Articles
From React to Svelte - A Devshop's Journey
In this episode of Dev Shop Stories, Josh and Kai discuss the capabilities and applications of AI technology, specifically ChatGPT. They share their first experiences with the software and discuss how it can be used to create summaries, ads, content, and even generate stories.
[00:00:00] Josh: Welcome to another episode of Dev Shop Stories. My name is Josh and I have Kai here with me. And today we're gonna share a story about AI and ChatGPT and Midjourney and all that kind of stuff. It's pretty, pretty amazing technology when you get down to it. So the story begins with my first introduction to a kind of ai.
software that just kind of blew my mind and that is kind of playing around with Midjourney. I remember when that was kind of first talked about in it. I think it was only a few months ago, to be honest. when it started becoming popular and I got into the Discord channel, where I can actually ask it to make my images of a certain type.
I just saw it kind of on the news or the buzz around it and my experiences, I remember just going on there and asking it to make me, a picture of a castle. And it, it did, and it actually was really cool. And then next I asked for it to make me a picture of Donald Trump riding on a velociraptor holding AR-15s, you know, and, um, What it didn't quite get it right.
What it made was it made a velociraptor that looked like Donald Trump. So he had kind of had his, his combed over hair and just kinda the blondish kind of thing on a velociraptor. And then, um, when I added back in the guns, it, you know, it just, it was like holding them and stuff. And so it wasn't quite right, but it was pretty funny to say it'll take that one and, and make me a high-quality version of it and stuff.
And it was, it was actually pretty cool. So, some of the interesting things about AI are just the possibilities and the capabilities that you have with it and the things you can do. And so Kai, what are your kind of thoughts on some of that?
[00:01:40] Kai: I think this is something we've been expecting in the industry.
You know, being in a tech that was coming down the road. I think one thing we didn't see coming was to see how far it is along the process. I started playing with ChatGPT just cuz I started seeing different videos of people talking about it and talking about how it's just an amazing AI tool. And you know, I was skeptical but I started playing with it and I was like, wow, I can't differentiate this between like a human being.
[00:02:06] Josh: Right. and I remember when I started playing with it, I thought it was pretty cool. You know, you ask the questions, and it gives you instant answers. you can give it very specific directions. You can say, I think, you know, one of the first things we said was, you know, write me a thousand words story about, three bears and the desert and a family
You know, like that was it. And it actually. Put out there. And it was actually an interesting story about how they went and they got lost in the desert and they were saved, and, and it tied like back into a conclusion and you're just like, how the heck is this even possible? But then I mentioned it to my son who's currently in college, and he's like, oh yeah, I've been using that for my assignments.
like, well, like what? He's like, yeah, I've, my last three assignments, um, you know, I, I've used it and I've gotten a's on him. And by the way, He signed up and he got, you know, registered as an account and they, they have some verbiage in there that says like, these are generated and they're your intellectual property.
You know, so they're not plagiarism and, whether the school really looks like that, is a different story. But I, I just thought it was, So fascinating that he had already started incorporating that into his lifestyle at college.
[00:03:17] Kai: I think the big reason we wanted to talk about this today is it's not often you find technology like this that's so transformative.
Like I think this is really gonna change how a lot of different industries and the everyday lives of people will. Operate.
[00:03:31] Josh: Right. And, from the point of where it's currently at right now, it's ChatGPT 3.5 is kind of what they're calling it. And it was trained against only like 75 gigabytes of data, like 175 billion different articles and content and stuff.
and 75 gigabytes might sound like a lot, but it has nothing compared to like what the entire internet kind of holds, you know? they're talking. ChatGPT 4.0 that's coming out, and that's gonna be essentially open to the entire internet to do its learning and training and stuff. And it's, so, it's fascinating to see what's gonna happen and change from, you know, three five to four to five, you know, and, and so on. Right?
[00:04:09] Kai: Yeah. one thing we've noticed or I've noticed is I think the value of these AI bots or engines or, or tools is essentially gonna be your data sets. Open ai, they created ChatGPT I think they have a version called Dolly that does images as well. anyways, they provide an open API for anyone to be able to create tools and things off of so that you can actually use those engines and create your own things.
And feed it data sets and, and train it on your own kind of platform or your own company really.
[00:04:41] Josh: Right. And going back to kind of transformative technology that can kind of change the way, some of the big tech companies that aren't. As involved in it. Like Google is kind of scared, you know, they're, I think they're, they're looking at it and be like, you know, it might get to the point where people are sick and tired of searching on Google for a recipe only to be blasted with ads and stuff.
And then even, I know when I search for a recipe for like brownies, right? You click on it, you get to see the surf first results, you click on those results and. Read through, you know, six or seven pages, first of a blog article about somebody who's telling us a story. This is a favorite grandma's recipe and this and that, you know, and then you have to scroll and scroll and scroll and, and, and then, you know, as you're scrolling, a popup happens and, and stuff.
And then it's just like, I just wanted this recipe. If you go into a ChatGPT and say, gimme a brownie recipe. It just spits it out there, you know how to make it. And then you say, okay, turn that into a grocery list. And it takes that and turns that into a grocery list of things you'd have to buy.
Right? Absolutely. And so, you know, this could really eat away at some of the people's need to go actually to Google to ask these types of questions. I would much rather type, you know, have an app on my phone that is just my QA app, and I just ask it, you know, it's like an oracle, you know, like, give me the answer to this.
And it just gives it to you straight. It doesn't give you, you know, 1 million search results.
[00:06:01] Kai: See, this is kind of where it's gonna start to make ahead, in my opinion. we're gonna take everyday questions that we have, especially knowledge base, things that are pretty well known across maybe certain sects of people.
Getting that data is gonna be really, really, really easy. I think where some of the limitations that come with AI are gonna be generated by meaning things that we don't know about yet or new concepts, new ideas. I mean, ChatGPT or any AI engine is pulling from a data set, you know, that exists somewhere.
It learned from a couple of different iterations. So anything new? That's the future.
[00:06:36] Josh: Yeah. let's take a minute real quick and just kind of talk about some of the uses of ChatGPT and how we've, how we've at Red Sky Engineering have actually used them as well. one. An example that I thought was really interesting was the summaries for these podcasts that we're doing right now.
we just had to take the entire transcript, drop it into ChatGPT and say, summarize this. And it actually comes up with a paragraph summary, you know, of the entire podcast. And it's like, that is awesome. That's a perfect use case. For example, a mundane thing that you'd have to do and try to come up with a title now.
You might think about it like we might need to optimize it for SEO or try to catchiness or whatever, but it at least gives you a starting point of something that is grammatically correct and has the key points pulled out into a summary.
I
[00:07:23] Kai: think that's a big thing for us in sales and marketing.
Even when we, you know, create a new ad or create a new case study, we can use, some of these AI platforms to create a bunch of different iterations and then we can look through these and maybe select from a list of 15 to 20 different options and start to form our kind of versions of these ads and, and it makes our ads and our content.
A lot better. It's more, directed toward the user. A lot of mistakes that are just over that we normally kind of walk through are cleaned up, and so that's a, I think that's one thing we notice is just improved product.
[00:08:01] Josh: Right now. I know, several people are starting to use it. ChatGPT or other AI engines for actually writing article content to try to rank higher in SEO And Google's trying to combat that by having, AI detectors and, and that kind of stuff.
And so, the current thought is, you should. Use it to kind of generate, that, but you kind of have to write it, rewrite it, a little bit of it in your tone, in your personality, and try to give some differentiation, maybe add some more infographics in and stuff to actually get those articles to rank higher.
[00:08:34] Kai: I mean, one thing we, we talked about was, you know, ChatGPT great, but then, the problem is it's so confident in everything it says. It can be completely wrong the information it's giving. Maybe some of the instructions or whatever. And you gotta go in and you gotta clean that up and make sure everything's correct.
[00:08:53] Josh: Yeah, I, I heard that it actually will do, fake references, you know? Oh, yeah. It, it'll basically say, you know, write me an article, give me three references. And the references sometime will just be, you know, microsoft.com/how to do this. But if you go to that link, it doesn't even exist. Error 4 0 4.
Yeah. It was basically, you know, faked out and stuff, which is really interesting too.
[00:09:15] Kai: So if, if you're looking to use these AI tools, I would look at 'em as that a tool to help improve whatever you're doing, whether that's streamlining the mundane tasks or improving your, your overall or product that you're giving out to your client. It's, it's not meant to replace you. And I, I don't think it will ever replace, a user because again, we're going back.
It's training off a data set. Anything new, anything groundbreaking. It's not gonna have,
[00:09:45] Josh: I'll give another great example that I think was, was awesome for my next-door neighbor, bless their hearts. They are the Worst technologically adept people ever. You know, like I remember one time I had to go over and help them with the printer cuz it wasn't printing well, it wasn't even plugged in, you know, kind of thing.
and so, you know, the, but, oftentimes in our. Religion, they ask members of the congregation to come up and give talks on Sunday, right? And so that person had to do a talk that week and I'm like, well, you should just have ChatGPT write it for you. And so I, I showed 'em, they're like, well, we heard about that and stuff.
And I showed 'em and basically, I said, you know, write a talk. You know, a church talk on principles of Jesus, whatever, and, you know, cite three stories and do that, and it just printed out the whole thing and you read it and you're like, man, that's something that sounds like somebody else would've given at the pulpit.
Right? so, You know, she's like, she says to me, well, just so you know, my entire speech would be for me, but next time I'm gonna definitely use this tool, you know, kind of thing. So, really interesting use cases for, that. one thing that was interesting that I thought was, is that Microsoft recently announced a big investment into open ai, the company that kind of created the ChatGPT upwards of like $10 billion.
[00:11:04] Kai: Yeah, there's a lot of speculation of what they want to do with that investment, but they, they view it as the future.
[00:11:10] Josh: Well, if I was Microsoft, you already know you're behind in the search, the goo you know, the search engine wars, right? Yeah. I mean, Google is the big behemoth. People aren't going to bring to search for it.
But what if they were able to come onto another transitive technology and be the leaders there? I mean, I can see why they would actually be like, let's get out of the, ahead of the race in this one.
[00:11:31] Kai: Yeah. And with the introduction of GPT 4, which, you know, we can talk more into, I mean the technology is gonna get better.
GPT 3, I think Josh, you were seeing was trained on 75 gigs, of data. Mm-hmm. , whereas GPT 4 is promised to be trained on the entirety of the internet. Itself.
[00:11:53] Josh: Yeah, it, it's extremely different. a lot, a lot more capability could be coming from that. So, what now I've used AI recently over the last probably six months for coding and, um, not ChatGPT, but it's another.
Microsoft product called GitHub Co-pilot. And what they did was they actually trained that. So lemme take a step back. Microsoft, I think it was a few years ago, purchased github.com, the largest, website that holds open-source code repositories for the world. And so they, they bought that and people, when they post on there, they, they say what type of license they have and whether it's just free and open and, and everything.
in that regard. And so Microsoft used their chatbot to run across their now-owned company's software and their source code and stuff. and anyway, so now you have this co-pilot developer, and it really feels like that, like when you're actually coding it's a paid service. It's about 10 bucks a month.
But as you're typing into your code editor, it's giving you suggestions and prompts, and some of the things that it does are just, it blows your mind of just like you, you start typing and it finishes a thought for you even before you get there. one example that I have is sometimes you have to create, a bunch of like, let's say calendar dates, right?
And so you'd write January, February, March. And you normally have to type all those things out, you know? And what you just start typing January and the co-pilot says, well, do you also want to have February, March, April, you know, all the way through, and you just hit tab and it just completes the whole thing, you know, for you right there.
But let's say you didn't wanna spell the whole name of January out, so you just say j an n, period for January, f e b, period. And then it just says, oh, do you mean. also March or m an r period and blah, blah, blah. All the way out, you know, just, saves you time from typing all that. It's just, it, it just blows your mind every time you try it.
[00:13:50] Kai: And I wanna say too, we at Red Sky Engineering have been using similar tools to help streamline some of our processes. Mm-hmm. , you know, we, we find that a lot of projects have a lot of the same starting. Josh, you can talk more to this about Resta and our backend builder.
this is not something that's new. This is something we're excited about cuz we're already kind of implementing this.
[00:14:15] Josh: Yeah. Yeah. So, Developers by nature are lazy. Yeah. And they, they don't wanna do the same thing over, you know, that's not exciting to type the same words into your editor over and over and over again.
So they invent tools to help streamline their lifestyle. Um, obviously this GitHub co-pilot is one of those tools that was created by developers to kind of, you know, remove the mundaneness out of it. at Red Sky Engineering we found that. Almost every single website or web app we did had kind of the same origins.
It's just the path that it took after it got to a certain point was different. And so we actually created a tool that would auto-generate all of the standard stuff you do, like authenticated a user and check permissions and connect to a database and you know, do all these things that are, that Required on a standard application. And so we made a tool that would just auto-generate, you know, essentially a hundred plus files, 10,000 lines of code and just, you know, save all of that for that, those steps for you. And you know, if we didn't have that, it would probably be, you would go to an old project, you'd copy your code, you'd paste it into this new project, and then you'd start cutting stuff out and removing things that you didn't need and, and all that.
But what that does is it makes it air prone, you know, you. Copy, inevitably copy something, but you'll forget to rename it. And so then you have to go through those errors. So by making a tool, it made it a lot easier. Now we are taking it a step further and saying, well, all of those things that we do, there are again still similar, so let's make a little engine that you just feed it the data that you want and it will.
Automatically connect to the database, and automatically generate, your SQL queries that have to go and pull that data. and let's build a nice little UI around it. So you just kind of click some buttons and it just, you know, figures it all out and, and does that for you.
[00:16:08] Kai: So Josh, you know, we're building these tools and obviously ChatGPT here, maybe some other tools will come out.
Do you think this is the end of development as we know it? Is it being replaced by AI tools?
[00:16:21] Josh: Yeah, I'm brushing up on my, you know, yoga. I think I'm gonna become a yoga. Didn't do that. no, I don't, I don't think so. I think, um, it, it's gonna be a transitional state. I think the way that you used to code is gonna change and, and that's kind of, That's kind of happened multiple times though.
So anybody that's ever been stuck in their ways with how you do something, in particular, will always get left behind. So you just have to morph with it. For example, you know, before the internet you had to use textbooks. You had to. You know, basically, go to libraries and figure out how to do stuff, how to go to meetup groups and, and talk to your, your fellow coders in person and stuff to try to figure out the best ways to make a machine do what you needed to and stand out, spend all that When the internet became proliferate, you had to learn how to become a good Googler. You know how to synthesize what your question is and find the answers and paste them in and stuff. What I think's gonna happen is you're gonna have to be good at learning to use these AI tools to help. You know, solve your problems.
And again, with the example of the co-pilot, your job isn't necessarily so much anymore to just kind of type out the required code because it's gonna figure that stuff out for you. your job is now gonna create and do the things that are gonna be more difficult for an AI to take over right away and add that component in.
So, no, I don't think developers are just gonna go away. I think they're just gonna have to morph with the times and, and use these tools to their benefit.
[00:17:52] Kai: You know our, our last episode, you know, we talked about layoffs and how the industries. Going through a change right now, we're seeing a lot of people let go, and companies kind of tightening the belt.
If I was a person that was gonna get laid off, I think I would spend a lot of my time learning more about ai. I, I think this industry's fairly new. Nobody really knows how to use it in a business setting. At least not yet. I think the minute that some of these big companies do, you know, Microsoft's investment, maybe, Google or someone comes up with their own AI tools.
There's gonna be a lot of opportunity.
[00:18:27] Josh: Well, I, I, I completely agree with that, and I actually wanted to share, a really cool use case of ai. we recently hired a sales guy, and it'd be cool to automate not having to do you know, the cold calling and all that kind of stuff while a company out there actually did that already.
And this is Almost two years ago now. I have this recording that I think is awesome. Essentially what it is, is it's a company that was trying to get more people onto their software platform, and Kai was actually there, part of this company. So maybe you can describe this a little bit better, kinda what they were doing.
[00:19:03] Kai: Yeah, so the pitch was, is this company came in and they pitched, Hey, we work with a bunch of phones. Um, specifically in India and the call center where they talk to, thousands and thousands of customers. We've analyzed those calls and we've created an AI engine that basically listens to people's responses, their questions, everything, and we're able to, with 90% accuracy, give a good response.
Set up a call. So this is super great for any company that's looking for like a setter closer model in their sales pipeline. Like just setting the appointments, then passing them off to someone more experienced.
[00:19:43] Josh: And the amazing thing is I think they took a, a sample recording from, you know, a, a native English-speaking person.
And kind of chopped it up into standard questions, standard responses, standard reactions, and it's listening for the prompts from the customer, potential customer, and choosing the best response. And it's near immediate, so it actually feels fluent, like you're just talking to somebody normal. And so I'm gonna play a kind of a clip here of this and just kind of listen to what it, what it sounds like.
[00:20:13] Call Voice: Hello. Hi! This is Brian. I'm calling volcanic retail. Could, could I speak with someone in either marketing or sales, please speaking. Oh, hey, this is Brian. I'm calling volcanic retail. I'm sure you're busy, so I can be super quick for you. How are you doing today? Fine. Good.
Well, you know, I'll be quick for you today. I just need a moment of your time. First of all.
[00:20:38] Josh: so what I, what I kind of pick up right there is it, it waited for the response and then it, it immediately responded, which somebody listened on the end of their end of the line, they're not gonna know that that is like an AI bot that they're talking to.
Right. fascinating and. Just lends credibility cuz I've, we've all answered those, those robo dialers. And you just know immediately like, okay, this is, this is not a real person, you know? So I'm, I'm gonna fast forward a little bit, but the next kind of segment that they're trying to do is they're really trying to get to the point where this person is a warm lead, meaning that they're willing to kind of engage and talk a little bit more.
And so what he does is this, this AI bot is, it passes it off onto a real person that's gonna take over and try to close the deal. Right. any comments on that Kai?
[00:21:27] Kai: Yeah, I think that the magic here is the bot. Being able to recognize and give a good response, but then notice, hey, let's try to lead this customer down this funnel, and at what point do I need to bring someone else in?
[00:21:40] Josh: Right. Okay. Let me, lemme grab this here.
[00:21:44] Call Voice: Okay. Awesome. That's perfect. Great. You know, do, do you have a couple of minutes for us to explain the platform? We, have someone right here that can walk you through it. Would, would that be okay? Awesome. You know, I'm gonna send that email over to you right away. It outlines exactly what we need from you to set up your account.
Now I do have my product and demo specialist available right now. If we can have a quick, you know, just 5 to 10 minutes, can I bring them on the line? Would, would that be okay? Yep. Awesome. You've been great. We appreciate the opportunity to show you how volcanic can help your business grow. Now, Logan is coming on here.
Logan, are you, are you there? This is Logan. Huh? This is Logan. How are you doing? Good. What's up? Good. Yeah. Thanks for taking a moment of the time with me today. So what's your brand? Scully Customs, Air Ride Suspension. Okay, so what trade shows are used to going to.
[00:22:37] Josh: so basically he was able to kind of transition right over ask a couple more engaging questions and then close out the deal at the end and, and get them to sign up for their platform.
a really unique way of using ai, I think.
[00:22:51] Kai: thanks for sharing that story, Josh, and sharing that recording. I think it really speaks to the power of AI and really it's potential of it. Um, I think that's one of the biggest reasons why we here at Red Sky have been really forefront.
Of trying to learn this technology, learning how we could best use it, how we can serve our customers who are, you know, building up their own business, you know, how can they use AI in their business. Um, it just opens the doors to so many different possibilities.
[00:23:19] Josh: Absolutely. I a hundred percent agree with that and we, obviously there's the AI portion of it, but there's the human element to that. In Red Sky Engineering, we believe we have a great process for leading people through that. That's just the AI bots I don't think are there yet. And so we have, you know, the human touch that'll actually go out and extract from you and help you build the product that's perfect for you.
Well, thank you for listening to our story and we'll be back next week with more stories, personal experiences, and advice on running a dev shop. Thanks, Kai.
[00:23:51] Kai: Thanks, Josh.
In this episode, we discuss the value of code reviews and what happens when a developer pushes unchecked code.
H.R. Jacob joins Josh and Tanner in this episode to talk about what it means to not be a Karl. We talk about Karl and the difficulties we faced when dealing with people who didn't suit our cultural environment.
In this story, Josh and McKay, explain the importance of having a fixed bid and a clear definition of what "done" means in order to prevent misunderstandings and frustrations.