Joshua Warren 0:06 there is a system called D script D script.com. You can upload your podcast audio to it, it generates a transcript for you. That's cool. A lot of services that do that, well, the magic of it is you can then edit that transcript. And as you edit that transcript, the audio changes. And I'm not talking about how you remove a word and you then no longer hear that word, you're gonna add words, and then you will hear those words that you never spoke in your own voice come out of that podcast Darin Newbold 0:37 today, and welcome to commerce. Today, we're excited to have you on board. My name is Darren and as always here with our fantastic host, Josh, to talk about what's happening in the world of commerce now. Now, Josh, for today, this one is a fun one, I saw this topic, and I can't wait to hear all about it. So our topic today, three ways or actually, I think we may even have a bonus fourth, but at least three ways to use artificial intelligence in Commerce today. So first up, we have a thing called any word, what is or it's at any word.com? What's that all about? Yeah, so any word has released a social post generator. So basically, it can generate marketing copy for you. In this case, you can, you can go in and we were experimenting with this earlier. And it's a little scary just how accurate it can be. But you select which social platform you're writing content for, you tell it in a few words, what this post is about, you hit generate you wait a minute or so. And suddenly you have some relatively well written copy for your social media. I won't say it is like it's not ready necessarily to just connect this straight up to your account, you're going to want to edit it a little bit. But even when we threw some fairly complex things at it, you know, some specific ecommerce platforms and asking why you would buy one over the other. It generated some solid content. And so that's where I think even today, while I think a lot of this artificial intelligence stuff, we've been thinking, Oh, this is this is not something the average retailer, Brandon, can you use? Yeah, I think some of this is ready for the primetime. So I think you definitely start using this to at least get some ideas for your marketing copy. In didn't you? Weren't you sharing with me? And maybe it wasn't this one that was a specific one that someone had use this to create an ad that was along the lines for Tesla, if I remember, right, yeah, yeah, there was actually a real interesting story recently where a group went out. And they use several of these different things. So they use something that generated copy one that generated images. And they just fed it and basically said, create a marketing campaign for a Tesla Model X. And without them having to do any editing without them having to do any suggestions or anything like that. It actually came back with a really solid example. And they combined it all. And it was an interior shot of a happy person driving a Tesla Model X with a great marketing slogan out to the side of it looked exactly like an ad you would see in a magazine. Wow. That is amazing. Well, you brought up our the second piece of this is a service that's that's called dowel EDA, ll dashi to generate photos and stuff. And we, we had a little fun with this one between the llamas and the penguins. I'm not sure which one went out. But tell us about that one. Yeah, so this is a unifying this one at labs dot open ai.com. And you there's a chance you've been hearing about this one. It's basically one of these tools that has used machine learning to develop a basically a training set a data set to where you go, and you type in a detailed description of what you want even down to saying, I want this to be look like a photo, I want this to look like an advertisement. I want this to look like pixel art, you know, you told those sorts of things. And it spits out four or five images that is created. It's not, it's not a system that's searching through stock photos, it is creating brand new art and photography from scratch. And I gotta say, some of the people that generates a little creepy, I don't think it's quite quite there yet. However, yeah, we generated some awesome penguins wearing sunglasses laughing while walking on Rodeo Drive in Los Angeles. And it did a really good job on that. Well, yeah, and the neat thing is, is once you find one that you like, you can actually have it generate more just from that existing image. So you can kind of take this in a lot of ways. But the other thing that you you did a test on was just even generating. And even though the people may have been a little bit odd looking in a way, but generally just happy people in a store buying certain types of products. So if you're looking for maybe some basic art or something like that, it could be helpful. Oh definitely. Yeah, you can you can use this to generate your own stock photos basically of Yeah, dramatic example I did to happy women wearing sunglasses and a high end jewelry store. I will say that with faces, it seems like telling it to put them in sunglasses really helps I think Joshua Warren 0:00 Oh Darin Newbold 5:00 that covers up some of the the issues that it has. And I know there's other models and other systems like this out there that are designed specifically for recreated human faces. And those are a little little more accurate, and in some ways a little more creepy. Darin Newbold 5:14 Well, and I'm sure there's probably a price tag on those as well. But so that kind of brings us to the third item in all of this. And this one, at least, this could, could be very helpful for those that either don't have or may have a very limited developer group or people that are doing the actual development on on their site and what they're doing. So this one is, is the GitHub copilot. So tell us about that one. Yeah. So Darin Newbold 5:43 I feel like basically, we've talked about how the machines are coming for the marketers job we've talked about the machines are coming for the artist or graphic designers job. Now the fun part, we can also talk about how machines are coming for the developers jobs. Luckily, GitHub copilot isn't quite there. But what it is, is it's a tool that you can go and give it comments, basically programming comments and say, This is what I'm trying to accomplish. And it will actually then write the code for you. You can also just watch along and follow along as you're writing your own code, making suggestions and things like that. And the way this system works is it was trained like all these systems, it's got to be trained on a data set. And so Microsoft who owns GitHub actually went out and train this on all of the open source and open repositories that are hosted on GitHub, which initially caused some issues. There was a problem kind of very early on people realized if you would like start writing and you would name a variable, something like my Amazon password, it would then suggest someone's Amazon password that it found that they had accidentally uploaded to get up. But oops, yeah, yeah, big groups there. They've since installed some safeguards for that definitely shows how there's some rough edges in some of these systems. But I think we're GitHub copilot really shines is if you have, I don't think it's ready to replace a development team or developer yet. But if you do have a developer team, used to kind of way back when, whenever I was just a lad, studying software engineering, really common practice was a concept called pair programming, where you have two developers, basically sharing a computer sitting next to each other. They take turns where one's writing the code, and the other is just kind of watching and giving feedback and catching errors and things like that. Realistically, I don't think I've ever seen a commerce merchant have a developer team where they're spending enough money to have two developers with only one writing code, like that whole idea of pair programming. Just it's an awesome idea. It just doesn't seem to be financially viable. We'll get up copilot can replace that second developer that pair program scenarios only amplify each of your developers by having this machine learning machine intelligence following along with what they're doing. Darin Newbold 7:56 Wow, that's interesting. That's, that's very powerful. That's got to create some, potentially some challenges in I don't know, copywriting and trademarks, and those kinds of things. Are you seeing any of that? Or is there? What are the worries there? Darin Newbold 8:13 Yeah, there's definitely some concerns there. So you could definitely, you got to be careful. It's not necessarily a proven part of the law, yet. There's a lot of people that think this is going to fall under fair use. But I've mentioned a few times that all of these systems are trained on a data set. And lots of times that data set is just coming off the web, it's just been people scrape the web and use that as input to train these things. Now, when I say train these things, it's not then going to say, Okay, you asked for a laughing Penguin, I'm going to find somewhere on the web where somebody has an image of a laughing penguin. And I'm going to copy that exactly what it's saying is, hey, I've seen hundreds of images of penguins, I've seen hundreds of images of people laughing. I'm going to combine those two through the magic of neural networks. And I'm going to output something that I think matches what you look like. So I think for a lot of people, you can look at that and say, okay, the machine is generated that the machine I guess sort of gets credit for that, not the authors of the data that was trained on, but the authors of the data it was trained on definitely could come back and say, Well, wait a second. No, I have some claim to that image that was output because it was trained on on what I've done. Again, when kind of the the proponents of all this think that this falls under fair use that it was, you know, information shared publicly, they're not copying it. Exactly. So they kind of feel like fair use doctrine applies. As far as I know, that hasn't been tested in court yet though. Darin Newbold 9:41 Well, yeah, I mean, but I can only imagine that's, that's coming down the road and something for us. We neither one of us are lawyers, nor do we play them on TV. So we would definitely encourage any of our listeners to take a look at and seek the proper counsel, legal counsel for any any use is on these things. One of the things I was gonna bring up, I just thought for you is, I believe I had read somewhere that a lot of this machine learning came from all of the captures, that are out there that we've taught machines, what a bridge looks like, what a chimney looks like, what a stoplight looks like those different things is that somewhere some of this machine learning and AI really comes from Joshua Warren 10:23 that is definitely some of the training, especially Google with their capture system for a long time. People have known that they're using that to train a system. And we've gotten to the point to where a lot of people are theorizing, and maybe this is even public knowledge at this point, that it's not that we're training the system anymore. It's that we're testing the system and the system is testing us. So Google is looking and saying, okay, you know, Darren said, there's three chimneys in this image. This artificial intelligence we trained says there's four, who's right, who's wrong kind of testing the accuracy. And I did want to mention, there is one system that is coming soon. That's really interesting. And speaking to that. Speaking of, you know, potential legal concerns, I think it's almost everyone's favorite social media network to beat up on in court and or to have controversies about but Facebook slash meta has released will have it released, they have debuted and demoed a system called make a video, where similar to these other systems, there's a text prompt, you say, hey, I want a video that looks like this has this and this happens in the video. And artificial intelligence makes a video for you. That one is not available to the public yet, so can't use that one in Commerce today. But you can definitely use it soon may have opened up signups to allow you to express interest in using that one. Darin Newbold 11:40 Interesting. Well, that's very interesting. And then kind of the bonus one that we talked about here at the end. And this could create some questions for you. And I'm maybe even on this podcast. But tell us about D script. Joshua Warren 11:53 Yeah, so first, first, they were coming for the the marketing jobs, then they were coming for the designer and graphic design jobs, then the developer jobs, then the videographers. Now they're coming for me and you dare and there is a system called D script script.com. You can upload your podcast audio to it, it generates a transcript for you. That's cool. A lot of services that do that. Well, the magic of it is you can then edit that transcript. And as you edit that transcript, the audio changes. And I'm not talking about how you remove a word and you then no longer hear that word. You can add words and then you will hear those words that you never spoke in your own voice come out of retinol, just Darin Newbold 12:36 wow, how do they do that? Joshua Warren 12:38 That is some magic they have on their end of just modeling a lot of different voices and then using the data that you yourself are provided about your voice to generate basically a model that can then mimic you. Darin Newbold 12:53 So can we do this podcast is like Captain Kirk and Leonard Nimoy. I mean, you Joshua Warren 12:57 know, that is the other thing I've seen you can do with it is yeah, you can actually change the voice. I think the most popular one I've seen so far is one of those Movie Preview announcer guys, and I think that could be a lot of fun. Maybe that's how we'll do the intro for our podcast. Well, then Darin Newbold 13:13 I don't get to use my awesome radio voice. No, that's true. That's true. All right. Well, this has been quite an episode where we've, we've gone through a lot of different things that can really help out from a commerce standpoint. Yes, some of it can be a little funny, maybe a little bit crazy in some areas. But obviously any tool used for good, we can definitely make a big difference. Any last thoughts on on these 334 different ideas on the Commerce world? Joshua Warren 13:39 Not necessarily about these, but I would just say keep your eyes and ears open. Because really the past year there's been an explosion of these tools. We've hit a tipping point for sure. So I think, you know, looking forward 612 months, there's going to be amazing what else AI and machine learning can do for us. Darin Newbold 13:57 Absolutely. All right. Well, artificial intelligence is here to stay. And we have definitely found some ways and for commerce to make a difference. As always, thanks a bunch for joining us at Commerce today. Please hit hit the like button and grab a subscribe if you have a comment. We'd love to see those as well and as always, have a great day and best wishes for your commerce future.