FirstPic: Forever Love

Loughlin Rodd, the founder of FirstPic, joins the Behind the Bots podcast to discuss how his innovative AI-powered dating app is revolutionizing the online dating experience. FirstPic leverages Artificial Intelligence to analyze users' photos and bios, providing tailored feedback and optimizations to help them create more engaging profiles. Rodd shares insights into the development process, including training the AI model on thousands of images and incorporating best practices from dating experts. The conversation delves into the potential future of AI in the dating industry, exploring ideas such as AI-driven matchmaking and relationship advice. Rodd also touches on the challenges and opportunities surrounding Artificial Intelligence in the realm of human connections. 

Throughout the episode, the discussion covers the impact of AI on dating app success rates, user behavior, and the overall online dating landscape. With FirstPic leading the way, Artificial Intelligence is poised to transform how people find love in the digital age. Tune in to learn more about this cutting-edge AI dating app and the visionary behind it, as Loughlin Rodd shares his thoughts on the future of technology and romance.

#ai #artificialintelligence #machinelearning #onlinedating #tinder #aidating #bumble #datingprofile #datingproblems 

Creators & Guests

Host

Ryan Lazuka

The lighthearted Artificial intelligence Journalist. Building the easiest to read AI Email Newsletter Daily Twitter Threads about AI

What is Behind The Bots?

Join us as we delve into the fascinating world of Artificial Intelligence (AI) by interviewing the brightest minds and exploring cutting-edge projects. From innovative ideas to groundbreaking individuals, we're here to uncover the latest developments and thought-provoking discussions in the AI space.

Loughlin Rodd: 20% of male profiles get 80% of the right swipes from women.

Ryan Lazuka: I mean, a lot of people don't think about that, right? Like Tinder wants you to meet somebody, but they don't want you to meet somebody so quick that you're not gonna use their app.

Loughlin Rodd: They want you to have hope, hope but not necessarily result. First pick is there to make building your dating profile super easy. It uses AI to determine which of your photos are best for your dating profile, show you how to order them and give you feedback on those photos. And we didn't adhere to like traditional beauty standards. The results were based on the quality of the photo, not whether you were an objectively attractive person based on society standards. So we're giving you instructions on how to give us the details that will allow us to craft a piece of engaging and unique content that will allow you to stand out on these app.

So for everyone watching, I'm Lachlan. I started out my career in early stage venture capital. So investing in nascent technologies that affirm called Kineac Ventures and what they call what that really taught me to do was how to stay abreast of what was on the cutting edge, right? And I've always kind of stayed with that throughout my career in terms of being on things that were kind of in nascent technologies, right? And have stuck with that for, you know, since leaving there, right? And in doing that, I've always kind of kept up with what's new and what's on the cutting edge, right?

And when chat GPT came out a couple of years ago, got very like into using that and learning how to make the most of it. My educational background is in computer science. So I have a computer science degree and my day job is as a product manager at a tech company. And so I know how to code and how to like leverage different piece of technology, right? And with chat GPT, I was able to, you know, kind of do things a lot faster and was able to start like iterating on different projects and ideas I had. And that's what kind of led me to my current project, which is called First Pick.

Hunter Kallay: Cool. Yeah. So that's how I heard of you is through First Pick. Very interesting tool. I don't want to tell too much about it. I'll let you tell about it. What is it briefly and, you know, what's the purpose of it? How did you get inspired to create this?

Loughlin Rodd: Yeah, absolutely. So First Pick is there to make building your dating profile super easy. So it uses AI to determine which of your photos are best for your dating profile, show you how to order them and give you feedback on those photos.

So things like, hey, we're not seeing your eyes, take the sunglasses off, or there's too many people in this picture, try not having a group photo, things like that. And it will also take some of your biographical details, answers to a couple of questions that you give us and write your bio or prompts for apps like Tinder, Bumble and Hinge. And it started out as a photo app originally.

So the content writing was a secondary component that I added onto it. Originally, I had wanted to solve this problem of folks that I knew in my life who were otherwise like very presentable, like good looking folks who just had no idea how to market themselves on a dating app and would select like the worst possible photos of themselves. And they were wondering why they weren't getting results. And I thought it was a problem you could solve with AI. So I started digging into the different libraries out there that you can use to solve this problem. And I found out about something called ResNet, which allows ResNet 50, which allows you to basically train your own image learning model. And I trained it to learn what a good dating app photo looks like and what a bad dating app photo looks like. And from there, I thought it would also be cool if you could understand where you kind of stood in the dating market, if you will, based on your gender. I can tell you things like, hey, your photo is scoring better than 75% of male users or something like that. So you could at least know, all right, my photo is pretty good. It has a good score from this model, but also contextualizing things with how do my results compare to other users of the app, right? And then when I originally pushed that version out, I got some feedback from early users saying, hey, I wanna know why my photo is scoring a certain way.

How can you help me understand that? And that's when I layered in other elements to my model, which learned to detect things in photos. Like for example, if I detect a phone in your photo, I can make a reasonable deduction that that's a selfie, a mirror selfie, which generally speaking scores pretty low on the spectrum of quality dating app photos. And I can say something to you like, hey, this looks like a mirror selfie. Try taking a photo that's in portrait mode on a nice camera where someone else is taking a photo for you.

Ryan Lazuka: I'm sorry to interrupt, but I know Hunter is thinking the same thing I am. We have an uncle that's like, typical crazy uncle, if you think of the crazy uncle out there, he's like probably 70 something. What do you think Hunter? His name is Vic, but he does these, he's on a Tinder, all these dating sites, and he takes the worst profile pics I've ever seen in my life.

The hand up with the phone in the mirror, untucked shirts, not shaven. He shows us our matches, his matches, and they're pretty horrendous. Just some strange folks that he matches up with. So we'll have to show him this app. But your app is, it can be for anyone, right? It doesn't have to be a certain age.

Loughlin Rodd: Yeah, any age would work. It's just learned from thousands of images what a good dating app photo looks like. And so in that training data are folks of all different ages, shape, sizes. One thing I was really proud of is we had some press a couple of months ago that where someone tested the app out and in the article said that we didn't adhere to like traditional beauty standards so that they were able to upload photos of all different kind of looking people, right? And that the results were based on the quality of the photo, not whether you were an objectively attractive person based on society standards.

So that was something that was important to me was making sure that it was accurate, not based on whether you're attractive or not, but based on whether you've taken a good photo or not.

Hunter Kallay: Yeah, so it seems like there's two parts of this. I mean, there's a couple of working parts, two major parts, one is the photos and then one's like the bio as well, right?

Can you just tell us a little bit more like, I mean, go in as much detail as you want. How do you figure out which photos are effective and which ones aren't? Are you like, cause like how are you pulling the data for this? Like you're actually going on to like statistical websites or you just like estimating that like high quality photos are gonna get more views. How are you pulling in those estimates?

Loughlin Rodd: So your question is around how we're kind of determining this is a high quality photo versus this is not a high quality photo. Right, right. Right, so what I did when I first saw it to build this was look into what exactly is a good dating app photo, what does it look like? And you know, the name is first pick, right? The idea was what would, what should your first picture on a dating app be? And I kind of triangulated and you know, correlated all this different information that was out there from dating coaches and things like that.

Your first picture should be and what it should look like. And then also gathered information on what was kind of a negative image to have or what wasn't gonna be a good one and came up with a framework for grading images. And then it was a matter of going out and gathering data to train my model, which I went to several different sources from and then using that framework that I came up with to label the data and train the model on it.

Hunter Kallay: Cool, so what did you find? Some people here might be like, oh well, you know, I wanna know what kind of pictures I should put on my profile and what kind of pictures I shouldn't put on my profile. What are just like some general things? Like I know you talked about like your eyes weren't shelling or like the phone in the picture. What were some like key indications that you found that were indicators of good photos and what were some that you found were indicators of non-successful or bad photos?

Loughlin Rodd: Right, so there's a couple of things that go into that. I think first and foremost, seeing the majority of your figure is really important. Seeing your face, seeing your eyes, so no sunglasses, no close ups of your face, not necessarily something that's just from the chest or waist up.

Show your full profile, show your face, a smile is generally very important so that you appear warm, welcoming, what have you, like those generally score very, very well according to the dating experts that I aggregated my information from. And in addition to that, like a high quality photo, right? So you don't want, you know, one of the funny things I found when I was researching this was one thing people fail to do that makes their photos like really unattractive is they don't like clean the lens of their camera, right? So they're taking photos themselves through a smudged up lens and then it doesn't look good, right? So just basic stuff like that where the lens is clean, it's in portrait mode, you're using a high quality camera and then you're in a relaxed pose with open body language and a smile and showing your full figure that generally is what's going to resonate the best for that first photo. And that's what the model is trained to help detect, right? Is, you know, how close have you gotten to that ideal?

Hunter Kallay: Very cool. And what about some poor indications? I mean, you kind of hinted at those, was there anything else that kind of stood out is don't do this, this is a big red flag in your photo?

Loughlin Rodd: Right, I mean, you maybe wouldn't be surprised at just the number of photos out there of largely men shirtless in the gym or taking a selfie of themselves, a close up of their face, photos with sunglasses, photos that are blurry, photos with multiple people in it. You know, it's interesting, you wanna have like one or so really high quality group photo on your profile where it's maybe you and two friends, right? And you all are looking presentable and what have you for like some social validation, but you don't want that as your first photo and you don't want that group photo where there's eight people in it and you have no idea which person is which, right? So those photos will tend to score pretty low, but yeah, those are some of like the low lights, especially selfies.

Hunter Kallay: Yeah, that's one thing that whenever you go on a dating app and you see the first photo and there's just a group, you're like, well, I really hope, or two people and you're like, I really hope it's that girl. And he's scrolling like, of course not.

Loughlin Rodd: You want people to know who you are, right? And I think that's like equal for men and women. You don't wanna have to play a guessing game of like, who is this and having to scroll to other photos to know whose profile you're looking at. It's a game of snap decisions, right? It's as quick as you can swipe left or swipe right. So if you're not giving someone a reason to swipe right on that first photo, you're putting yourself at a tremendous disadvantage. And if I can't tell which person I'm looking at, I'm more likely to just move on.

Ryan Lazuka: How do you deal with, you sort of hinted at it a little bit, but how do you deal with like cultural norms are different? So you might have someone from India and their photo is gonna be completely different than someone that was born in the United States. So is your tool objectively like doesn't care where you're like, what background you're from. It's just gonna say follow these rules no matter what or is it cultural based as well?

Loughlin Rodd: Yeah, so I didn't really identify any differences in terms of like this culture wants a dating photo that looks like this versus, what you would expect in the States. I did make sure to train the model on folks of different ethnicities, right? So that you didn't have a model that was biased towards white people or black people or what have you that could give an accurate score based on the quality of the photo regardless of someone's like ethnicity, right? But in terms of cultural differences, there wasn't anything like baked into the model around that. Got it, got it.

Ryan Lazuka: Yeah, maybe if you're like, if the app, because this is a US based app, maybe if the app was in a different country then it would be different, but it sounds like everyone sort of lives in the United States and it doesn't matter what your ethnicity is as much as how well your how well your pick's gonna do on the app. How does your, can you walk us through how it works for the end user? Like if they just went on your site, they want their profile pick improved.

Like how does it work on your site when they go on and log in and then the process of optimizing their picture and then putting it on say Tinder or something like that.

Loughlin Rodd: Right, so there's a free experience and a paid experience. So the free experience lets you upload a single photo and get a score for the photo from the model, one to 100. And then it'll also tell you your ranking among other users of the same gender and it will give you any feedback we have from the model. So if you have a killer photo, we probably won't have a lot of feedback for you.

But if there's issues like we can't see your eyes or things like that, there's no smile, we'll try to give you that feedback. And the end user can rate up to three photos out of the box for free. And then the paid experience allows you to upload multiple photos at the same time and go through that process in bulk and add your input, your like kind of answers to questions for populating your bio and prompt content for the different apps. And at the end of that process, you'll see the bio we've written for you and any prompt responses for the app you've chosen, either Tinder, Bumble or Hinge. And we know like what prompts each app has and will match your information up with the best prompts. And the photos that you provided will be ordered in the way we think you should order them on your profile. And you'll see feedback and the photos score and the ranking among other users for each photo.

Hunter Kallay: So I wanted to transition to the bios here. If you're following along, you can go on to firstpick.ai and you can see kind of sift through this as we're talking about it. Do you said like it'll generate, will generate different bios for Twitter, Hinge, Bumble based on the type of profile you put on those types of sites or is it kind of standardized across them? And does it generate the bio based on the pictures or is it like, does it prompt you on like certain questions about yourself and then puts together a bio for you? How does that work?

Loughlin Rodd: So it is going to ask you a set of seven questions. It'll ask you what your hobbies are, what you're looking for in a partner, what you like to talk about, what your job is, and things along those lines.

And it'll also give you some instructions on how to best answer the questions. So for example, if you go on Tinder, Bumble or Hinge, there will be thousands of people who say they like running, right? Or they like skiing. What our framework tells you to do is to be detail oriented. So I like running around the McCarron Park track at sunrise, for example, like that's differentiated, right? And it's content that when someone reads it, it can stand out as opposed to I like to run, right? So we're giving you instructions on how to give us the details that will allow us to craft a piece of engaging and unique content that will allow you to stand out on these apps where there's a sea of profiles.

And it's hard to kind of puncture through that and be matched with, right? So after you answer those questions, we then have like based on the app, you're wanting us to write content for our kind of AI will say, okay, you selected Hinge on Hinge. There are no bios, there are only prompts based on the information that the user gave us, determine what the best prompts to use will be and then craft responses to those prompts that are 150 characters or less or something like that, right? At a high level for Tinder, it'll say, write a bio that is X number of characters long and synthesizes all of the user's information and craft something that's engaging and detail oriented.

Don't use curse words, don't use any kind of sexually explicit content, don't seem like overly needy. It's basically giving a persona to the AI telling it how to take this information and craft something that's engaging and also not off-putting.

Hunter Kallay: Yeah, that's pretty cool. The thing you said about the specific details was really interesting, right? I mean, it's like, y'all like running around whatever park at sunrise, like that was a lot more catching to me than just saying that I liked running. That's just something I wouldn't think about. Are there any other tricks you find that you found in the bios that help people be successful?

Loughlin Rodd: Yeah, I think one of the most important things is like a positive outlook or positive tone, right? I think one of the things that like inspired me to add the bio functionality was, when I first started finding initial users for the app, I was going on Reddit and basically DMing people who had asked for profile advice on like our Tinder or our Bumble or our hinge. And one of the things they would come back with is like, I also need help with my bio and prompts. And when I dug into what people were putting on their profile, some of it was like negative, right? Or in a way, off-putting. So either either overtly like saying something that was negative, like I'm not looking for someone that is XYZ, right? Or something along those lines, or they would say something that was like off-putting in somewhere, remember someone had like a poll on their hinge profile, you can put like a poll, like what should we do for our first date?

And like it was, you know, go on a walk, get coffee, like try Blacktar heroin, like something like that. Or just like, you know, not like I get what you're trying to do, like you're trying to put something unexpected there that maybe gets someone to laugh, but that's not, you don't want to stand out in a bad way on your dating profile, right? So those were things I noticed that people were doing and I trained the AI to avoid, essentially. Yeah, so those are like some of the things that come to mind, you know, in answering that question.

Ryan Lazuka: What's your vision for this? Do you, cause I'd imagine like Tinder or Bumble or whatever is gonna start incorporating these things. Are you trying to get bought out by these companies or do you have a longer term vision around your product?

Loughlin Rodd: Yeah, so candidly, this was something I put together over a couple of like, I stacked a bunch of PTO days at the end of 2023 and built this like in my spare time. And yeah, I've been like impressed with the amount that it's resonated with folks and that it's been like picked up by, you know, folks like yourselves and other AI newsletters and things like that. I think in terms of my vision, I'd like to just continue like building it and learning what is gonna be impactful for people in terms of making a difference on dating apps that they can be successful, whether that's adding in messaging or some other functionality. I don't know that I have like a grand vision beyond helping people have better experiences on dating apps. So I have like a day job, right? So this is kind of a side project I've done in my spare time and I can't necessarily see it becoming a good thing, right? It's just a kind of hobby for me.

Ryan Lazuka: Cool, no, those are the projects that usually work out the best. So awesome job with creating it. One thing that was really interesting, we did a story on it a couple of months ago is there was a guy, I don't know if you heard of him, a guy from Russia that he did something similar but he started and you mentioned messaging. He used Tinder to, he used AI on Tinder to automatically message, you know, the people he was dating. So he was dating like, I think 100 people on Tinder at once and he just had the AI talk to these girls and eventually he ended up marrying one of them but he had them filter out. So he talked to like a hundred of them and then it would filter out like the best 10 girls that, you know, according to the AI that he should go on a date with and then it would automatically set up dates for him to go on and, you know, before he went on the dates he would just review what the conversation and what the messages back and forth between him and the AI and the girl were. So he knew, you know, what to expect on their date but it's just like, it's like real weird things that are starting to emerge in AI that I think that's one of them. But can you see the dating app sort of going down that road because the initial conversations people have are all kind of generic anyways, like you're just trying to get to know each other. I know there's a human aspect to it and it's fun to start talking with people but at the same time there's a lot of people out there that are lazy and sort of just want to go that route and not deal with, you know, introducing themselves to a hundred different girls before they find the right one.

Loughlin Rodd: Yeah, so I remember reading that story. I thought it was super like creative and interesting what he did. I suppose I struggle to see that approach working at scale. I mean, I just like imagine if everyone is doing that, like no one's actually having a real conversation. Like for what he did to work, there has to be a woman on the other side who thinks she's interacting with the real person, right? Now, if everyone were just kind of doing that, you'd just have AI bots talking to one another, setting updates. Maybe that's what it comes down to. I mean, maybe like, you know, there's people out there having like AI girlfriends right now, right?

So, you know, I guess I hesitate to say it had never happened, I could always be wrong. It just seems like something where if everyone did it, it wouldn't work but in isolation with a couple of people who are creative and tech savvy doing it, they can kind of level themselves up in their dating life. But if everyone did it, you'd have potentially a problem.

Ryan Lazuka: Yeah, but you mentioned earlier that you are thinking about implementing some kind of messaging via AI as well on your app or no?

Loughlin Rodd: It's definitely something that's crossed my mind. I know there's a bunch of companies out there that are doing it, which is one of the reasons why I hesitate to go after it, right? Like I think what's unique about what I built is no one's doing the photo part of it. There's companies out there that are doing AI photos, like taking a photo of you and using AI to make another photo of you, which I think is kind of disingenuous to put on your dating app. But no one's doing the photo angle, which is something that I think is cool and unique for me. And the bio angle some other people are doing, but I think if I can kind of own the, how to make your profile part, like that's a nice corner to be on. The messaging component, And the ones I've seen have been kind of gimmicky, right, where it's kind of trying to come up with something clever or what have you.

Or yeah, I don't know, when I use dating apps, the messages that kind of work the best for me are ones that are unique to that person's profile, right? And say something either funny or unique or that I noticed. And I think you could train an AI to do that. You know, and it's not something I've started working on yet. I think I've kind of got to decide whether I want to go deeper into, you know, really building something that nails making your profile or whether I want to kind of add this functionality that lets you, you know, start sending messages. So it's it's up in the air for me.

Ryan Lazuka: Okay, cool. Yeah, but sometimes you just got to nail what you're good at, you know, so no one else doing it. Then you got to leg up on your competition, which is no one really. So that's awesome. But what like one thing with the messaging, though, is even if I feel like people are getting more savvy on understanding if they're talking to AI or not, like you can go on Twitter right now or x.com. And on my post, at least like all the replies I get, I'd say probably half of them are AI created.

I can just tell. But I think people are getting more savvy on that. So if that does become a thing, you know, on Tinder or Bumble or any dating apps, people are going to start to like, know right away, they're going to feel if they're talking to AI or not. Like there's a lot of little subtleties that humans add that AI doesn't right now. And maybe that will get better over time. But if someone sees that they're talking to AI or figures it out, they're they're they're going to abruptly stop talking to you, I'd imagine. Right.

Loughlin Rodd: So I mean, the thing that I kind of like the idea of and the way I would probably want to approach messaging is, and I think this is a reason why maybe it wouldn't work with my approach is it's probably putting more work on the end user. But I like the idea of optimizing a message that you've already drafted. So you want to send this message to this woman on Tinder, you know, it's three sentences long, it's maybe a little too wordy, it's maybe the ideas you're trying to say aren't as clear as it could be. And our AI can take that in and spit out a, you know, a 90 character message that is worded the way it should be and resonates and gets the point a little quicker and maybe draws that person in better. I think what most of these kind of apps are doing now is they write the message for you as opposed to taking your original ideas and optimizing them.

So I think that would be an interesting way to approach it for me as opposed to just making the message for the person so that it doesn't lose some of that like personal nature.

Ryan Lazuka: For sure. I think that that's a gigantic market right there because, you know, I met my wife online and everybody that goes through online dating is thinking to them. A lot of people are thinking themselves, should I message now? Should I wait three days? Should I, should I, you know, ask this question or what timing? Like what time should I ask this question? We just went on the date. When should I ask her again? I mean, just thousands of questions run through your mind because they're excited. It's always exciting part of part of dating when you first meet someone. So like almost like a psychologist to say on top of this whole process, like an AI psychologist to say, Hey, like, I can imagine it being something like you start typing your message and you say, Hey, AI review this.

And the AI says, Hey, Ryan or Lachlan, pump the brakes. Don't message right now. You know, maybe message her tomorrow and do, you know, don't ask for the date right away or something like that. Like someone to keep you on the right track. That would be super helpful.

Loughlin Rodd: So, you know, it's interesting you bring that up. It's this isn't the first product I've fooled around with on the front. I actually worked on something a few years ago that would pull in your Tinder data. Basically like spoofed the Tinder API, which is like closed. So, you know, it couldn't actually be released to anyone in any kind of scalable fashion where you weren't running it on your local machine.

But it allowed us to pull in. If you signed in with your Tinder credentials, all of your Tinder data, your messages, your matches, super likes, boost matches, all that kind of stuff. And we built a dashboard that would basically say you get the most message responses at eight PM at night on Tuesday, or your best conversations are with women who are 27 years old, or you get the most super like matches from women who are 30 and went to an Ivy League school or something like that. And, you know, that that was like kind of trying to build like a dating coach in your pocket where it could help you understand when the best time of day went what day so you could kind of like optimize your efforts on the app and say, all right, I get the most hits on Wednesday night. That's when I should be on the app. That's when I should be sending messages. It would also tell you like the the length of message that was best for getting a response. So you'd know, hey, messages around 15 words seem to be like my sweet spot, like that kind of thing. Sadly, couldn't release it to anyone, but it was a cool project to work on and kind of helped like it. It was a bug that kind of stayed with me and like led to me building, you know, the first pick.

Ryan Lazuka: Awesome. And you couldn't release it because the API is not public. Is that the main reason?

Loughlin Rodd: Exactly. So the API is not public. It was like a spoofed version of the API, which like someone, some smart person figured out how the Tinder API requests work and put it on GitHub. And we were able to like plug into that and suck in all your data and do like analysis on it with Python and give you this nice dashboard with JavaScript and chart.js.

And yeah, if you tried to like sign in, like if I tried to host it on AWS, like because it was using Facebook's authentication, it would not work because my Facebook app that I was building to build this dashboard needed to use like the same app ID as the Tinder app ID from Facebook and Facebook didn't allow that. So it just, you know, it wouldn't work. So it was a little like technical there, but long story short, you couldn't do it.

Ryan Lazuka: And that was that before AI came out like chat, GBT and stuff like a few years ago. Yeah, this was like 2019 ish. Yeah, it's too bad. Like it sounds like it would be, it would be impossible to do something like that unless these dating sites like publicly released their API. Because what you're doing is off. You're doing is an on the site. Like, you know, like if you go to your site, you don't have to, you're not tying into any API or anything like that. You're just doing it completely on your own site to help people out. Exactly.

Loughlin Rodd: The only way you could do it is with an extract of the end users Tinder data, which we thought about playing around with, but just like you, when you ask for your data from Tinder, they take three days to get it to you. By that point, like your desire for this is probably gone. Right. Like it just would have taken kind of too long. And we wrote the code to look to work with this spoof version of the API, not a CSV that we were getting from Tinder. And so we just kind of, you know, tabled that and moved on.

Ryan Lazuka: Yeah, that's too bad. Unless someone came out with their, their very own Tinder, but that's a little, it's a tough task, tough ask.

Loughlin Rodd: I do, I do think there's like room in the market for something that leverages AI as a dating app. And, you know, we'll help match you with the right person. Uh, based on AI, one of the things that I think is a really interesting potential is quality of profile. So I think you can use like AI to score someone's dating app profile, right? Based on the quality of their photos, the quality of the content they put in their bio and prompts, and then at least match folks who have similar quality of profile together, which would potentially take away at least like all of this kind of wiping through nonsense profiles that you have on today's, uh, dating apps. Now that's probably not in the best interests of Tinder and Bumble and Inge.

Who, yeah, do kind of have vested interests in folks staying subscribed and staying single, right? So I don't know that they certainly would want to streamline that process, but it's an idea that I think is interesting and someone should take up.

Ryan Lazuka: Yeah, that's a really interesting idea. Well, we're interesting that you said that, I mean, a lot of people don't think about that, right? Like Tinder wants you to meet somebody, but they don't want you to meet somebody so quick that you're not going to use their app.

Loughlin Rodd: So that's, they want you to have hope, hope, but not necessarily results.

Ryan Lazuka: Yeah, it's almost, I don't know. It's fun talking to projects because it gets my mind going on like what future apps could be like. So like, what about the model where you meet, like you pay once you meet somebody or something like that? Or, I mean, I guess it's hard to prove, but I don't know. I'm just thinking out loud, I guess.

Loughlin Rodd: Yeah, something like what I built could be beneficial to a dating app just in terms of raising the floor of the quality of profile that is on the app. You know, one, there's an interview with a guy who was a data scientist at Bumble that came out recently and there was a summary of it on Reddit, unlike our Bumble and the, what the guy said was basically 20% of male profiles get 80% of the right swipes from women. So it's very concentrated in terms of that top, you know, 20% of profiles, the men that are getting interest on dating apps. So if you as Bumble had a way to raise the floor of the dating profiles of men and women, you could potentially make like a much more engaging app where the likes were more distributed across different people and it led to a better overall experience. Right now, they don't really help you build your profile. They have like some things that may be like, oh, we'll try and pick your best photo.

We'll try and give you some tips on writing your content, but there's nothing that is really like making it so that you have like a killer profile on their app. Right. And that could make a big difference in some of the, you know, kind of fatigue, I think, is out there with these apps right now.

Ryan Lazuka: For sure. I mean, like even if somewhat like if Tinder's or Bumble, their motivation isn't, is to give you hope, but maybe not necessarily find you the right match right away. If another app did come out where it raised that floor up a little bit, the quality of your dates are going to go up. You're going to be more happy to use the app and it's still going to take you. I don't know what the average like time is that someone or how many dates someone goes on on these dating sites. But, you know, the quality of those dates goes up. They're going to be more happy to use the app in the future, I'd imagine, even if they find someone they like. Yeah.

Loughlin Rodd: I mean, I would love to see an app come out that did a little bit of turning the revenue model on its head where I think you could implement some kind of framework that's more results based, right, as opposed to a subscription model, maybe, you know, based on like a phone number being exchanged or something like that, that deducts like a credit and in order to, you know, get more matches, you've got a, or what have you pay for some kind of credit and you only deduct a credit when a phone number is exchanged, something that like turns the kind of subscription model on its head and makes it more results oriented. And then I think you could, there's actually like room for something that's donation based, like if I went on an app that was free and I found my girlfriend who became my wife on that app, I would be very likely to open up my wallet and give that app like a significant chunk of change. Like that's not a worthless thing to me, right? Now, like how do you run a business based on, you know, irregular donations? Like that's another thing. But, you know, there's, there's ways to like change this model up so that it's not like lock someone into a six month subscription and then trickle matches out to them and maybe they result in something, maybe they don't and they have to keep subscribing.

Ryan Lazuka: For sure. And then it also helps. I like say if, you know, you may or you end up marrying someone on one of these sites, like that is going to be free marketing for the rest of that person's life. Right? Like, how'd you guys meet? Oh, we met on Tinder.

We met on maybe, maybe some people won't admit that. But if you have a more reputable site that people are like, Oh yeah, that's a really good one. Like eHarmony, you know, had a better reputation back in the day than, you know, okay, Cupid or some other sites out there. I don't even know that site's still around, but it, when people think of Tinder, they think of like going on dates just to hook up with people or, you know, it's got a bad connotation to it. So if you could build a site that has more of a quality reputation around it, I mean, that could really help it as well. I'd imagine.

Loughlin Rodd: Absolutely. Um, you know, it brings up an idea for me. There's a book by the woman who was her name's Logan Uri. Uh, it's called how to not die alone. And she was like the chief data scientist at hinge. And she has like an anecdote about one of the, like her women friends in her life, like kind of put a bounty out there.

If you like match me up with my future husband, I'll give you $5,000. Right. Uh, and then it made people far more likely to like keep their eye out for single men who could be a potential match for her. And it actually, it worked. And, you know, she found her partner that way. Uh, I wonder if there could be like a bounty based system on some of these apps where it's like, all right, if I find my girlfriend and delete this account, that's worth $500 to me. Right. And when that happens, you know, we charge your card 500 bucks. No, I'm just throwing things out there.

Ryan Lazuka: That's not a really cool idea, but, you know, like I'd happily put a bounty out there and just say, Hey, if you find my next girlfriend, that's, you know, a thousand bucks, right? Uh, whatever the, whatever the number is.

Loughlin Rodd: Uh, you know, and then maybe that incentivizes the app based on how high the bounty is to send you quality matches. You know, there's ways to like turn this on its head where it's not this like kind of swiping nonsense. I just don't think there's a lot of incentive out there for the existing the existing participants in this market to do anything super innovative. It's super interesting.

Hunter Kallay: One of the things that, uh, has been coming up a lot in discussions with different developers and people in the AI space is this idea of AI relationships. So I wanted to talk about that for a second with you since you're in the space.

Um, what do you see? What do you make of AI relationships? Like in, um, in here, I'm talking about people who are in relationships with an AI.

Uh, boyfriend, girlfriend of all different sorts. I mean, we've seen them in virtual reality. We've seen them just in versions of chatbots or little avatars on different apps. Um, such as replica or something. What do you make of all this? Where do you think it's headed?

Loughlin Rodd: Yeah, I can't say I'm super like in the weeds on this. Most of what I've heard is kind of anecdotal from other, you know, tech forward folks who've like gotten, uh, exposure to it and they're like, yo, this is kind of wild, right? What's happening? Uh, you know, I guess, I guess it's like a little worrying to me.

Right. I think there's, you know, if you, if you go on Reddit, you'll see like, uh, you know, posts about like male loneliness, right? And I think like it's not just men.

I think it's women too, right? Where there's, you know, a lot of like kind of loneliness out there. And I would hope that we as a society aren't like trying to cure that with an AI chatbot, right? Like that seems like a, like, I don't know, that's not a life I want to live.

Right. I would hope other people would, would rather live a life where they can, they can overcome that and find like hobbies and things like that that get them exposure to more people and potential partners, uh, where they, where they don't have to be engaging with an AI bot. And then on the other hand, I'm sure it's like a very lucrative business to be in, right? Uh, where, you know, uh, from what I've heard, there can be like very like sexually explicit, like AI girlfriends, things like that that are almost like kind of a, uh, like a pornographic experience, right? And, you know, when new technology comes out, you know, a lot of times like the, the kind of, you know, porn industry, things like that are the first to adopt it, right? Uh, and kind of, you know, turn it into a new revenue stream and channel. So I think, uh, you know, that's kind of one of the areas that I think it'll go. Uh, and I don't know that that's like a great thing for society, right? But, you know, that's kind of the way things have gone in the tech industry in the past and, uh, you know, we'll have to see how it plays out.

Hunter Kallay: Yeah. It is, it is an interesting topic. I like to think about it's, it's hard to imagine, uh, you know, in Ryan and I have talked about this a couple of times before, it's like, it's hard to imagine at some extent, you know, if you're lonely, you can't find a girlfriend, why you wouldn't turn to something like this. I mean, like a, a girlfriend who's like always going to be there for you when nobody else was somebody who's going to like know everything about you, who's always going to be there to support you.

It's like, you can definitely see the pull of it. Like why some people would turn to an AI boyfriend or a girlfriend or just an AI companion and AI friend. But yeah, I mean, as you say, there's a lot of negatives to it as well. Um, but yeah, as far as what's going to happen and what should happen or sometimes and oftentimes two very different things, it's just going to be very interesting to see how this plays out.

Ryan Lazuka: It might be one of those things where like, you know, you, you say you break up with your girlfriend or boyfriend and, you know, you get depressed or whatever. And you know, a lot of people will eat or have some kind of vice, like go out and drink. Maybe that's going to be the next device, right? Like, oh, you have, you, you're talking to the AI chat bot to help you out through this. Like maybe it's one of those things. Cause I don't think anybody has a good connotation of like having it.

It's, there's not a good connotation of having an AI girlfriend. No one's like, Oh, good job, sunny. Like great. Glad you're dating someone again. Getting back out there. You know, most people are going to be like, what are you doing? That's crazy. So I don't know. I mean, it's hard to predict the future, but I can see it going down that route.

Hunter Kallay: Yeah, absolutely. I mean, and we've seen this like over and over again with like AI therapists too. I mean, maybe it's turns into something similar. There, there's some crossover between those two industries where it's like these AI girlfriends or boyfriends kind of act as AI relationship therapists in a way. Like how I'm going to help you get over your ex by letting, you know, being there for you, you know, so you're not alone or something like that.

Loughlin Rodd: Yeah. Or even like, how do I manage my relationship now? Right. Like that's something that's kind of interesting. It's like, okay, you know, my, my partner and I got into a fight. This is what happened. Help me like see it from her perspective. Right. You know, I can, I can see that being useful, like kind of like an AI therapist. Right. Yeah.

Hunter Kallay: I've seen some of those tools actually out there. We do a lot of research in the, in the emerging tools and things. I've seen some, some of those. Oftentimes they're just too scared to give you real advice.

Right. But if we can get to the point where they actually do give you real advice, maybe Mistral can come out with something like this. And then when we get to the point where they actually do give you real advice, like you should break up with her man.

Like she's toxic, you know, like something like that. I think that might actually be really helpful and pretty cool. Obviously they'd have to like put like some disclaimers on there. Like don't blame us for ruining your relationship.

Ryan Lazuka: You're going to fight with your girlfriend and you do something to her because chat, she told you to, you know, make it mental with her, but you did it. Like told you something really stupid to do. And she's like, she breaks up with you and she's like, why'd you do that? You're like, chat, she told me to do it. You know, like, I could see that kind of stuff happening.

Loughlin Rodd: Yeah. These, these AIs are very, uh, they're very cautious about like the information they'll give you sometimes, like just kind of a little off topic. But one of the things I do in my spare times, I do a little stand up. Right. And awesome. Yeah. And one of, sometimes I'll have like a joke idea and I'll not quite know exactly how I should word it. And I've plugged it into chat, GPT before and been like, Hey, I think this is like a funny concept.

What do you have? Like that comes back here. I want to joke based on this and they'll be like, this is a very like sensitive topic. You should be very careful about this and what have you. Right. So you're like, Oh, all right.

This is not actually useful to me. And I can see like they're being, you know, they don't want to say something that's like offensive, right? So these models probably wouldn't want to say like, Hey, go break up with your toxic girlfriend, right? They're probably going to be a little more cautious in what they tell you.

Ryan Lazuka: For sure. You got to have you played with like, have you heard of a llama before? Lachlan? No, I haven't. It's like this tool where it lets you download LLMs and AI models on your local local machine. It's really cool.

One of my friends told me about it. So like there's, and then you can download mist, mixed role, mist role, you know, llama, any, any model you want, but dolphin, dolphin mist, mixed role on Olamma. If you download that one, I mean, these models are big.

They're like 25 gigs. But once you have a downloader, you can start talking to it and it's hilarious. Like I told it to like roast Elon Musk or something like that or a roast Taylor Swift and it just goes into this tirade of like swear words and like just ripping them to pieces and it's hilarious. Like it is so refreshing. So I bring that up because if you wanted to write a joke, that would be a perfect place to go or have a review your jokes because it does not hold back at all. It's like eye opening the first time you use it. All right. That's something I got to check out. Olamma. Yeah. Olamma. Specifically dolphin, the dolphin mixed role model. It's really good. Yeah.

Loughlin Rodd: That, that, that sounds like very useful. Like, uh, for that kind of use case and I don't know, it'd be fun if we could like, you could probably train a model, right? On like different like jokes and voices from comedians, right? I mean, then you get into like copyright stuff, right? And then it becomes like, Oh, did I, did I like actually write this joke? Did I steal it from a combination of Jeff Ross and Bill Burr like from an AI? Like, you know, I don't know. Yeah.

Ryan Lazuka: I mean, well, that's the thing with comedians, right? It's like, if you steal someone's joke, the people get really pissed off. So, you know, you could just be like, Oh, it was chat. Gbt. It wasn't me. You'd still look like an idiot saying that, but at least you could put the blame somewhere else, I guess.

Hunter Kallay: Right. Ryan, you say it's refreshing because it really is like, it gets so annoying when you try to, I do a lot of studying and ethics and AI ethics, among other things. And you can't use, you literally cannot use these models to help you with anything because every time you try to talk to it about some ethical concern, it's all of a sudden, like, like you said, when it's like, you get flagged, this violates our content policy, or you need to be very careful about this sensitive stuff, like make sure that it's like, no, I just want you to tell me, like, is this, is this structure, like, is my take on this structured well or something like that? And it just, it just won't give you any feedback.

It's very frustrating. So when you see something like that, or you see AI like go rogue or something like that, it is kind of refreshing in a way. Like, yeah, it's like, it makes you like step back and be like, whoa, like, this is kind of scary.

But at the same time, it's like, yeah, it's pretty cool. Like I'd love AI to just like throw out some opinions every now and then. I think it would be pretty cool. One of the, one of the thoughts that I've had is like, and I think Mistral wants to do something like this.

But of course, it's going to be, I think it's going to be censored down. But one thing that would be really cool is just like put as much, I mean, it's hard to find like unbiased data. So just find as much data as you possibly can and put it into a model. And then just like, see what, see what it says, like, see what it pops out and says, see what its opinions are on certain things. And I just think that'd be kind of interesting how it reasons and how it thinks just given the data without any censorship. I think that'd be kind of interesting to play around with different data sets doing that sort of thing. Yeah, it certainly is refreshing.

Ryan Lazuka: Yeah, for sure. It'd be, it'd be really awesome to be a, you know, an employee at OpenAI. And then you'd be able to see, I'm sure they've got uncensored versions to play with, you know, like in, within the company. It'd be fascinating to see that play out, you know, type in something and see what it actually really would say without any filters. Yeah.

Hunter Kallay: I want to be like, okay, well, I'm going to, I'm going to use this dataset and now I'm going to throw in like this dataset from the New York Times and see how it changes its answer to this. I'm going to throw in this dataset from here and see how it changes its answer to this controversial issue. I think that'd be kind of interesting. I think, I think also there's potential there to like assess the lean of various datasets on issues.

Cause like a lot of people want to say, well, you know, this news, news source has this political lean or this news source has this ethical lean or something like that. Like we can actually test this, just set apart that set and train a model on it and then ask it certain questions. I mean, I think that'd be a cool thing to experiment with. For sure. For sure. Yeah. Do you have anything else, Lachlan, that you'd like to promote or say about your project or anything like that? I mean, I know we cover a lot of stuff.

Loughlin Rodd: Yeah. I think we had a great conversation. I appreciate you all taking the time. And, you know, definitely like raise some ideas for me about, you know, things I could build, right?

Yeah. For anyone watching would love to have you try out first pick and let me know your thoughts that's at first pick.ai. And, you know, if you have any feedback or thoughts, I'm just Lachlan, L-O-U-G-H-L-I-N at firstpick.ai. I would love to hear from you. Always want to get product feedback, ideas from folks who are interested in the space. So feel free to reach out.

Ryan Lazuka: For sure. It's an awesome tool. And who knows, like give it a shot. Like it's free to try out. And who knows, you know, you might find your future wife, like, you know, I did a long time ago. So but thanks. Thanks. Thanks for coming on, Lachlan. I really appreciate it.

Hunter Kallay: If you're watching, be sure to subscribe below so you can get all of our videos and see our cool conversations with developers like Lachlan. And then also subscribe to our weekday newsletter fry-ai .com.