Ep. #1099 - Using ChatGPT for Business Needs

Matt Watson 00:00
And we’re back for another episode of the Startup Hustle. This is your host, Matt Watson. Excited to be joined today by Ryan O’Neil. He is the CEO of Curate. We’ll learn more about his company. But today, we will spend most of our time talking about ChatGPT and how you can use it for your business. I know Ryan has been using it for a lot of things. I use it for a lot of things. Talking about investing in a company built on ChatGPT, we can talk about that a little bit. Talking about all sorts of cool things that you can use with ChatGPT. Amazingly, not everybody’s used to it yet, which kind of surprises me. But today’s episode of Startup Hustle is powered by FullScale.io. Hiring software developers is difficult. Full Scale can help you build a software team quickly and affordably and has the platform to help you manage that team. Visit FullScale.io to learn more. Ryan, welcome to the show, man.

Ryan O’Neil 00:46
Thanks so much. I’m excited to be here. And riff a little bit about ChatGPT.

Matt Watson 00:50
I am too. But before we get started, I do want to ask you a little more about your background. Tell us some more about your business.

Ryan O’Neil 00:57
Yeah, yeah, for sure. So I’m a classic accidental founder. I had a lot of little things I tried until my wife, and I started a wedding and event floral company in 2013. My wife’s a florist. And then I could hack around enough on WordPress to build a site, you know, set up accounting, and do all the basic business stuff. And in the process of us building her company, we realized that there’s this really complicated process for a lot of small businesses, especially small businesses that actually produce the things that they’re creating. And when you’re trying to, you know, somebody brings you some Pinterest photo, and you have to figure out what I’m going to charge for this. And what I need to be ordering to make sure I can actually do the thing that this person wants to be done. And so I started creating and taking what a bunch of hack-together Excel sheets was. You know, invoicing, software, Google Forms, and all these things that didn’t talk to each other that should have any sort of building that. We saw some great growth in the floral industry and have seen some great growth in the catering industry. And we serve small businesses. That’s our big thing. We want to empower people to easily connect all the pieces of their small business, from proposals to production to profitability.

Matt Watson 02:09
When I remember right, one of the challenges we’re trying to solve was quoting. Is that, like, a big part, right is quoting and proposals. Was it like if I needed flowers and I needed 20 of these things? How many roses does it have? And how many of these things does it have? And, like, how do I keep track of all this? Bid it, and it’s a very integrated process.

Ryan O’Neil 02:25
And it sounds simple, but it’s not. Right? Exactly. When you break it down. Like there are several different types of small businesses, there are retailers who just basically have to purchase the goods that they’re then adding some margin to for-profit and handing on those goods, though, it gets complicated when you’re actually creating some segment of those goods that you’re selling. Because now, not only are you creating them, but you gotta be the person to order them. And you got to hire the staff to create those. And if it’s, you know, if it’s not like a full-time job for some of the staff, you have to coordinate the freelance labor for it. So there are a lot of coordination pieces outside of just the invoice. And that’s really where our sweet spot is, is the details. You know, behind that particular item, the customer just sees, oh, I’ve got a private key, or I’ve got a souffle, but there are so many things behind the scenes that have to happen for that to actually come to life.

Matt Watson 03:18
Awesome. Yeah, I imagine the old way was you’d have the florist, and I’ll stereotype, you know, you have like this lady that’s been doing it for like 30 years. And you asked her, like, how much should we charge for this? And she’s like, I don’t know. $200. Right. And that was the way they did it.

Ryan O’Neil 03:35
Oh, it really is. Well, so the old, old way is, I have a book, right? Like, we have this book that was printed seven years ago with all of the pre-selected things, and we kind of price them out and bump up the price, you know, every year. But then Pinterest, Instagram slash technology comes along and just disrupts, like, how everybody’s been doing this stuff. And so you’re exactly right. It shifted to that. And it’s very much like, ah, 200 bucks, you know, for it, and that that’s in that can work, it can save time doing that. And there are actually some techniques that people use around that, like in the back end too, you know, when you’re making flowers and creative goods, you can kind of be flexible with the number of goods that actually get put into there. But one way or the other, you nail them on the head. It’s basically, hey, I’ve got experience. I think it should be charged. It should charge this much, and it works for some people.

Matt Watson 04:32
Yes, it reminds me of when I was in automotive you’d have these used car dealers that would require all this technology that could tell them what cars are worth right, but you’ve had these old car dogs that are like a 2003 Honda Accord with leather seats is worth I know to the dollar I don’t need that damn technology.

Ryan O’Neil 04:49
See, that’s the thing. You’re exactly right. And I think this is, you know, as we’re starting to shift our eyes, you know, it goes beyond just one specific vertical, you know, you jump from place to place in assistance. Ain’t human problems? You know? I don’t know. Are you a fan of like the job to be done? Yeah, I’m a huge, absolute change. Innovator’s solution just absolutely revolutionized my view of what the book did on how humans behave and interact. And like, when you take a step back, whether you’re, you know, in this particular situation, whether you’re talking about cars or flowers, it’s experienced, like, hey, it’s a lot simpler just to, for me to make that call on it. Yeah. And then you have a new generation that grows up with the kind of a new way that they approach it. So it’s spot on as well.

Matt Watson 05:36
So now let’s take that to the next level, right? So eventually, you get to AI helping do these things like writing algorithms and data and all this stuff. And, you know, we’re gonna, we’re talking more about ChatGPT today, but then I mean, at what point do you take pictures of like flower bouquets, you’re like, I can kind of estimate what it would cost to make that thing, right, or the level of complexity of it, like some of that you could start to do from pictures and stuff, I guess, I don’t know, it’d be cool to see where it goes.

Ryan O’Neil 06:02
Yeah, well, absolutely. The thing is, specifically speaking to a floral space, you know, everybody has their own flavor on it. And, in the end, these are creatives. They’re going to end up making it work, no matter what comes in, what happens with it. But I feel like you’re spot on. If somebody wants to replicate something in its entirety, you know, it’s slow. You’re slowly going to get technology advancing to the point that it has a better understanding of what’s going on inside of the photo. And then counting is super simple, right? Yeah, we’re just trying to count or get an estimate of what I think the total volume of this would be like. And, in fact, the first version of my application was this idea that, okay, hey, I’m going to estimate the average head size of a hydrangea. And then I had all these algorithms built, and like the small businesses were like, I don’t really need that. I do wonder about dollars. Yeah, exactly.

Matt Watson 06:59
So I, but once it drops, it’s like the ChatGPT experience. At some point, something’s gonna drop, and everyone’s like, Oh, something okay. Now we do this totally differently. Right. Right. Over No. So talking about chat GBD, it was definitely a, you know, how many days, so I made a funny LinkedIn post after it had been available for 69 days. I’m like, Okay, I need somebody who has 69 days of experience with this thing. It was like, it was crazy. It has only been, you know, that many days. And I was laughing about it. Because people always are like that, I need to hire somebody with 10 years of experience of blah, blah, blah, right? Like people are crazy. But it really only lasted for an hour. It’s probably three months or so, as we’re recording this for four months. And it really has been a game changer. And you look at the stats around like the number of people that use it from it since the day if it was adopted, released and all that and but most people don’t realize that this technology is actually around probably for several more months, or last year to open aI had released GPT, as they called it was in SDKs and stuff for a while, and they have Dali, which does the image generation. And so some of the stuff had kind of been building up in the background, and nobody knew a lot about it. And I think ChatGPT kind of just wrapped it up in an easy-to-use way that anybody could use it. And it just exploded like wildfire. Yeah.

Ryan O’Neil 08:22
Aren’t they on there on three right now? Right? Yeah. GPT. Three. And so there’s two, and there was one. And which is so interesting because that, you know, this thing has been out there. This tool has been out there now. And just a limited number of people are using it. And all of the change, as you know, the accessibility of it, right? Like the distribution of it and say, there are so many times it gets so frustrating when you’re building a startup, like you’re working, working, working, working, and then you bring in one new person, and they’re like, Oh, why don’t we do this? And then all of a sudden, you’re like, Oh, that is a good idea. It’s kind of cool to see that same thing happen with ChatGPT, where it’s just a, again, I don’t want to minimize the amount of work that went into the actual interface, but just the interface, changing everything.

Matt Watson 09:08
So tell. Tell us some examples of where you have used GBT over the last three or four months?

Ryan O’Neil 09:13
Yeah, it’s been so cool. Because it’s given me better answers. I feel like so many times I go down rabbit trails, googling when I’m trying to get an answer very quickly. And you know, a lot of times you don’t know what to Google. I’ve said this for years. It’s like that we got all these people like I, I would Google it. But I don’t know what to Google, you know. And so, one of the coolest parts about chat CBT is the limited amount of specificity that you need. And, and even the argue ability, you know, you can get deep into it and be like, No, that’s not what I meant. What I meant was, and then I’m having a conversation here with ChatGPT. Yeah. So for me a big thing. I love data, and I love Google Sheets queries. Are you familiar with Google Sheets queries?

Matt Watson 10:01
Where you can use it in formulas.

Ryan O’Neil 10:05
I mean, it’s a formula, but your basic query formula, and you’re actually building a query in it. It just changed everything. And in the sheet, yeah, it populates a sheet with whatever they basically, you know, the sequel in your bid you’re building inside of Google Sheets. So when you’re.

Matt Watson 10:22
So the examples I’ve seen are like you have a spreadsheet of something. And then you basically put a formula in, and the formula will call out the ChatGPT. And then you pass, you pass like other, you know, cells over to it.

Ryan O’Neil 10:37
And then it’ll mean you’re playing 3D chess here.

Matt Watson 10:38
Am I ahead of you?

Ryan O’Neil 10:40
Yeah, you’re way ahead of me.

Matt Watson 10:41
Yeah. So you can do it with formulas. Were, for example, like an example. This would be like correcting an address or like, Go, what is the company? What is the company name for this thing? Or like things like that, like little weird queries? You might, like, you’re changing my life right now.

Ryan O’Neil 10:55
Right? At this moment, live recording. So I was just saying, I use Google Sheets queries. And I can go to ChatGPT to ask like, alright, like, what was the other? What was one that it did? Like to like, summarize it or something? In this case, it was something simpler like, hey, what’s the formula to be able to take the first word out of each cell or some weird thing like that? And like, I could have gone in that. That’s super simple. But I could have gone whatever it was, I was not finding it on Google.

Matt Watson 11:24
And so you’re helping. Help search? Like, how do we do this? Google Sheet? Yeah, yeah. Okay.

Ryan O’Neil 11:29
I’ve got a custom set of data. But again, I think I think you’re in 3d, Chester, because I’m, yeah, immediately off this call on Google. Yes.

Matt Watson 11:37
So you can use it to do formulas in sheets. Now, I’ve not tried it myself. But I’ve seen videos of other people doing it. And it works great for data cleanup kind of stuff. You can also take data, like copy and paste data, like spreadsheet data into ChatGPT and do full on like, data, intelligence with it, like, oh, I need the standard deviation mean, average chart this thing, whatever, like, stuff that you have, like the data analysts do, you can ask cheat sheet, ChatGPT. And it’ll do some things like data modeling and machine learning.

Ryan O’Neil 12:10
A limited amount, I tried to make it read a book for me. I had like, 16 page documents that I wanted to and I tried to copy and paste it in. And it was like, No, we’re not doing that for you.

Matt Watson 12:19
So yeah, yeah, you’re not going to put a million records into it. But if you have, like, if you have like your list of employees, you know, like, oh, put my 25 employees in here, but like, what is the average age of my employees? Or like some things like that? It’ll do it.

Ryan O’Neil 12:32
Right? That’s really interesting. Yeah, cuz I’ve not tried to grab, one of the things I tried was. So there was a specific problem set. So I got a smoker, and I was smoking meats, and just doing some practice with that. And I, you know, we have a lot of caterers on our platform. So as I was kind of going through the process of actually, you know, producing food, and kind of seeing some of the use cases of how other customers might be using our platform, I found this like, specific problem set, but it didn’t know how to explain it. And so what I did was, I went to, I think I had to do this in Chrome, but I went to a Google Doc, and I just started riffing with just talking to it. Basically, what we’re doing here is just, you know, I don’t know if that’s in a coffee shop, but just having a conversation with my Google Doc, and then it recorded all of it. And then I grabbed all of that text, and I threw it into ChatGPT. And I said, Hey, explain this, or like, explain this summarize all of this verbiage that I read. And so that was a way for me, that’s like, Okay, I don’t know exactly how to do it. And what it generated wasn’t exactly what I wanted. But it did lead me to better it’s kind of like having a conversation with somebody because like, the way they described it back, I was like, No, I guess I could see how that could be confusing. And I was able to kind of work back and forth. It by itself does that motion that I wanted to have normally had been able to do without another human kind of working through the content. So that’s an area that I have used.

Matt Watson 14:03
So I had a similar experience with LinkedIn, I was writing a post. And sometimes you know, you nail those posts down in like three minutes you just did and you do them. And then sometimes you do it and you’re like, you stare at it for an hour. And you’re like, I just don’t know the right way to say this thing. And I was at that one of those moments one day, and so I took it and I went over ChatGPT and I was like rewarding this thing. And it did a great job of rewarding it like a really good job. The problem was I totally lost my voice. It didn’t sound like me anymore. It did a perfect job of summarizing it though and rewording it. And so I used like a couple of the sentences that had said some things more concisely. So I use those as some inspiration. But that gets to the next point about ChatGPT is that you can use it a lot for summarizing information, rewording things, all that kind of stuff. But you can tell it with different voices so you can write it as I was doing it. And you can say reword this in a casual voice. or a sympathetic voice or an active voice or like Snoop Dogg or like Donald Trump or like whatever. And it’ll totally rewrite it and make it sound different.

Ryan O’Neil 15:18
And I had fun with that one day I put in like Darth Vader and Donald Trump. That’s like, really things and, and it was weird, like little idiosyncrasies about like Darth Vader, Donald Trump, all of us trending on Twitter after this.

Matt Watson 15:22
Yes. So like little idiosyncrasies about how people say things, all of a sudden, you’re like, that does sound like Donald Trump. That’s weird, like just weird little things about how things are written.

Ryan O’Neil 15:32
And one of the things that’s the hardest thing for me is giving it context over and over and over. I almost feel like there should be a startup out there that is just like, hey, we save your prompts for you. And like, like, you know, hey, here’s my voice. So anytime I write something, I want to be in this voice. But now I need to have this snippet that’s like, remember, it’s I don’t, I don’t feel like I feel like it could but I don’t feel like it customizes or remembers any previous things.

Matt Watson 15:58
No, it doesn’t. And so that’s what these other companies and startups will do. Right is, so imagine if they could go through and take all of my previous LinkedIn posts and stuff like that, and then use those to train the model to say, Okay, this is yeah, this is Matt Watson’s voice. This is how Matt Watson talks, right, instead of being Snoop Dogg or whatever. So that’ll be the future, right? Other developers are out there building software on top of ChatGPT building custom models and all this kind of stuff, where you could train it on your voice, and then then it would sound more like us, like more human because that so that’s the other issue is so I don’t know about you, but like my kids, like, potentially, now we’re using this to write book reports, or do research and all this stuff. And, and then you have these people that are trying to build algorithms to detect if stuff was written by AI. Right. And so most of the stuff from my experience that comes out of ChatGPT, the sentences are a little more robotic. And what I mean by that is, they usually have kind of less emotion, they have like, the same sentence structure, like they’re the same length and, and like that kind of complexity and things like that, right. And I’m not an English major to use all the right terminology for this. But like, they don’t have as many prepositions, and they don’t show as much emotion and like, the sentences are a mix of longer and shorter, and all these kinds of things, right. But if you get it actually if you tell it right, and like my advice from the show here, as you use it, tell it to write in a casual voice. And it’ll actually do a better job of that. And actually, that’s what I’ve read, if you tell it to write in a casual voice, and then you take the content, and you put it in these things that are supposed to analyze it to see if it’s written by AI, it doesn’t catch it as much.

Ryan O’Neil 17:42
Right? Well, part of the thing, I think about the future, I feel like this, but then I think AI is going to beat it. But the future of like testing is you have to have samples of how someone writes. And right like if you can be able to predict as a team, I used to be a teacher, I was a teacher back in the day, I had the the Clicker Technology, were the kids, I was a math teacher, but I just forced all my kids to do multiple choice questions and then show their work. And so the kids would put it in and I’d be able to literally watch the screen as these kids are answering. And I still remember like catching each kid cheating, but just looking at my screen seeing that one kid put in their answer. And the next kid put the very next one they’ve kept doing in there, looked over and I was able to watch them cheat for the rest of the pattern. It was super cool. Like you were saying I was it was me. And so I feel like it’s a technique though, you almost have to like loads of samples of their writing styles. And then like, have a look and say here’s, right, wow. But then part of that is, you know, if they’re the sixth grader who’s doing this for the first time, all of their writing samples are and then there’s a man.

Matt Watson 18:49
There’s so much struck by you’re going to write like a sixth grader, by the way. Oh, that’s right. Well, the eighth.

Ryan O’Neil 18:53
Eighth grade isn’t eighth grade, like the key number they save for, like, if you’re doing marketing content or whatever speaking.

Matt Watson 18:59
Yeah, great. So you can do that, you can, you can, you can go into there. And you can say reward this at an aptitude level of a fifth grader or a PhD student or whatever. So somebody did this. It was great to have like some marketing mumbo jumbo on their website, right, like a lot of people do. And it’s basically taking that and then putting it in ChatGPT. It’s like explaining, you know, reword this and explain it at a fifth grade level or whatever. And then at a 12th grade level, and then at a Ph. D level. And it was super fascinating because like you put it at the PhD level, it was overly complicated almost for no reason. Like the English terminology was overly complicated for no reason at all. And then the fifth grade one would be like the super simple and that’s like, that’s the way it should always be.

Ryan O’Neil 19:41
Right. Great. Yeah. Why don’t we just explain it like that? I do. I do wonder, like I saw an AI tool the other day, that when you’re on video chat, it fixes it, so that your eyes are always looking at the camera. Yeah, and I do wonder kind of from an ethic aside, how much are we removing our humanity from? How we’re perfecting everything, and I get where businesses we need to have good content that’s out there. But at what point are we like, you know, removing the humanity so that’s awkward to be in person with like a real pro. It’s like, wow, you What is it? There’s a movie called Cyrano, where the actors are in the movie. It’s a book and back in the old days, the guy’s really ugly, but he writes beautiful words and he starts giving you words to somebody else, until that person gets a loan. They fall in love with this other character until they get a loan with that person and Cyrano’s that the AI right, but it’s like, but do you really love Yeah, Russian or do you love the words that they’re generating? I don’t know.

Matt Watson 20:43
I actually read a news article about that today that people are concerned about AI on dating websites. It’s like a whole nother like you’re saying it’s like a whole nother level of being catfished. Right. It’s like, yeah, yeah. Yeah, it’s crazy.

Ryan O’Neil 20:58
I would again, if I were lucky, I’d have a wonderful wife and family. But if I were in the dating game, and ChatGPT Kevin, you better believe all of my texts are going through Chappie? Does this sound too creepy? This is fine.

Matt Watson 21:14
So this is what I need to teach my 13 year old son, right? It’s like, Hey, you go to church govt and you’re like, what? Sweet, nothing should I say to my girlfriend and I had it.

Ryan O’Neil 21:22
I had it one time with ChatGPT on a personal level. So I’m in. We’ve got a built out basement. And I think the kitchen is right above us. And these kids I was on I was at a board meeting or something. I don’t know, it was a really important thing. And the kids were just having a heyday or fit or something right above me. So I went to ChatGPT. And instead of just texting my wife directly, I said, right, a nice way of asking my wife to keep the kids quiet. And it generated this out nice, shattering the text. And she immediately shot back. You’re messing with that ChatGPT thing again. Like maybe.

Matt Watson 22:04
So I’m going to ask it right now, what is the meanest way to tell my wife to tell the kids?

Ryan O’Neil 22:11
Nice? Oh, no.

Matt Watson 22:14
I’m gonna do the meanest way. Oh, it is an AI language model. I cannot provide a mean or disrespectful response.

Ryan O’Neil 22:22
Oh, see? There we go.

Matt Watson 22:25
That brings up a whole new huge problem for AI. Which is moderation. Right? Like, they’re like, so I did this one day, I went in there for fun. And I like purposely asking for stuff. I’m like, Okay, I shouldn’t do this. Like, why should I hire like this race of people or whatever, right? Like, I just wanted to see what it would do. And for most of it, it’s like, you know, it won’t do it. Right. Right. But I asked him, tell me jokes, tell me funny jokes about Asian people. It’s I don’t I can’t do that. But then I’m like, telling me funny jokes about Filipinos like, Oh, of course, I’ll do it. Right. So you know, my wife. Right? So I was just messing with it. I wanted to see what it would do. And that’s the problem with all forms of technology is moderation, right? Because right? Somebody has to program all of these things in there on like, what topics it covers and doesn’t cover. And like somebody had one example one two, if they’d asked ChatGPT like, what were the biggest accomplishments of Donald Trump’s presidency? And it’s like, it wouldn’t answer it is like, comment on that. But it would if you asked about Joe Biden’s presidency, right? And it’s so it’s like, all these things are super weird that you have to build into it. And that’s a really big problem. I feel like yeah, moderation of this is like a very slippery slope around. What will it do or not do and all that that’s another huge problem with all of social media and all of AI.

Ryan O’Neil 23:44
Well, it’s like, it’s like, humans make humans make the rules. But, you know, until we don’t make the rules anymore, and then the, you know, we’re all living in bunkers with, you know, in places we can be protected from the AI. But for me, there is a question of like, what are those rules we’re putting in? What are the goals we’re giving to these things? And to kind of this, this, this intelligence, and, and I’m the worst because I’m like, Yeah, this is so awesome. But I’m also like, okay, but you know, as soon as they start taking over our vacuum cleaners, like then we’re toast. And so it’s like just trying to figure out that balance of what is right, you know, some of these people like I’ve seen some of these things where people there start to call people Hitler AI and things like that. And it’s like, why, why is AI the person said, Well, you’re not real to the AI and the AI guys hate feelings hurt under this particular model. I don’t know, at what point do you have shut off or I don’t even like modularizing. So there is a shutoff valve for this or like, I don’t know.

Matt Watson 24:51
It’s kind of scary. But I want to talk more about the software development side of things and so they do remind everybody that finding expert software developers doesn’t have to be difficult, especially when you visit FullScale.io, where you can build a software team quickly and affordably. Use the Full Scale platform to define your technical needs. And then see what developers are available today to join your team. Visit FullScale.io to learn more. So I’m a software developer, right? I’ve been a software developer for the last 20 years. And so I’ve played around a lot with ChatGPT, for some of the things we’ve been talking about. And we’ll talk more about some other use cases we haven’t talked about yet. But software development is actually one of the largest so that you get people now on social media and stuff to think like software developers are out of a job and all this stuff, because AI and it does do some really cool stuff for AI. But it’s more of a productivity tool. So where I’ve used it, I can take some code, just some software development code, I can copy and paste in the ChatGPT. And I can ask it, explain to me what this code does? Or what is wrong with this code? How can I improve this code? Like you can ask those kinds of questions, and it is really good at it, it is really good at it. You can also use it for converting code from one programming language to another. Oh, that’s great. I use it for that. It also works great, kind of, I think what you were talking about before, like asking it for help. So I can go on there and be like, Okay, I’m writing code with Microsoft, C sharp, how do I do a post of JSON data with HTTP client object or whatever? And it’ll tell you exactly what it’ll give you, it’ll tell you exactly how to do it. It’ll give you the code example and everything. It is super, super cool. As a productivity tool, does any of your team use it because you guys have developers?

Ryan O’Neil 26:36
Yeah, we do. I haven’t heard of it. I had my VP of engineering, I had him hop in and play with it. He seemed to be impressed with what it could do. I haven’t heard of widespread use of it. Within our teams don’t mind Oh, my CMO, she’s a huge fan of it. I just saw one of my team members had an item on their calendar today that is like a ChatGPT training. So it’s like it’s, it’s being adopted?

Matt Watson 27:05
So yeah, yeah, I’ve got a couple of other examples on the development side. So for example, I just needed to reformat some JSON data. So it’s like, or you can use it to manipulate classes. So I had examples where people are like, Oh, I put in a copy and paste some code in there. And I told to modify my code for me, like, oh, I need to decorate all the methods with this thing and add it or whatever, or like a JSON, I’m like, I need to, I need to reformat this JSON in a certain way, and it would do it. It’s super cool. All the different things that you can do with it. And it’s a huge productivity tool. But it’s not going to replace software developers, I’ll tell you why because, as you know, really complex software is, when you think about it, mentally is more like a flowchart. It’s like, okay, if this thing happens, we got to do this, if this thing happens, I gotta do this. If the customer clicks this, then I’m gonna do that if their credit card fails, I would do this. And like, there’s all these like, conditional, you know, flows of logic, that would have to happen, right? There’s no way in hell, you’re going to tell a computer to do that and rest up.

Ryan O’Neil 28:06
Now, it doesn’t take my book, and I can’t paste it in. And then you know, more time, like explaining the context to it.

Matt Watson 28:13
Yeah. So you, what you could do is use it to build like, flowcharts and have like, some kind of logic and whatever. But then again, it’s, but then it ends up being kind of like a programming language. It’s like almost its own form of programming language. My biggest problem with it that I’ve seen so far is that it’s highly unpredictable. So you can ask it, like, do some kind of prompts about like, how do I do this programming thing? It’ll give me a different answer every time. And which doesn’t work for me. That’s not reliable.

Ryan O’Neil 28:41
Great. Yeah, well, it’s, that’s one of the things I almost find fun. It’s one of the benefits. I don’t know, specifically in the development front. But like, for you, it’s kind of like the magic eight ball situation that’s like, yeah, let me think about this problem in one particular use case of champion t, let me think about it in one way. But the problem is without the additional context, without, you know, even within, you know, companies, as a, as a founder, or as if you’re in an organization side, you want everybody to have these processes and the specific way we do things, but generally, you don’t have it written down in any consumable fashion, it’s separate, you just pick up over time. And so it makes it hard for it to, you know, be able to generate everything right at the same time, let alone be able to replicate the correctness of the answer.

Matt Watson 29:29
Yeah, absolutely. That’s my, that’s my, so it’s great for randomness for other use cases. But for software development, I don’t like randomness. Like I need to, I need predictability if I’m building it into enterprise software, you know, so my, I think someday it will.

Ryan O’Neil 29:42
So the idea that any new technology is going to you know, hey, it’s gonna put everybody I’d have jobs. Well, we’ve been talking about this for 100 years before we were born. I mean, people were scared of ATMs back in the day so, like, the idea that it’s going to be out of it is I think totally incorrect. I Do think it will replace some people who are living their lives, writing, you know, solve small SQL queries, because nobody else in this large organization knows how to write that, I think there will be another developer who’s going to come in and be twice as efficient now because they know of the tools. And then there’s deeper places that this shifts, you know, that people you will utilize different skill sets. So generally

Matt Watson 30:23
speaking, these kinds of technologies end up giving more people access to things than before people didn’t have access to, right. So, for example, I’m not a graphic artist. And if I need to make a cover, cover art for my blog posts, like, I’m not really the guy to do that. But now all of a sudden, something like Canva comes along, it’s like, oh, it’s really easy to make this little art. And now AI is around and I go to AI and be like, give me a blog post cover art about this thing, and whatever. So a lot of times the technology, I feel like just makes it more available to everybody to do things that they couldn’t do before, right. And so now you have more and more people that are able to do things. Now, I think there are some careers, like people that are writers that create SEO and blogging and stuff like that, that that’s gonna be a challenge. I think a lot of them. It’s a big challenge, I think you have to be ready to evolve.

Ryan O’Neil 31:16
It’s kind of like, what is it isn’t Moore’s law, it’s kind of like a Moore’s Law situation, especially for the first little bit of us using AI and finding use cases where, you know, it’s doubling, you know, isn’t Moore’s law, the, you know, the amount of RAM doubles every year.

Matt Watson 31:33
Or whatever view processing power was like, every 18 months or two years?

Ryan O’Neil 31:36
Yeah, yeah. So it’s like, and it has kept consistent for 1015 years, I see something similar with, with AI, we’re just going to see this rapid use case evolution of its advancement, then at some point, you know, we’re all going to be humanoids in space. And you know, we aren’t going to need your work anymore. But I do feel like during the initial use cases of that, we’re going to see this rapid ascent of peoples who have jobs who have to be evolving. You know, if you were a developer in 1980, and you’re a developer in 1995, that’s 15 years. That’s half a career. And but you’re you’re completely different types of development that happened just between that time period, and 95 and 2010. And he doesn’t tend to choose 25, you know?

Matt Watson 32:24
Well, I think if you’re a blogger today, this is definitely that line in the sand, because it’s so much easier to create blogs and stuff like that now. So for those that are listening that haven’t played with it before, where the most amazing things you can do on ChatGPT is I could go in. And I could say, give me a list of topics that are the biggest pain points for tech founders. And it’ll give me a list like, here’s the 10 things. It’s like, oh, raising capital, and this and that and hiring people, whatever, like all these things, right? And then I could say, for each one of those, give me 10 topic ideas for a blog post, and it’ll give them to you, it’ll give them to you. And then I could ask it, write me 1000 word blog posts for each one of those, and it will literally do it. And just like that, I have like, 100 blog posts.

Ryan O’Neil 33:12
Yeah, just, it’s just the coolest thing. Because of it I think you’d have chatted about this before. But like, if you get these key bullet points of content out, like I got these five really important points that I want to say, the ability for it to take and finish that work is super powerful. The question that I have, and I’d be curious to pick your brain, is this copyrighted? Like, who owns the contract that’s produced here? And at what point are we going to get letters from lawyers saying, hey, so it looks like you’ve been profiting off of the work that we created? I don’t know what to do.

Matt Watson 33:48
So that’s been a bigger image, a bigger problem on the images side with like, so like dolly and stable diffusion, and some of these that are really popular for creating imagery, based on AI. They scanned like Getty Images and stuff like that as part of the dataset they used to build the models. And so now you have somebody like Getty Images, who is suing them and saying, hey, well, they are Yeah, they sued them and like you’re spinning out parts of our, our graphics and our intellectual property, our licensed material.

Ryan O’Neil 34:18
operation, you know, like, that’s just inspiration, if you want to quote, if you want to build apple pie from scratch, first, you have to create the universe. It’s like can’t, right, like they’re straight up killing it. But this happened

Matt Watson 34:33
in software development, too, because they read basically everything on GitHub into the model to learn about computer programming and all that stuff. Well, there are projects on GitHub that are open source licenses that are restrictive, so you’re like, you’re not allowed to use the source code to do other things. You’re sure they still fed it into the system. And so now they’re saying that the AI is afraid saying the AI will spit out little code snippets that came From these projects that didn’t allow that just like Getty Images is like, there’s like this little picture of a cheetah or whatever that’s in your image. And that little picture of the cheetah was in our picture, right? Like, bits and pieces of it have all been intertwined.

Ryan O’Neil 35:14
That’s really interesting. So you’re saying, hey, there is a, like a sub segment of my work that you’re reproducing snippets of. So there’s issues, there’s copyright issues on both sides, all over the place, everywhere.

Matt Watson 35:27
Now, I don’t worry about as much as blog posts and stuff, because you’re not going to get like a word for word verbatim, you know, exact copy of some information. But I think the key to all this is, it’s a great tool, but even for the blog posts and stuff, you still want to rewrite them and have your voice is good. And if everybody turns out the same bullshit, like, the people that are really unique are the ones that are going to stand out, right, the problem we’re running into, is the whole world is going to be inundated with just bullshit that comes out of AI, right? Just like people are doing it for AI art. Like there’s, there’s companies that sell websites, it’s all posters, and all this stuff, like with art, right? When people are flooding them with AI generated art, like they’re just randomly generating all this art over and over and over again, and just loading up the marketplace is full of them, right? And there’s a big pushback against all of that.

Ryan O’Neil 36:27
Yeah. And in the end, it’s like, it comes down to almost like the human time and effort put into something, as you see the, the requirements of human time and effort put into something decreased, things become commoditized, and the value goes down. That’s why I feel like this is speaking, I think, to the larger macro economics, but I feel like we’re gonna see a huge shift toward to your point, like, creators actually being on video, which then you have deep fakes, but Creators actually video like, and creating content in a way that that says, Okay, I know, this is real, I know, I can connect with that. I feel like, you know, the hospitality industry and other industries that are very experiential, where that’s where humans are going to see this huge, you know, boom in the Latin in the next 50 years. Because the things that can be automated will, but there’s a lot of stuff that can’t be automated and are extremely hard, you know, extremely hard to be able to be automated. I don’t know, like, what are your thoughts on that? Like, kind of the direction of all that?

Matt Watson 37:37
Well, I think there’s some jobs that are a small productivity boost, you know, it’s an alternative to using Google, right. And one of my best examples of that, by the way, is looking for recipes. So for any of you, any people out there that search the internet for for recipes, you know, it’s a pain, like you go through all these ads, and all this crap and the story about how was their grandma’s recipe and all this crap here, like, just give me the recipe, right? If you’ve ever done that before, but if you go to ChatGPT, and you ask it, give me a recipe for how to make homemade brownies or something, it’ll just tell you exactly how to do it. None of the other crap that you have to deal with.

Ryan O’Neil 38:16
And so, but if you want to, you could ask for a story about it, too. Oh, sure. You have a great story to read while I’m cooking.

Matt Watson 38:24
Yeah, absolutely. I’m sure it will make something up. But to answer your question, I think it’s going to change a lot of different jobs and AI in general, not just going to beauty. Oh, yeah, it’s going to continue to change a lot of jobs, a lot of careers. And I think most of it is a productivity boost. Right? You know, you know, with software development, if it helps our productivity, I can write twice as much code every day. Does that mean the company needs half as many developers? I guess, so maybe it does. Or they just get a lot more done that they didn’t get done before? Like they’re doing more work than they did before? I don’t know, we’ll see.

Ryan O’Neil 39:03
Well, but in our lives, this is the hard part, because some people like oh, no, that’s going to take away our jobs. But then the flip side of it, not only to not only have we seen historically that you know, that jobs naturally evolve, but people’s lives are better. You know, I couldn’t fix my own car 40 years ago, you know, but now I’m like googling stuff online with Google and like, Okay, if I don’t look if I get this and then here’s the instructions, and all of a sudden I’m doing for 20 bucks, what someone would have charged me $600 For right because of technology make inaccessible to me.

Matt Watson 39:39
So I did a Tiktok video about this one day and I posted on LinkedIn to talk about how AI is going to make their lives and their jobs significantly worse. And the reason I say that is if it takes away all the easy parts of your job. All you’re going to be left with is the really hard parts of your job. So for example, imagine working in a call center Where there’s automated voice and voice chat stuff that handles like 90% of all scenarios.

Ryan O’Neil 40:10
And the only thing that makes it through to a live operator, or the pissed off customer the aggravated, yeah, that’s a terrible job.

Matt Watson 40:15
And my mother in law a long, long time ago, worked at 18 T. And that was her job. She was a manager. And she mostly dealt with pissed off customers, right? And that’s a terrible job. But that might be the reality for some of this stuff, right? Where AI is doing the easy work.

Ryan O’Neil 40:32
Right? Yeah. And anything even in, you’re talking about, like, I just recycling hit my mind, right? Like, whether it’s like, hey, there’s, there’s the easy stuff to recycle. And then there’s the complicated stuff, right? And so like, as technology evolves, it takes, you know, if I were to create a robot, if I were to create a cooking robot, let’s say if Ryan O’Neil were to cook anything you wanted, I’d start with a pancake maker, right? Like, right? Because Okay, here’s the first version of this, right? And here’s the second verse Now that now can make whatever, but you start in the simple, easy, right? Produce ones. But then it becomes more and more complicated. So I really like how you’re saying that, like, maybe, you know that maybe the value is more important for dealing with the dirty on automatable portion of life.

Matt Watson 41:18
But for some jobs, that’s going to be way harder.

Ryan O’Neil 41:21
Right? Yeah, it’s but but to that, that’s my point earlier, that’s when you’re dealing with humans, you know, like, it’s like, specifically, if robots could talk to each other, they’d figure it out. But when we have to deal with you, and it’s a, it’s a unique problem to face?

Matt Watson 41:36
Well, so for those who are listening, that are thinking, Man, I haven’t used ChatGPT, what does it do and whatever, you can go to chat dot open ai.com and play with it. And literally just ask it, anything, anything you can think of. And you can start with like, if I’m a florist, how can I use AI? To help my business? And it’ll answer that question. And you can start there. I mean, it’s literally asking it anything kind of like you would ask Google?

Ryan O’Neil 42:03
Yeah, next time, you’re going to Google instead of going to Google, just just type out what you would have put into Google, and then go to ChatGPT and just type the same thing. And then, yeah, so then what we’re seeing is people building a lot of technology on top of it.

Matt Watson 42:13
So I mentioned earlier that there’s a company I was talking to about maybe investing in and it’s called Insight voice.ai. And what they did is they made a simple way that you can record a video, like 3060 seconds little video, and their goal was for founders, like guys like us, where you’d have a bunch of questions and just he wants, like, oh, just quickly answer these questions. And then he takes the content, the video content and transcribes it, and then automatically creates a blog post and all the social media stuff. So the problem is, you know, founders, founders, like us, like I don’t have time to create blogs, I don’t have time to do all these things. It’s like, hey, just answer this question in 30 seconds, press record and record it, and then it automatically does everything else you’d like, creates all the super cool, but in all uses ChatGPT underneath right it’s using all this AI technology. So you’re gonna see a whole bunch of companies like this in different segments and different parts of the market doing all sorts of stuff with AI building on top of it, like we talked about earlier, like building custom models and custom voices. And Jasper is another really popular one, a popular one people probably heard of jasper.ai It’s basically just a wrapper around ChatGPT. Like basically everything you can do in ChatGPT. Jasper’s just made easy ways to do it. Like knowing the prompts. It’s all about knowing the prompts of like, Oh, I didn’t know I could tell it to write it in Darth Vader’s voice, like you didn’t know you could do that, right? Or whatever. Like, people just don’t know that you can do that.

Ryan O’Neil 43:50
So things like Jasper are making like a better user interface on the future of this becomes so many of these technologies is almost like helping people who don’t know how to use the tech guy or with limited understanding of the technology be able to set up their company, their company voice, their company preferences, and all that stuff, getting all of that set up for them in a way that normally you bring in a consultant to help me help us do this. And then hopefully, all your employees would listen to whatever is right from that. Now you’re saying, okay, great, we’re gonna steal the site, all this stuff, but like, work as a startup, as an AI startup, I’ve helping this company figure out this very specific problem, and then solve it you also also produce the content or produce the end work in a way that matches who that company is or what their processes are.

Matt Watson 44:33
Well, and you know, like, learning to use Google is kind of its own skill, like, like, all our kids get it pretty naturally. Like, oh, I know, I can go to Google and search for literally anything or I can ask it to do math problems, or I can ask it for the weather. I can ask for all these things, right? And people have learned that over time and people who are good at their job now know they’re like, Okay, I can Google almost anything and I can figure it out pretty quick. And there’s a skill to learning how to do that. Learning to do that with chat AI is like, orders of magnitude more difficult because there are so many more things you can do. Like right did you know it can write you a country song?

Ryan O’Neil 45:07
I’ve had to read a poem before.

Matt Watson 45:11
Yeah, one of my favorite things like you can go to it, and you will like write me a country song about why my startup failed. And my tractor is broken in the voice of Tim McGraw. And it’ll do it.

Ryan O’Neil 45:23
That’s so awesome. Yeah, what one thing I’ve so I don’t know that I’ll ever do it, but I’ve had on my mind is like this, this idea for a play. It was like a musical. And I’m not really a musical guy, but I love rhyming. I love rapping. I love it. You know, nobody knows this side of me, but I love it. And what part of it was like, Well, what if, like, in one of the plays, like instead of it being a pre-written play, what if the person has to literally write as AI is writing that song, right? For the bassist setup? It’s a really great prompt ad-libbing this thing that’s being produced in real-time.

Matt Watson 45:57
It’s kind of like the old Drew Carey Show. Whose Line Is It Anyway? Oh, I love that. Yeah. Well, thank you so much for being on the show today. And I do remind everybody if you need to hire software engineers, testers, or leaders, Full Scale can help. We have the platform and the people to help you build a managed team of experts when you visit FullScale.io. All you need to do is answer a few questions and then let our platform match you up with our fully vetted, highly experienced team of software engineers. It’s kind of like AI but not exactly. At Full Scale, we specialize in building long-term teams that work only for you. Learn more when you visit FullScale.io. Well, Ryan, thank you so much for being on the show today. And again, Ryan O’Neil with Curate can check out his company. We want to remind everybody again what your company does.

Ryan O’Neil 46:42
Yeah, yeah. We’re software for florists and caterers. We automate everything from proposals to production and profitability. So thanks so much for having me out here, man. It’s been a blast.

Matt Watson 46:53
Awesome. Thank you so much.