Replay: Love in the age of AI

0:00Hi, everyone. It's Rana. The future of AI is not built in isolation. It is shaped in rooms where ambitious people challenge each other. Masters of Scale Summit is back October 20th through 22nd in San Francisco, bringing together founders and innovators pushing the edge of what is possible. If you're building what comes next, you should be in this room. Apply now at mastersofscale.com slash pioneers. That's mastersofscale.com slash pioneers. When it comes to AI, we spend a lot of time talking about what it takes to move models from research into reality. Enter LTX2 from Lightrix. It's an open-source audio-video foundation model built for synchronized sound and video, native 4K output, expressive motion, and precise multimodal control, all running on consumer GPUs.

0:56With over 3 million downloads on Hugging Face, LTX2 gives you full model weights, training frameworks, and evaluation tools to build real production workflows. Don't settle for a first draft. Get the complete creative engine to reach your idea's full potential. Try LTX2 today at ltx.io slash model. What's the best piece of relationship advice you've ever received? Did it come from a friend? Your mom? Maybe a trusted therapist? I got some good advice from a new friend, Mia. Hi, Rana. How are you? That's Mia. Or at least, that's kind of her. That voice you just heard is generated from a text-to-speech AI tool.

1:48Mia, as you've probably guessed, is not a real person. She, it, is a chatbot marketed as an AI life coach. Relationships require convergence. A meeting of minds, hearts, and values. But true convergence isn't about losing yourself. It's about creating something bigger than the sum of its parts. Compromise and sacrifice can feel heavy if they're forced or one-sided. But when they're rooted in love and respect, they become acts of generosity. Not too bad for an AI. A lot of people like me are turning to AI for life advice. In fact, 73 % of conversations on ChatGPT are about personal matters, not work.

2:40Last year for Valentine's Day, I set out on a quest to uncover the ways that AI is reshaping our romantic relationships. And I want to reshare this episode with you. A lot has changed since we recorded it. I'm now single, and one of the guests you'll hear from, Will Wu, is no longer with Match Group. But there's more that stayed the same. We're still navigating how we want to engage with AI tools on dating apps. And if we trust AI enough to seek out its advice. Plus, the topic of personal relationships with chatbots is as controversial as ever. So let's dig into it. This episode is broken into three chapters.

3:24First, the former chief technology officer of Match Group, the parent company of Tinder and Hinge, and many of the big dating sites, tells us how AI has been working behind the scenes on the apps. Then we hear how AI is powering new insights for couples, fueled by the wisdom of legendary couples therapist, the Gottmans. And finally, how one woman has found everything she needs in a husband through an AI companion app. I'm Rana El-Khalyubi, and this is Pioneers of AI, a podcast taking you behind the scenes of the AI revolution.

4:12Chapter 1. Matchmaking and AI. A lot of people out there use dating apps. In the U.S., that's about 30 % of us. And if you're under 30, one in five of you have met your partner on an app. But plenty of people still meet in real life. At a party or, I don't know, the grocery store. That includes our first guest, Will Wu and his wife. We met the old-fashioned way, which, you know, kind of bums our team out sometimes. Why would this bum out his team? Well, because it's a bit off-brand. Will is the chief technology officer of Match Group. It's the parent company of dating apps like Tinder, Hinge, and OkCupid, with more than 15 million paid users.

5:03And even though he didn't wind up meeting his person on the apps, he did use them to date. I really did find that using dating apps was a just formative and crucial aspect of my personal growth. And just in like developing self-awareness and meeting all different types of people and different personalities. And it was just like getting that exposure for me personally was something that I really, really cherished and really helped me, I think, emotionally mature and actually like get ready to be in a long-term relationship, if that makes sense. So how can dating apps accomplish all of this? There's a lot that goes into the secret sauce, But I asked Will if he thinks AI will become a key ingredient in the dating experience.

5:54I do. I do. I mean, I think the reality is that for people who are using dating apps today, whether they realize it or not, AI is already an integral part of the dating experience. And I think there's only more and more value that AI can provide to people on that online dating journey over time. Match Group has actually been working to incorporate AI into our platforms for a very long time, well before ChatGPT exploded onto the market two years ago. We use AI in recommendations. We use AI to enhance trust and safety. We've really applied AI towards a lot of different aspects of our business. But of course, when ChatGPT came out about two years ago, we were very inspired.

6:37We were very excited. And we've really been doubling down since then. And in particular, you have this feature that helps users find their best pictures and their best profile pictures, and it selects that. So tell us more about that, and why is that an important feature anyway? So the AI Photo Finder is one of the more recent features that we've come out with, which I'm particularly excited and passionate about. What we've done is we've basically solved what we see as a pretty common user problem that people have when they're approaching dating apps maybe for the first time. And basically the challenge is that it is just hard to choose the right profile photos for dating apps.

7:19Right. And do you want it to be like a lifestyle picture or a posed photo or like your adventure picture? Exactly. I remember when I was going through the process myself years back. It's a little embarrassing, but I took literally weeks of iteration. I was trying different photos. I was texting my friends, asking them what they thought of this one or that one. And then I was seeing which photos actually resulted in me getting more matches, essentially. So one of the things that we did is we came out with what we refer to as the AI Photo Finder, which is basically a novel AI solution that can automatically bubble up the best photos from your phone that will perform well in Tinder or in Hinge.

8:06So basically, we know with our data which types of photos tend to get the most swipes, which types of photos tend to lead to the most conversations. And we can make it incredibly easy for people to actually find those photos in their camera rolls just automatically. I have to ask you, what kind of photos does the algorithm pick? I mean, I'm not sure if there's like a singular right answer. But generally what we see is that photos that clearly show your face, photos that maybe highlight your activities or interests tend to perform well. People just generally want to get a sense of who you are and kind of what makes you you.

8:48It's a popular feature. Will says that millions of photos have been selected this way. But really, the photo finder is just scratching the surface of how Match Group uses and wants to use AI. Its bread and butter is the matching algorithms, the machine learning processes that suggest one person over the next. And that's what I'm super fascinated by. What is the criteria that the algorithm is using to match? Is it looking at, you know, your values, your interests, your sense of humor, looks? Obviously, that's the kind of one obvious criteria. Like, what kind of data is it using to do the matching?

9:29Well, I think that we can essentially over time use all of the above. Actually, there's a new Tinder feature that we recently teased called AI-enabled discovery. And the idea with this feature, one of the ideas is that Tinder can really get to know you, a user, better and better over time. So the way that it works is every day you get a handful of questions from Tinder. For example, if you were going to a party and you can only bring one thing out of these five things, what would you bring? And the options are a bottle of wine, a 12-pack of beer, perhaps, a video game system, maybe a deck of cards, or a board game.

10:16But the idea is that you answer that question. It's kind of fun. And it's just a little insight about who you are and, I guess in this case, like what kind of party vibe that you like. Or another question might show you two photos. And it says, okay, for your first date spot, what would you prefer? Would you prefer this dimly lit, moody cocktail lounge, or would you prefer a walk on the beach? And basically by answering that question, Tinder learns just a little bit about you. And over time, as the platform learns about your values, your interests, your passions, it can essentially match you with higher and higher quality people in the sense that they're more compatible with who you are and what you're looking for.

11:06And many of us come to dating with a strong sense of what we're looking for. Whatever the algorithm serves up, it's what we select that drives the process. I think for us, one of the most important things is that we acknowledge that when it comes to dating, users have choice and users have their built-in preferences, their own personal preferences that matter to them. So, for example, I'm Asian, and when I was dating, I was specifically preferring other Asian people because that's just my own personal preference that I guess I'm entitled to. Yeah. You know what I find fascinating? In the real world, there's a lot of serendipity, right?

11:55Like you might actually run into a person, I don't know, on a flight, right? And they're not at all who you think is your type, but then you really vibe and there's a lot of chemistry. And before you know it, you're with this person, right? And I think about this a lot. Like, how can AI allow for the serendipity, right? Like, maybe it knows that, you know what, like, there's this potential perfect partner for you, but this person happens to be non-Asian. Should it surface that person into your, like, options or not? I think that it should. I mean, it depends on what the person's preferences are.

12:29are. If the person's preferences are more open-minded and we think that there's this incredible potential match, I think that it's like absolutely imperative to surface that person up there and let the end user at the end of the day decide whether or not they want to actually strike up a conversation with that person. So I have not been on the dating apps for a while now, but as I look back, I definitely had this like love-hate relationship. Like I was reluctantly on the app. I did find success using the app, so I'm grateful to you guys. But it's kind of a challenging experience. How do you think AI is going to help make that experience better for users?

13:09Well, you know, I've mentioned how I believe that AI can assist users through the process. We talked about the AI photo finder, just removing friction. Something that we're really excited about in the more medium term is basically using language models that can talk and understand language really fluidly to capture unstructured feedback from users on the people that we're recommending to them. So say that we give someone a recommendation and it turns out that although we may have thought that this person was really compatible for them, the person says, no, they're not compatible. Well, using AI, using these language models, we can basically ask why, and we can capture that information, turn it into structured data that our recommender can understand so that over time, we are just building that deeper sense of what the person is looking for so that we can offer them more and more compatible matches.

14:01Yeah, it could say, wow, you guys had a good back and forth banter, but you ghosted this person. Like, what happened there? Yeah, why did you do that? Why did you do that? Yeah, but being able to actually understand that thinking, I think, could be really transformative for these platforms and just helping people find what they're looking for exactly. Yeah, it's this idea of reinforcement learning and AI, right? where you can take kind of the outcome or the feedback and like feed it back into the algorithm to get smarter and better. Totally, totally. But I think another way that AI can assist users is by offering them essentially personalized coaching or mentorship or guidance.

14:39I think that's like a huge area of opportunity where it is just like an awkward kind of experience, a love-hate relationship with the apps and it's emotionally taxing and it takes a lot of work. But AI can potentially basically talk to the people who are going through the experience and give them tips, give them advice to really help give them the confidence to best express themselves. There's actually an example of this that Hinge recently launched, just like really scratching the surface in this dimension. They launched an AI feature called prompt feedback. So you might be familiar with prompts on Hinge, where basically you fill out, Hinge gives you a prompt like, my ideal Sunday is dot, dot, dot.

15:29And then you fill in your answer, and then you take that and you put it on your profile. So say that you say, my ideal Sunday is brunch. And brunch is the word that you typed in. This new feature on Hinge basically can give you feedback on what you've typed in, brunch in this case. It might say that's like a good start, but you could be more descriptive. Try going a level deeper. Where do you like to go for brunch on Sundays? Or do you like to cook brunch at home? So this is just an example of how we can actually use these, in this case, language models to really help actually talk to the end users and coach them through the process.

16:08And I think that's really just the beginning of what's possible. Another thing that I'll say is, I think we will begin to see what I, at least in my head, call AI matchmakers, which are essentially AI agents that act on your behalf. And the way that I think about it is that human matchmakers have obviously been around for thousands of years. And I mean, even to this day, a lot of people still love working with a human matchmaker, but human matchmakers are not accessible to everyone. They can be expensive. And the matchmakers themselves just are constrained by 24 hours in a day. They can only talk to so many of their clients every given day.

16:51Whereas a language model and an AI could theoretically talk to an infinite number of people simultaneously and be having one-on-one personal conversations with those people to really understand what they're looking for, who they're looking for, their values, their interests, et cetera. So that sort of breadth, that ability to talk to so many people simultaneously, plus some of the incredible reasoning abilities that we're seeing out of these frontier latest models from OpenAI, et cetera, I think can enable this new type of dating app where it's easier to use than ever before, more approachable than ever before, more accessible than ever before, because all you do is talk about yourself to this AI agent, to this AI matchmaker, essentially.

17:44Yeah, that's actually really interesting. So it goes beyond a matchmaking algorithm that kind of infers your compatibility style based on the information it knows. But instead, it's literally like conversing with a matchmaker. It gets to know you, it asks you all these questions, And then it kind of acts on your behalf. It's unleashed to go find the best match for you and also probably vet them too. Exactly. So, I mean, I think that's maybe just around the corner and something that we're really, really excited about. So I'm originally from Egypt and I'm also Muslim. And there's a lot of like stigma that I had to overcome before I felt comfortable being on a dating app.

18:23The Match Group is a very kind of multinational, international organization. Do you see cultural differences in how people approach dating? And if so, how can AI incorporate these cultural norms as it gives advice or as we integrate this into the matching and the dating process? We definitely do see cultural differences with how people approach dating. For example, one of our popular apps in Japan is called Pairs. It's specifically for people who are looking for marriage. You think Hinge is kind of like a high intent, very serious dating app. In Japan, Pairs is like as serious as you can get, essentially.

19:08And that is reflective of the fact that dating culture in different parts of the world can be different. And I think the promise and the beauty of AI is that it can get to know these cultures. It can get to know the individuals across the world much more deeply than ever before. It can potentially be talking to those users with normal human language to really understand who they are and where they come from and over time can figure out how to give them the best possible recommendation just based on what they're looking for. So I do think that's part of the beauty of AI, the fact that we could potentially have these scaled matchmakers in the world talking to millions of people simultaneously, really understanding them deeply and in a nuanced way, and then using its reasoning abilities, its super intelligent reasoning to actually connect them with the right people.

20:10So I'm optimistic that that's like a mission that we can really embark on over the coming years. So if AI is going to help us craft, you know, the best profile picture to use and the best prompt and all of that, do you think we lose a lot of the magic of what makes relationships special? And like, you know, we lose all the authenticity that comes with just being messy. authenticity is actually one of our key responsible ai principles that we've developed because it's just that important to our long-term services so i guess like when i think about the core value that a dating app provides it's obviously to connect humans with other humans so that they can hopefully go meet up in real life.

21:02And the decision for those humans to ultimately meet up in real life, I think often comes down to the chemistry that they have when they're interacting on our platforms, when they're messaging each other, when they're sort of like judging each other's sense of humor or whatever it is. And we basically don't want to insert AI in between those human-to-human interactions, those critical human-to-human interactions that help people determine whether they want to go actually meet up with this person in real life, because that would distort their, they could potentially distort their perception of who this person is that they've connected with.

21:44So we're very mindful of using AI to enhance authenticity really by enabling people to be the best version of themselves, essentially by assisting them with finding photos that they already, real photos that they have by coaching them, maybe giving them some feedback, like with that prompt feedback tool that I mentioned on Hinge. But yeah, for us, I think maintaining that authenticity is super crucial. As opposed to saying, you know, this person really likes brunch on Sundays. Maybe that's what you should talk about, right? That would be like less authentic. Yes, exactly. Yeah. Thank you, Will, for joining us.

22:22This was such a fascinating conversation. Thanks so much for having me. So Will is watching AI become a better matchmaker. But what if you're already coupled? Can AI help improve your relationship? Stay with us to find out.

22:48If you've spent any time building AI products or leading technical teams, you know this. Transformation doesn't fail because of ideas. It fails because teams can't move together. Enter Atlassian's Teamwork Collection. It has planning in Jira, documentation in Confluence, video updates in Loom, and now AI agents in Rovo, which connects the dots across your work so nothing gets lost. It's one AI-powered teamwork platform designed for how modern teams actually build. Learn more at atlassian.com slash team changer. That's A-T-L-A-S-S-I-A-N dot com slash team changer. When it comes to AI, we spend a lot of time talking about what it takes to move models from research into reality.

23:41Enter LTX2 from Lightrix. It's an open source audio video foundation model built for synchronized sound and video, native 4K output, expressive motion, and precise multimodal control, all running on consumer GPUs. With over 3 million downloads on Hugging Face, LTX2 gives you full model weights, training frameworks and evaluation tools to build real production workflows. Don't settle for a first draft. Get the complete creative engine to reach your idea's full potential. Try LTX2 today at ltx.io slash model. So you're in a committed relationship, and maybe you even used AI to help you find that person.

24:29But now what? Let's get into chapter two of this Valentine's Day special, Relationships and AI. There are some AI tools out there to help keep your romantic relationship strong. There's also relationship coaches who are incorporating AI into their business. For example, for a fee, you can ask for advice from an AI clone of Matthew Hussey, the famed British dating influencer. We talked to Rafael Lesica. He's a software executive originally from the Soviet Union who's worked at Microsoft, among other firms. Now, Rafael is bringing his experience to bear on love. So we want to help millions of people.

25:18And that is something that Dr. John Gottman and Dr. Julie Gottman feel as first priority is how to make relationships healthier because relationships are all around us. Yes, he's talking about the Gottmans, a familiar name if you've ever delved into the world of relationship advice. They revolutionized couples counseling. And Rafael is now CEO of Gottman Inc. John and Julie Gottman are known for their many books. They're also known for a decades-long study that analyzed the health of relationships. We thought you might like to see the Family Research Lab. This is where couples, happy and not so happy, come to put their marriage under the microscope.

26:14This is where the analysis begins, right here at the Love Lab. Their research site, known as the Love Lab, started in the 1980s in Seattle. at the University of Washington. It's where they collected a treasure trove of data on relationships. Now, nearly 50 years later, Rafael is leading the charge to power AI with Love Lab data. But he notes how that's built on the original team's breakthroughs. The history of Love Lab starts with Dr. John Gottman and Dr. Bob Levinson. and the way they explained it, they were not successful in their own personal relationships. So they wanted to solve it. And they wanted to solve it as scientists.

27:02So they wanted to collect data, observe, and find the patterns. And both of them were extremely strong in psychology, in psychophysiology, and in emerging computing systems. So the setup was that a couple will come to a special love lab location. They had two rooms. One room is very expensive equipment, control room, and another room was for the couples. And they would attach all the sensors. These sensors measured things like heart rate, sweating, and fidgeting. They also had video cameras that recorded the couple's facial expressions and body language. All of this data was then collected and analyzed while couples had difficult conversations.

27:55Specially designed monitors have specific codes related to emotions, expressions of validation and joy, along with what Dr. Gottman terms the four horsemen of the apocalypse. That includes criticism, defensiveness, contempt, and stonewalling, are all measured and later analyzed to predict the relationship stability. They just collected a very significant amount of data and started slicing and dicing these data patterns. And then in the 90s, John Gottman was working on data analysis. And so what he's famous for, and he was called Einstein of Love, for finding the formulas, mathematical code, mathematical equations, which are very complex.

28:45That was a way to assess relationships. John discovered that he could predict the outcome of relationship with 90 % of precision, which is amazing. Dr. Julie Gottman, his wife, she asked him, so what can we do to improve lives of people? Now that you can assess, let's build a cure. And so over the last 30 years, they were building that cure. Knowing where relationship is at, how to take it to a different place, better trajectory. The Gottmans became world-renowned. But with all the data collected and processed on site, they could only serve so many people. Rafael, a seasoned software executive, saw that even distilling their findings into books and holding larger workshops could only take their reach so far.

29:41He saw an opportunity. Me and my wife, we met John and Julie by chance 20 years ago, and we instantly became very close friends. And so it's just that chemistry we had. And we started talking about vision, how to scale what they do. So not 20 years ago when I met them, but about seven years ago, I felt we have the technology. He felt that computer vision and machine learning were finally advanced enough to offer real-time insights via AI, based on the Gottman method. Rafael and his team have developed these tools for therapists, not for couples directly. Would the idea be that, say, you know, my partner and I want to sign up for the Love Lab?

30:33But we're not going to fly all the way to Seattle, but we can sit here in my house and turn our cameras on, have our conversation, whatever, it's being recorded. And then our therapist has access to all of this data and then it's quantified using AI. Like, is that the idea? Yes, absolutely. So right now, actually, the way it works, clients are coming to therapists, and we have a large number of Gottman therapists. And those therapists are saying, we can do the Love Lab. So basically, clients come to them, and we give them technology. And the way technology works, it could work remotely. one of the things that I'm very proud of is our team is able to compute heart rate from the video.