounds like me because I've been able to clone my voice. They've been able to not just, yeah, no,
Deep: it's funny.
You mentioned this, like my younger sister just, you know, texted me. She's like, Oh my God, do you know about this? You know, uh, these, uh, folks going around with like, uh, you know, being able to fake your voice. And I said, uh, yeah, I've been doing this for 30 years. I think the first, the first time I, you know, my, uh, brokerage account.
Uh, like, you know, authenticated me by voice. I immediately like got in and said, no way. Completely disable this from my account. I do not under any circumstances, authorize this.
Greg: Yeah. And for somebody like yourself who has all these podcasts and all this information, I can not only clone your voice, I can make it sound like you, like, you can make it look like me too.
Deep: I mean, there's enough pictures of me out on the way. I mean, all of us who like, you know, are technology leaders. You're you're out on the web. You've been there
Greg: for a while. Yeah. And I think those are, those are some real ethical problems. The, the ability to fraud, to drive, I mean, we saw it, we've seen it in elections already, and it's just going to, because one of the big benefits, and I think one of the magical things about AI and generative AI is it can be personalized to an incredible level, like, like an advertisement that looks like it was written just for you, because in a lot of ways it really was.
And that's kind of a magical experience. And you're like, I didn't even know I needed this. Thing and now I do. But whenever you do that in ways to to shape policy and to shape public opinion, um, and you can kind of prey on on the worst demons of people. Um, I think there's some real ethical concerns there.
And I, I probably go back and forth. I'm a little worried that, um. I'm probably a little bit of a libertarian at heart. I know a lot of a lot of us tech people tend to kind of lean that way you do your thing. Let me do mine. And so I'm a little worried if regulation gets out in front and tries to solve problems that don't exist.
Um, but I'm also worried that some of the problems that will rise. Will be so big will have wish we would have thought about how to solve them before they pop up. And so, so, yeah, I think there's some real ethical, some real concerns that how do you think about how do
Deep: you all at tailwind approach that with specifically with respect to your kind of marketing platform?
Like, do you do you take specific efforts to. Get involved in like, not, um, you know, not allowing particular types of marketing or is it not an, not an issue or no, no, it's very
Greg: much an issue.
Deep: Are there guardrails and what kinds of issues are there? Yeah,
Greg: we, we have guardrails, uh, up around, you know, just like normal things.
We, we have, you know, just kind of your basic. Recognize when people are using certain words, certain phrases, things like that, there's a whole flagging system that won't won't, you know, return results and flag and, you know, our customer service, let our customer service go and look at things. But there's also things that were very explicit about you're not allowed to use our platform for, um, we don't allow politicians to use our platform, uh, in any way anything policy related, it gets a little gray when you get into some nonprofit spaces, but there are there absolutely some nonprofits that use our platform.
But if you're you're running for office, uh, trying to push through legislation, you, you're, you're not, that's, that's against our terms of service. Um, and if we figure it out, you know, not, we have, you know, 60 some odd thousand customers. There's ways that people could sneak through, I'm sure.
Deep: But why ban political speech?
What's the
Greg: rationale? We're worried, uh, we're worried about people using our technology to, um, Shape public opinion. It's just not something we want to get involved in. Um, so, so we, we don't do political speech. Um, obviously hate speech, um, anything hate related and, and that gets really great to what's hate, you know, um, again, I probably have a little bit more of a, a laissez faire attitude towards towards that.
And then some. Um, because I think it gets gets really interesting on what's hate and what's not hate. Um, you know, uh, I live here, uh, in Oklahoma. That's in a lot of ways, a very red state and a lot of things that I hear on a daily basis from family members or whatever, like it's hate speech, you know, but, uh, but other people like, no, no, it's not, you know, and you're, you're trying to tell my ability, you know, to, to, to live my life the way I want to live my life.
And so. Yeah. Um, but, but we do put some guardrails. We don't do anything, uh, around, um, uh, I, I don't know, a nicer way to say than like pyramid schemes, the multi level marketing, um, you, we don't, you can't use our platform for multi level marketing. Um, so do
Deep: you have, uh, so I assume you have like some classification capabilities to identify these.
I'm curious, are you, are you, have you moved all that stuff up onto the LLMs and, you know, and you're using like maybe. GPT, three or five or four to get your training data. And as opposed to doing it, the old, this is so crazy. Cause this is like literally five months ago.
Greg: Yeah. Back before Christmas old. Yeah.
Deep: Because back then, you know, this was a, this was an effort. Like you had to go and grab, grab all your training data. You had to train the models. Usually from scratch or maybe with some transfer learning, but definitely you didn't start up out of the gate with like, like a trivial, like a one or two sentence prompt that pretty much nails it.
Greg: No, we've moved all that to AI. Um, just kind of like, uh, for, for all that we use GPT four, just, that's what we use. We can go through, we do have some, some different prompts and pieces we go through. Cause one of the part of the onboarding is just give us your website or your primary social media account.
We then go. Go through, run through your, your website. We do some parsing and some scraping and then push that data through, um, the large language model to get back what this website is about, uh, what industry they're in. And there are a few industries that were like, Hey, we appreciate you loving, loving tailwind.
We appreciate you giving us a try, but this product is just not, not for you. And so, so we, we can identify upfront and just honestly, that's really, really rare. It's really rare that that ever happens. Um, but it does happen. And we do have the technology in place where used to you had to self identify. And there are a couple of those, you know, kind of gotcha categories that if you self identified as, oh, if you say, well, I'm political organization, we could then say, okay, hey, you're not for us.
Um, and, and then we moved a little beyond that to trying to do to your points and some training and trying to some some analysis there. And now it's just all up on the large language models to go through and pull it out for us and make it really quick. Kind of
Deep: amazing isn't it how, how we like something is.
Straight as seemingly straightforward as identifying, you know, hate speech was actually an incredible amount of work, even just five, five, five months ago.
Greg: Very difficult.
Deep: Now, you know, the Like, you know, the open AI folks and others like those models are just so general that, you know, you can get your training data out of chat GPT for, for like maybe a lower cost LLM pretty quickly, you know, and then we're to the point where you're just down to the boundary cases that, you know, that
Greg: you care about.
Yeah. And those, those are the ones that we do still send those to a person because it's not a hundred percent, but it's really, really close. Yeah, it's really good.
Deep: Yeah, it's it's I just can't stop, uh, getting blown away with that. Perhaps you're not sure whether AI can really transform your business.
Maybe you don't know what it means to inject AI into your business. Maybe you need some help actually building models. Check us out at xyonix.com that's X-Y-O-N-I-X. com. Maybe we can help.
Well, um, I want to end with maybe a final, a final question. I like to always ask people to kind of, um, sort of imagine we're five or 10 years out and everything's kind of evolving at the, the clip that it's at now, which, you know, can be exponential, but given what you know today. You know, put on your futurist hat and like maybe take us out into the marketing world, um, you know, 5 to 10 years out.
Tell us what, what does it look like? Is social media even still a thing? Like, are people's bots talking to bots and like the humans are just kind of like ignoring it? Like, what does that world look like, you know, in 10 years? Um, and if 10 is too hard, then 5. Well,
Greg: I'll try. I'll try five and I'll miss five.
I'm sure because I always thought Moore's law was fast and Moore's laws is nothing compared to the pace that we're seeing these these large language models. Um, so I think where we're heading. Um, I think, um, when we talk about large language models, I can't think of the character, but you know, the snake that's eating its own tail, that mythological character.
Um, There's some aspect to that. And I because I think people will stop putting their own interesting. If you're making great recipes, you're probably gonna get pretty irritated that these models keep eating your recipes and then spitting them back out for everybody. Um, so I think we'll see. I think we'll see a reemergence of walled gardens.
I think we'll see a reemergence of people not wanting, you know, putting putting their technology or their their their, um You Their content out there, but but just a little bit of it. Um, and I think they'll I think we'll see a big push towards community. Um, and and putting those communities behind logins where these language models can't go and eat their proprietary content.
I think that'll be a really big push. I think one thing, um, that that people sleep on, um, is email. And the thing is, is People think, oh, well, I have a million followers or, oh, I have 10, 000 people over here. Well, you don't, right? You rent that audience, uh, Twitter can take that audience away from you anytime they want to, uh, YouTube can be platform you if, if they want to, or not even that they can just change an algorithm.
I mean, those of us who, I mean, we're, we're probably about the same age. I kind of got into this, but around 90, I mean,
Deep: they Google Google's text. Yeah. Millions of businesses at some point, not intentionally. There's no evil intent. It's just the new the new machine learning on top of all the teams machine learning.
Yeah, but out a new thing and your stuff's no
Greg: longer ranking. So yeah, an algorithm change puts puts creates winners and losers every time. And I think what we're going to see is, um, I've seen the the internal some internal studies from some pretty big companies. That they believe that these concepts like chat GPT these as opposed to going to search having like direct answers.
You know, I'm asking you a question. I'm getting an answer back, um, will cannibalize somewhere. I've seen the numbers somewhere around 50% of all search traffic will go away over the next 24 months, even if that's half right, that will create a whole new set of winners and losers. And so I think what we're going to see, I think we'll see the rise of influencers more so than we have people being able to drive traffic specific traffic because they have their own community, and they're not beholden to kind of the search engines or, or the different SEO tricks and techniques and people trust them.
They follow them, whether it's a twitch streamer on gaming or some, some stay at home mom that's gotten really good at recipes. They love that person. I think that person is going to be the person selling. Cookware that person is gonna be the one selling gaming controllers, and I think we'll see a shift in the way that that we were advertised to, um, a lot of times.
I don't think we think of search engines is advertising necessarily, but that very much is. I mean, that's how businesses survive is they have high ranking. So, oh, I mean, there's
Deep: there's hundreds of billions, if not, uh, dollars going through that that ad platform like 100%
Greg: I don't think so. I know for a fact this is not some inside information because Microsoft has said this and screamed it from the rafter rafters.
They they lost search and they will spend and stop at nothing to try to win this. And so I think there'll be interesting. I think we're going to see a big Google, um, Microsoft, you know, being versus. Google search by and I don't think it's going to end up. My guess is I don't think it's going to end up with 1 clear winner.
I think I think we'll see a much more fragmented search space, especially as we move more towards digital assistance. I mean, most everybody has some kind of an Alexis, Alexa or home or something in their house. As we get more and more comfortable with this, I think it'll be very normal to say, you know, Alexa is already exists, but I think we'll get more comfortable with Alexa order this for me or Alexa bring that in, which is a whole nother set of search engine or search optimization in some ways.
Um, how do you optimize your content where it's showing up whenever somebody said you ask you ask chat GPT or whatever version chat GPT is out there in three years. What's the best restaurant to eat at in this city? How do you how do you try to start to manipulate that data and that information where your restaurants the one that's coming up?
And so I think we'll see a lot more, um, lot of innovation in that space. But if I had to say just the one thing I would go back to, I think we'll see a massive rise of micro communities. That would that would be my guess. Is we have a whole whole lot of like minded and I don't think this is necessarily good I think that's one of the beautiful things of the internet and it's it's broken to a large extent But one of the beautiful things of the internet is is I could get online And I could learn things from people all over the world from different backgrounds than me um And and really kind of challenge my viewpoints and move things around I think that's what made the internet so beautiful for the first 20 ish years of the Internet.
Um, and I think we've seen, uh, the Internet turning more and more into an echo chamber as people have realized. People react towards towards things that they're like, but I think that we'll see a hyper focus towards that and there will be some good come from that. Uh, you know, you, you have that, that, um, chef that you love their recipes and you join their community and you're able to go in and bounce some ideas and there'll be some really bad come from that.
But that would be, I think what we'll see if I had to say 1 thing that I think we'll see over the next 5 years, I think we'll see a hyper focused niche communities community. And I really think on the AI front, the AI agents, which we, we didn't get a chance to get into, but I think AI agents. Um, we'll be handling the vast majority of our tasks from, uh, I'm going it knows who I am.
It knows what I like. And it knows that I'm planning a 7 day vacation and not just goes out and books an itinerary for me. It creates an itinerary for me comes back. I approve it. It goes through and books and plans it all out for me uses my airline miles
Deep: to yeah, like a lot of a lot of the stuff we're seeing with auto GPT a lot of those experiments.
Yes, gone. You know, you have like a long arc thing to solve, like plan my vacation, which involves a lot of queries, a lot of reading, a lot of distillation, a few like follow up, uh, clarification, like clarifications, uh, you know, much more so than. Um, yeah, I mean, I almost it's almost like you could just imagine, like, a much deeper personalization lens put on top.
In some senses, we're kind of coming full circle, like, take something like travel planning. Uh, we used to have this place. I don't even know if they're still there called, um, they were called the council travel or something like that. And you used to walk in there as a when, like, young people would go in there and they're doing a backpacking trip.
And you would sit down and you know, the person working there just knew everything, every little quirky thing all over the world, every plane trick ticket hack, like, and they knew what your budget was and they would just figure out everything. It was like amazing, you know? And then at some point. We've got into the hell of having to book our own tickets.
And it, I feel like we'll probably just come full circle back to that kind of, um, elite kind of concierge experience is kind of what I, what I tend to call it, where you've got. You know, people that can get you what you're really after, not just these little things that you have to, you know, where you have to waste three hours in front of Google, like digging around forever.
Like it'll be able to take up a lot of that small space. I
Greg: love that, that, that term. I'm going to steal that just so you know, that elite concierge experience. I think they're a hundred percent right. Um, uh, and, and I think that's one of the great things that's coming. I'm excited about that.
Deep: That's all for this episode of your AI injection as always.
Thank you so much for tuning in. If you enjoyed this episode and want to know more about utilizing AI in your business, you can check out a recent article of ours by Googling How to Spot Great AI Opportunities in your Business, or by going to xyonix.com/articles. Also, please feel free to tell your friends about us.
Give us a review and check out our past episodes@podcast.xyonix.com. That's podcast at x-y-o-n-i-x.com