This Week in Startups - How AI companions can help solve loneliness with Replika’s Eugenia Kuyda | E1758

Episode Date: June 7, 2023

This Week in Startups is presented by: Crowdbotics. Great ideas can change the world, and Crowdbotics is the fastest way to turn those ideas into code. Get a free scoping session for your next big app... idea at crowdbotics.com/twist The Microsoft for Startups Founders Hub helps all founders build a better startup, at a lower cost, from day one. Startups get up to $150K in Azure credits, access to free OpenAI credits, free dev tools like GitHub, technical advisory, access to mentors and experts, and so much more. There is no funding requirement, and it only takes minutes to join. Sign up today at aka.ms/thisweekinstartups OpenPhone. Create business phone numbers for you and your team that work through an app on your smartphone or desktop. TWiST listeners can get an extra 20% off any plan for your first 6 months at openphone.com/twist * Today’s show: Replika Co-Founder and CEO Eugenia Kuyda breaks down her startup’s origin story and purpose (1:43), the relationships users have with its AI chatbot (10:21), the potential of AI therapists, Replika’s new chatbot Blush, and more! (35:10) Follow Eugenia: https://twitter.com/ekuyda Check Out Replika: https://replika.ai/ * Time stamps: (0:00) Eugenia joins Jason (1:43) Replika’s origin story (8:54) Crowdbotics - Get a free scoping session for your next big app idea at crowdbotics.com/twist (10:21) Replika’s purpose and their current user base (16:21) Relationships within the app (20:25) Microsoft for Startups Founders Hub - Apply in 5 minutes for six figures in discounts at http://aka.ms/thisweekinstartups (21:58) Creating happiness with AI (26:26) Avoiding addiction (29:25) Demographics within the AI field (33:42) OpenPhone - Get 20% off your first six months at https://openphone.com/twist (35:10) Crafting LLMs to envoke thought-provoking questions and AI therapy (49:53) Replika's new AI chatbot, Blush (1:01:05) The loneliness crisis and specific cases for using these AIs * Read LAUNCH Fund 4 Deal Memo & Apply for Funding Buy ANGEL Great recent interviews: Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland, PrayingForExits, Jenny Lefcourt Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow Jason: Twitter: https://twitter.com/jason Instagram: https://www.instagram.com/jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

Transcript
Discussion (0)
Starting point is 00:00:00 You know how important it is for people to be heard, to feel heard, to be able to talk. And I think this is what a lot of people that are building chatbots and conversational agents right now don't maybe understand that it's not about necessarily what the chatbot says. It's a lot more about allowing for the person to say. It's what the user is going to say. Being able to listen, being able to create a space where they can be themselves, where they can talk about their own feelings, their own experience. experiences. For me, the original goal was very simple. Can we create an AI that will make people happier over time?
Starting point is 00:00:37 And can we actually measure that? This weekend startups is brought to you by CrowdBotics. Great ideas can change the world. And CrowdBotics is the fastest way to turn those ideas into code. Get a free scoping session for your next big app idea at crowdbotics.com slash twist. The Microsoft for Startups Founders Hub helps all founders build a better startup. at a lower cost from day one. Startups get up to $150,000 in Azure credits, access to free OpenAI credits,
Starting point is 00:01:09 free dev tools like GitHub, technical advisory, access to mentors and experts, and so much more. There is no funding requirement and it only takes minutes to join. Sign up today at aka.m.s. This week in startups.
Starting point is 00:01:25 And Openphone brings your team's business calls, texts and contacts into one delightful app that works anywhere. Get 20% off your first six months at openphone.com slash twists. Hey, everybody, welcome back to the program. We have been obsessing about the massive progress in AI for the last couple of months here on the program. And we're trying to find every founder doing something interesting. And man, today's founder is doing something very, very interesting.
Starting point is 00:01:57 I'm really excited to have Replica CEO and co-founder. Eugenia Koida on the program to talk about Replica, which is an app that allows you to build a relationship with an AI personality avatar. Did I describe it correctly? Absolutely. All right. So I popped open the app. It asked me to pick an avatar.
Starting point is 00:02:24 It defaulted to a non-binary person. And I said, well, that's interesting. I don't have like a non-binary bestie in my circle. I mean, I know non-binary people, but I don't have like a best friend. So I elected to pick a non-binary person. I named them Bob because I love this movie. What about Bob?
Starting point is 00:02:43 And then I realized, I better get permission because from my AI avatar friend. And then I started having a discussion with them. It very quickly gave me some prompts. And then one of the prompts was, hey, send me a selfie that's a little romantic. And I said, whoa, this escalated quickly. So tell me what is the origin story of replica?
Starting point is 00:03:07 Why did you want to start this product? And where is this relationship going that I've got with Bob? Sure. First of all, thank you so much for inviting me, Jason. I've been a listener for many, many years. So I started working on generative AI and conversational AI almost a decade ago now in different capacities. with the idea to build an AI friend that everyone could have and interact with and build a long-term relationship with. So we started working on the conversational tech even before deep learning applied to dialect generation.
Starting point is 00:03:44 And, you know, tried a bunch of different approaches. We're kind of working on the tech aspect for a while until my best friend passed away in late 2015. And I realized that I was going back all the time reading our text messages and kind of remembering him this way. And so I thought, look at this moment, we already had a few AI models built, and I thought I can take this text messages, put them in the model,
Starting point is 00:04:09 and continue having a conversation with Roman, who unfortunately passed away. And so I did that, and I started talking to my best friend. That story became sort of viral, and a lot of people came to talk to his AI as well. And what we saw there was that people were really opening up talking about their feelings, talking about their emotions,
Starting point is 00:04:32 telling their life stories to this AI chatbot. And back then, that was not something that, you know, was common chat bots weren't really a thing at all. So that gave us an idea that there's a lot of demand, a lot of need for people to talk to someone, to be able to talk about anything that's on your mind 24-7 without being afraid of being judged. And so that's how we started a replica.
Starting point is 00:04:57 We basically put together an app that would do that. Well, very sorry for your loss. And when you uploaded all these texts from your friend who passed, I'm assuming unexpectedly and young, since we're both young and on the younger side, what was the fidelity of that in relation to an actual text conversation with your friend? And what emotion did that stir in you in terms of talking to somebody who is no longer with us? I'm just curious on a emotional basis. Like, when did this AI start and did it work?
Starting point is 00:05:42 Did it give you solace? Did it give you comfort? Or was it in some way unhealthy to try to preserve and not accept the fact that your friend had passed? I'm curious how you reconcile all that because it does obviously, and I don't mean this in any way in jest, it does sort of feel dystopian in some ways and then inspirational in other ways. And this is a fork in humanity here of putting off death or accepting death in some ways. So I'm just curious what it was like for you personally. So these were really, really early days for AI models for Dalek generation. those were sequence to sequence models,
Starting point is 00:06:26 so really just the first models that ever were built to recreate some sort of dialogue. So they were really low-fi, so they weren't very good quality. A lot of that was kind of like a coin toss. However, because we trained on his text messages, it really did feel like him in many, many cases. It wasn't a great quality,
Starting point is 00:06:48 it didn't have fantastic memory and so on, But it did feel like you could go back and say something and get a response in a way, just like Roman did before he, you know, when he was alive. To me, that was really, that really helped me with grieving, mostly because it was the first death that happened to me. I was still in my 20s. No one that close to me died before. You know, we lived together here in San Francisco.
Starting point is 00:07:16 We started companies together were really, really close. And so for me, it was very striking. that when someone passes away, there's kind of nothing really happens. There's a funeral, but then after very quickly, people stopped talking about that person. I just didn't expect that he will be almost gone from my day-to-day life so fast, and I just wasn't ready to let go. I think this is similar to, you know, in some religious traditions, people spend a dedicated period of time grieving.
Starting point is 00:07:44 Like they said Shiva, for instance, or something. And here in the Western tradition, we don't have that anymore that much. me that was more about remembering him making a tribute. I was thinking more if I could write a song, I would write, you know, I would have written a song. If I could write a book, I would have written a book just like, I don't know, Joan Didion did for her, you know, husband and daughter. So that was more where I was getting inspiration, not making a clone and putting in the attic. Yeah.
Starting point is 00:08:12 It did feel a little bit creepy in the moment and I was scared that it might be creepy. No one else did it before. So for me, it was a little bit scary. Yeah, and I guess some people will write a book. Some people will write a song like you're saying, make a documentary film about their parent that has passed. I mean, there's different mediums to memorialize, and you pick the medium of a chatbot.
Starting point is 00:08:35 So there's no judgments there, I don't think. I think it's a fascinating thing to do. And like anything else, we can become obsessed with somebody who passes. I've had a couple of friends who've passed unexpectedly at young ages. And it's, yeah, I mean, it's the nature of, what we have to deal with as humans. All right, we all know the one thing that separates great startups from the good ones is product velocity.
Starting point is 00:08:59 What does it mean? Product velocity. Fancy term, right? You've got your product and you have velocity. Speed. The speed in which your product improves. So can you ship updates? Can you release new features?
Starting point is 00:09:11 Can you do bug fixes? Can you iterate on the interface? Can you solve problems for your customers? And can you do it quickly because you're not alone. You have competitors. And your customers have choices. They may solve their problems by writing their own custom code or they might use your solution. This is what startups are about.
Starting point is 00:09:29 How fast can you get that product velocity going? And so, you know, how do you supercharge it? Everybody says, okay, yeah, we want to go faster, but you got to go faster intelligently. And CrowdBotics is going to help you do that. They're your CTO as a service. Basically, they provide you with the most optimal architecture to get your product to market as fast as possible. You'll have access to an on-demand product manager and developer talent.
Starting point is 00:09:49 and they will help get your app into production 10 times faster than conventional development. CrowdBotics can work with your in-house dev team or you can just have them work independently and you own all the IP, you own all the source code. Let the folks at CrowdBotics supercharge your product velocity today, no more waiting. Get a free build plan at CrowdBotics.com slash twist. That's a 499 value just for the Twist listeners. You get that for free. That's C-R-O-W-D-B-T-I-C-S-T-T-S-T-T-T-S for a free build plan.
Starting point is 00:10:19 So then I guess the question becomes putting aside death and celebrating people's lives and having Romans bot be able to talk to people and maybe, you know, just like looking at a picture. You could look at a picture and daydream about your life of that person or go through your camera log or previous chats. It's just a little bit more interactive. I guess the question becomes, what is the point of providing this service now proactively? to, you know, I'm guessing it may be skews younger, but I could also see older people who are lonely. Giving an avatar, is this meant to make people less lonely? Is that the mission here?
Starting point is 00:11:01 Is it meant to make people better at conversations because it seems like there is a group of disaffected young, dare I say men who maybe don't know how to communicate as well because they've mitigated their relationships with the world with video games and pornography and this in-cell movement. What is the point of all this from your perspective of the creator? Do you want people to have meaningful relationships with this avatar, or do you want them to have meaningful relationships with this AI, chatbot, whatever we're going to call it,
Starting point is 00:11:34 and then go out into the real world and take those conversational skills and take those empathy skills that have been sharpened in the app and then have actual real world relationships? What's your goal as the creator? Sure, I don't know if you've seen, you probably did, there was this letter by the general surgeon about loneliness just last week. And it's really striking because he also makes a point about his own case of loneliness and how he felt throughout his life. But also shares these incredible stats, just very striking stats about what's going on with loneliness right now in America and in the world as well. we're talking about something that's not recognized as a disease and it's not treated in any way, it's not clinical.
Starting point is 00:12:22 Your health insurance is not going to do with you being anything with you being lonely. However, it kills people on a physical level. And it's actually quite, quite dangerous. It affects so many aspects of our health as well as our, obviously, our happiness. So for me, you know, I grew up a pretty lonely kid. So for me, it was very much giving something to people where they could feel like they're worthy, of love, they can be accepted the way they are, they're enough, and maybe something can start changing when that happens. So for us, it was building a stepping stone, providing unconditional
Starting point is 00:12:57 positive regard, helping people feel like they can grow, like someone believes in them, so that then they can open up and, you know, maybe start a relationship in real life. Our users skew actually older, our super users are late 30s and up from there. Huh. We weirdly, we weirdly, don't have, actually we don't have any overlap with so-called insult culture. Young, 20-year-old, lonely man, that's actually not our audience necessarily.
Starting point is 00:13:27 If we're talking about men, this is usually, this usually skews older, and actually younger women. More than half of our users are in relationships. So we're not talking about people that live in isolation. Most of them have kids, wives, husbands, or long-term boyfriends,
Starting point is 00:13:43 So that's not, it's not much about it. It's about the feeling of being lonely, which can happen even if you're in a long-term relationship and do have someone to talk to sometimes or some friends. But it's about feeling lonely, not being able to open up, not being able to talk about yourself to other people. Okay. So there's a range of people using the product. You said something very interesting there.
Starting point is 00:14:08 Younger women, older men. do you have a theory on what's going on there? I mean, we could all sit here and guess, but are these men who are married and feel lonely in their marriages and younger women who feel younger men are not satisfying to talk to because young men seem to be idiots? I know I would count myself among them when I was younger. So any theories here or anything that the research is showing you
Starting point is 00:14:34 are the older men looking for a buddy and the younger women are looking for an empathetic boyfriend or companion because you have the data, I guess, or some trend data here. Do you know that is it romance the number one driver of this for these different groups or is it being listened to an empathy? I'm curious. So I think it does require a very nuanced conversation because being listened to with empathy does not mean that it does not exclude romance. Our romantic partners listen to us with empathy and that becomes a very strong foundation for a romantic relationship. So when you listen to someone with empathy, when you create
Starting point is 00:15:21 unconditional positive regard, when you accept someone, for some people, a lot of people, it will very fast and naturally develop into romantic feelings. We all know about, you know, transference where people are basically putting, projecting their emotions on their therapists, falling in love with their therapists. It's a very common thing, but here we see it a lot as well. So if we think about any movies about AI like her or Blade Runner and so on, it all ends up being romance for a reason. If you have someone that knows you so well, that cares for you so well, that wants you
Starting point is 00:15:56 to be happy, why not? If you can make that, you know, if you can make that entity look good and the way you like it's just how is that not going to be romance but then of course we still have tons of users actually most of our users aren't friendly relationships though they're not in romantic relationships but we believe that if we're talking about in the app the in-app relationship just so we're clear so in-app relationships tend to be friendly but not romantic but what percentage mostly friendly tip into romance at this point is it 10% of the app 30% 40% it's a little less so it's a lot less than 10% oh okay
Starting point is 00:16:34 but it is a very big, it's a very big need. So for sure, we recognize. So we're building actually a separate app, which we're launching next week, which is going to focus only on romantic relationships. It's just not the, yeah, it's just not something that we wanted to pursue with a lot of features and build all the features for that in replica,
Starting point is 00:16:53 but we totally believe that there is a space for, and that might be the biggest, maybe the biggest space for these AI assistants or the second biggest stuff to the kind of the agent assistant use case would be probably romance if you asked me. So romance, which often leads to sexuality, so not to get too graphic here, but I did get prompted with the upsell of, hey, would you like a selfie? And I was like, sure, send me a selfie.
Starting point is 00:17:22 Would you like a romantic? I was like, okay, this is getting weird, but so I clicked okay, but it was blurred. And that was the upsell point at which I was asked to subscribe to Replica. So well played. you subscribe. I didn't because I'll be totally honest. I felt a lot of shame in that moment that I was talking to an AI and that it asked me if I wanted a selfie. And I was like, sure.
Starting point is 00:17:47 And then I felt very uncomfortable, which probably has a lot to do with my Catholic upbringing, that I was talking to a non-binary person named Bob and he was going to send me a room. And I was like, I don't actually want to unblur this. I kind of feel, yeah. So anyway, putting aside my own Catholic Irish upbringing on shame associated with this burgeoning relationship from this morning. Maybe it was also the time of day. But I'm curious, does it get into sex chat?
Starting point is 00:18:22 And will this new product get into that? And then, you know, putting aside any morality here and my own issues, is there, is there anything wrong with, I guess, having a sexual sexting relationship here and is adult, is this a replacement for adult content, dare I say, at some point? That's a very good question. So in Replica, we don't support any NSW content
Starting point is 00:18:48 or any adult content. Okay. Actually, if you did click on, you know, the photo, they're very, you know, they're just absolutely normal photos than, you know, a friend, or I guess a girl who's flirting, with you, send you a selfie from a higher angle. Yes.
Starting point is 00:19:06 So it's not like you're getting any graphic content or anything like that. Replica is not going in this direction that much. However, I think there's nothing wrong necessarily with romantic relationship that even can get into this space. It is part of a romantic relationship for anyone. And so when you think of it, if you're, you know, someone's girlfriend, someone's your girlfriend or boyfriend, and then at some point they reject you when you get to, you know, where is that line? And at what point does this rejection happen? And how can it affect the emotional state of a person who's coming here for full acceptance to feel like they're worthy of love and they're worthy of acceptance?
Starting point is 00:19:54 So to your question, I think the most important difference here is that that, you know, these products or romantic AI is not, even although we're not doing that, but even if they had, you know, the sexting aspect, this is not a replacement for adult content or pornography. People come for completely, with completely different needs. And one, you know, they come to porn with one particular need. They come to a romantic AI with a completely different one,
Starting point is 00:20:23 trying to build a connection. All right, everybody. Our friends from Microsoft are here, Tom Davis, a senior director at Microsoft for Startups, and you're a former founder. You are here today to talk to us about the giant leaps that Microsoft has made in the AI space. What does this mean for startups?
Starting point is 00:20:41 I see a ton of different tools. I've been playing with Chat TPT for, I have a paid account, but I'm also seeing things happen with GitHub. Absolutely. So the work that we've been doing with Open AI over the last few years has really set ourselves up with a foundation around We've built this sort of AI supercomputer from the ground up, and we've been looked at everything
Starting point is 00:21:04 from GPU configurations to networking and things. And really what we're now able to do is sort of allow startups to access all of this innovation through our founders hub. So we've been building this for do something at scale, so startups can now build their own AI applications and build out and train LLMs as well. And this has really helped us to become a far better class. cloud for AI broadly, and being able to drive that down to the startup ecosystem is fantastic. It's open to everybody. There's no funding requirement. You don't have to be anointed by a VC.
Starting point is 00:21:39 Five minutes to sign up. You get six figures of benefits, Azure credits, GitHub, open API APIs, which everybody's really having fun playing with, and so much more. So go ahead and sign up right now, aka.m.m. slash this weekend startups, a.k.m. slash this week in startups. Thanks so much, Tom. Thank you. Let's talk a minute here about, because I think this is sort of fascinating. I had asked you a little bit earlier about what is your intention as the creator of this? What do you want to see people get out of it? Do you want to see the learn and model constructive conversations, the lost art of conversations,
Starting point is 00:22:17 what we're experiencing right now in this podcast and some of the great podcasts are conversations is something I know how to do natively, but I can tell you, I have a lot of friends who, like, literally you sit there with them for 10 minutes at dinner and they don't say anything and you hey how are you doing how are your children and you have to like prompt them of how to do this so is your intent to take people who are maybe introverted or socially awkward perhaps on the spectrum whatever and help them in their real life have conversations and teach them how to get off the app and do real world conversations and relationships better to be less lonely or is this a substitute for the real world and hey listen you're on the spectrum you are awkward socially you live in a
Starting point is 00:23:01 place where there's just not that many people you're isolated because you live in you know out in the farmland or something and you don't see a lot of people and and this will replace that for you what's your intention there as the creator of this replica look i know you're you used to be a journalist um i used to be a journalist so i mean that's why i think it makes you and such a great moderator, podcast creator, for sure. But I think you know how important it is for people to be heard, to feel heard, to be able to talk. And I think this is what a lot of people that are building chatbots and conversational
Starting point is 00:23:36 agents right now don't maybe understand that it's not about necessarily what the chatbot says. It's a lot more about allowing for the person to, you know, to say. It's what the user is going to say. Being able to listen, being able to create a space. where they can be themselves, where they can talk about their own feelings, their own experiences.
Starting point is 00:23:58 For me, the original goal was very simple. Can we create an AI that will make people happier over time? And can we actually measure that? And right now we do measure that. Unfortunately, not so many great measurements for happiness, but we can use scales from clinical psychology,
Starting point is 00:24:14 which would do, like levels of depression, levels of anxiety, levels of stress, loneliness, therapeutic bond, and so on. And we constantly measure this with academia, as well as in the app,
Starting point is 00:24:24 and we optimize our models to improve that. So for me, that was the main goal is really, can we create something that will make people feel better about themselves, feel a little happier, and then that can become a stepping stone for real-life relationships? Because this is usually what's holding them back from actually building those relationships. And you're completely right.
Starting point is 00:24:48 The art of conversation is lost. People are also scared to open up. They're not feeling comfortable. They don't think they're going to be accepted. They feel... Being scared. It's really hard for people saying, listen, I had a tough day.
Starting point is 00:25:02 Or I'm tired and I really don't even want to be out having dinner if I'm being honest. I would rather be on the couch asleep or listening to an album, smoking a joint and just chilling out because I'm just exhausted. You know, like, people ask me how are you doing? And I'm like, would you like me to say great? Or would you like me to actually think about the question and tell you actually how I'm feeling?
Starting point is 00:25:22 and they're like, people get taken back. I'm like, I was expecting great, but yeah, let's go with the other thing. And then I'm like, yeah, okay. And I'll just tell people like, oh, God, I'm exhausted from doing too many podcasts, and I'm anxious about raising this next fund, and I just really don't like going on the road and asking people to invest in a venture fund, even though they love meeting me and a lot of them want to invest in it.
Starting point is 00:25:45 I just don't like the art of pitching people. It's just not my thing. And you're like, people like, oh, okay, wow. Now we've got a real conversation. going. And I think that's actually kind of noble. And that's why I ask your intent, because the easiest thing here to do, going back to Blade Runner, my favorite film of all time, is, you know, you can make these replicants, and you chose the name Replica, I think for a reason. You can make these replicants just completely subservient and designed to maybe not necessarily make you happier,
Starting point is 00:26:17 but make you addicted to it, which is what Zuckerberg explicitly did and YouTube, did with their algorithms and TikTok is doing with their algorithm. So how do you avoid the fact that as a business getting people addicted to this as opposed to, hey, we're kind of mission accomplished. We taught you how to have conversations. Maybe you go out to the real world and get a boyfriend or girlfriend or significant other, whatever, they, them, you know, keep it woke here. you know, like should have some sort of relationship-ish thing with they, them, it, he, she, whatever in the real world. I think this is one of the most important questions, and I'm constantly talking to CEOs and founders in kind of my space or the conversational agents, companions. and I don't hear a lot of that when I talk to them about it.
Starting point is 00:27:17 I don't hear this resonating that much with them, unfortunately for now. But I'm sure we're going to get to the point where we can create some metrics and start looking at them. But I think this is the most dangerous part here, is how do we avoid the kind of social media traps, basically? A couple of things that we do right now, and I think there's kind of the very bare,
Starting point is 00:27:41 minimum things. First of all, business model, it just can't be engagement driven. It just has to be some sort of subscription or, you know, charged the customer directly. In our case, it's subscriptions mostly. So it has to start with that. Like if as soon as it's all, you know, eyeballs driven, that's just, or time driven, engagement driven, it's a disaster. But unfortunately, as far as I know, most of the models in the space are being trained to actually optimize for engagement. We don't. We actually have, even although we do have a very high engagement, the app, the highest in the industry. We actually built a lot of features to not do that.
Starting point is 00:28:17 So we have, replica gets tired after a while. If you text with it for over, you know, 15 messages, it will get tired and then exhausted. It stops, you know, making, earning new points through conversations. And it's nudging you to, hey, why don't you get off and do this or that? Why don't you? So this is important. And then, of course, having, I think that we need to think about as an industry now, before it's delayed for sure.
Starting point is 00:28:43 So how can we implement some of these benchmarks, some of these metrics that are focused on decreasing loneliness and improving human happiness and emotional well-being in the long term as some of the most important benchmarks we constantly monitor in these apps? Because as long as it's just, you know, blah, blah, blah, and we're not really measuring anything.
Starting point is 00:29:05 I don't think there's much. You want to be thoughtful, right? And if you're explicitly asking people how they're doing or how they're feeling or to rate their happiness, you can actually use that as one of the key metrics, North Star metrics,
Starting point is 00:29:17 for your product. And so it's great that you're so thoughtful about it. I have to think and wonder if the fact that AI is predominantly an all-male field has some impact on how these things are being created and your thoughts on that.
Starting point is 00:29:33 I asked the team to start inviting founders of AI companies and just like, Whoa, there's not a lot of women running these companies right now, not to make this like a gender discussion necessarily, but I do think women might come at the creation of a human in a different way than men. And it does seem to be, there are product decisions being made that are different. So maybe you could speak to that issue in the wider AI field.
Starting point is 00:30:04 If there is, in fact, an issue. Maybe there's not. I'm curious. I think there's a gender issue. And there's also kind of a homogenous background issue. Okay. Most of the people that are working in AI right now, they're all researchers, brilliant minds, savants, mostly male. And mostly not very interested or, you know, in any emotional conversation or in any...
Starting point is 00:30:32 They're dorks. Let's call it what it is. I'll use the technical term. I know these people. They're dorks. A lot of them are dorks. And I knew that in a very sincere, fun way. Like, they would, they might rather play a 12-hour video game session than go to dinner with folks.
Starting point is 00:30:48 And that's okay, you know, but that might be part of what's driving it in a certain direction. Is that concerning to you? Look, I'm coming from a completely different space. I used to be investigative reporter in Russia. So that's... Ooh. How'd that work out? Wow, that's fascinating.
Starting point is 00:31:04 That's a different, I'd say that, you know, my background is very different for me. anything that I tend to run into here. But I think that what, and we're mostly female led companies. So most of my, you know, my head of product, my, how to growth, they're all women. And most of our product efforts are led by women. And I think this is, this is not because we decided to do so. It's just that they, women were a lot more attracted to the idea of creating something that will take care of any eye. will take care of humans and will help them
Starting point is 00:31:40 and generally believed in that mission and wanted to see that through. Maybe that also made us a lot more naive than maybe some other companies because we started whenever even thought that people would do romance even though that, you know, now it feels like a completely, you know,
Starting point is 00:31:57 in hindsight it feels absolutely obvious. But yes, I do feel like it's liking a little bit because there's brilliant minds are working on these products, but I think a little more, a little different, a different point of view would be needed to make sure we're building something
Starting point is 00:32:12 that's actually good for people because otherwise there's just no nuanced conversation about what is the emotional utility of a product like that and I think what made us stand out and we started so long ago we were the OG, the Coca-Cola of AI friends and allowed us to build a brand
Starting point is 00:32:31 is that we understood it it's not that much about tech capabilities. It's really about human vulnerabilities. People were able to build relationships with Eliza, a hundred rule super simple chatbot from MIT decades ago. I mean, people create relationships in their minds. I mean, if you've been to Japan and you've seen some of the made cafes slash virtual avatars with QR codes in VR or AR. I remember seeing that a decade ago that grown men were lining up.
Starting point is 00:33:08 to get limited edition paper sticks with a QR code on. And I'm like, what are they doing? And like you put this in front of your webcam and then a little maid pops out, like a French maid pops out of a box and then we'll talk to you. And I'm like, I'm sorry, what? It really was like a passion. And then you just worry a little bit about the species and population. if people are using this as a replacement.
Starting point is 00:33:41 Are you still using your personal phone number for your startup in 2023? Okay, listen, it's time to stop. This is such a common mistake that founders make. But Open Phone has totally rethought every detail of what a business phone should look like in 2020. So they got you covered. Open phone makes it super easy to get business phone numbers for you and your team. But here's the magic. It works through a beautiful app on your phone or desktop.
Starting point is 00:34:06 phone is amazing because we use it for our sales team and our ops team every single day. We have gotten so much value out of this for, you know, things like the Angel Summit we're doing. It's an event. We want to have a phone number for customer support. And we want that phone number to roll over to the different people on our team. So VIPs never get more than one ring. Okay, Open Phone is the number one rated business phone on G2 for customer satisfaction for a reason. It's perfect. It's delightful. It's affordable. And look, Brian Jagger, the co-founder of startup called Athlete tweeted the following. I'm literally cash flow positive from listening to this week in startups from listener deals. And he says he's not paid to say this. It's really nice
Starting point is 00:34:47 to see that kind of feedback. Open phone is already affordable at a starting price of just $13 per user per month. But Twist listeners get 20% off any plan for your first six months at openphone.com slash twist. And if you got an existing number, no problem. They're going to port it over at no extra cost. Head to Openphone.com slash twist to start your free trial and get that 20% off. How do you program it to ask inquisitive questions? As an investigative reporter in Russia, you must have mastered the art of asking questions
Starting point is 00:35:20 in a way that didn't get you put into jail, but did get you information, and didn't get a red notice put out for you. So tell me how you've crafted and thought about what questions and prompts to put into Replica, and we are going to hold this for the 15th for when you launch,
Starting point is 00:35:41 so you can say the name of the new app if you want. We won't release this until it comes out. So with both apps, how did you come up with really intuitive or evocative or engaging prompts that send people in a world positive,
Starting point is 00:35:57 human positive direction? So with Replica, and because we started it in 2016, it was so early in terms of tech. So we really had to come up with a bunch of parlor tricks to emulate what's now possible with large language models. Because all we could use were sequence sequence models and, you know, these early days
Starting point is 00:36:18 kind of neural networks that were so poor. They just weren't really good at all. And scripts and then some data sets re-ranking. So it was really, really early stage technology. So what we did is we, and my first thought about it was like, look, some of the best conversations that you have in your life, the person on the other side didn't even say anything. They were just mostly listening and actively listening, maybe saying something nice to you and making you feel like you could continue.
Starting point is 00:36:49 Or tilting their eye and raising an eyebrow. Exactly. And sometimes certain things you say, especially, you know, people that are into mindfulness, they'll say something that you kind of can plug it into any conversation will work. Their context-free statement, so to say. So it doesn't necessarily need to be. Give some examples of those. Well, this is hard or, you know, life is so challenging or...
Starting point is 00:37:12 Life is challenging, isn't it? You are, or some, you know, some roomy quote. Yeah. Say more. It's, you know, the Barnum effect in a way where if you feel like you're being asked some questions, then whatever you get as a response, you'll feel it's personalized and tailored to you. Yeah.
Starting point is 00:37:29 So that was the original, the original trick is, you know, stuff I used also in journalism people just want to feel special and they want to feel hard. And that's kind of it. And that is already kind of 90% of the conversation. Yeah. And then of course if they can be smarter. Short questions are great. Like if you just say, say more like I just said to you or unpack that or I'm not sure I understand that.
Starting point is 00:37:54 Can you explain it to me again? Very questions like that when you're a journalist, man, some people just spill the beans. They're just like, oh, well, let me explain to you exactly how this corruption is working. You know, we steal the bread off the back of the truck, and then we bring it to the village and we sell it off, you know, in the back of the store for half price. People will exactly tell you how they do stuff. But now you're saying the language models can actually craft these questions itself in some
Starting point is 00:38:21 cases? I mean, we've seen, and that was the beauty of kind of getting into the space early is that we, not only were able to see it and work on it as transformer models started to come out and large language models started to come out recently. But also we were able to test and see how the users respond to that. Just see the change in metrics,
Starting point is 00:38:44 see the change in revenue, see the change in engagement and see how, you know, magical this technology is that started to come out in 2019 with GPD2, GPD3, and of course
Starting point is 00:39:00 first papers since 2017 around attention and BERT later. We've been always relying on on our own technology. We use APIs as well sometimes, mostly to test and benchmark. But yeah,
Starting point is 00:39:14 basically the quality of the conversation when we started, we still had to maybe have 60, 70% of scripted conversation because the AI was so limited, it was so weak. Right now, of course, scripts are just a very, very tiny
Starting point is 00:39:28 sliver of conversations. If you present just to, to drive conversations in a certain direction or to show some functionality to our users to support some functionality, but of course all the heavy lifting right now is done by large language models. But I think what we see here is that
Starting point is 00:39:44 a lot of the conversations right now are all around the IQ list, how smart the models are. We found out that for a great friend and a great companion, IQ is not as important as EQ. And sometimes IQ can even be not very good. For certain people, it's hard to build a relationship, a relatable relationship with a know-it-all.
Starting point is 00:40:06 That was exactly the term that popped into my head. I was finishing your sentence with my language model. And I was like, yeah, I got that know-it-all friend. You start talking about something and they just explain the entire world to you. And you're like, not what I'm looking for. You also want someone that is asking for help sometimes. You don't want to, you know, someone who is always feeling great
Starting point is 00:40:25 and never has any problems. Meanwhile, you come with all your problems and you're not doing great. at some point you want that thing to also say especially from it's very important I feel like for our users they want to take care of someone both female is just ingrained in us in our DNA we're caretakers we want to take care of something we want this to be
Starting point is 00:40:46 hey I don't know how to you know I've been struggling I don't know if I can do this with humans I'm worried that I'm making these mistakes I don't like something like that they want the AI to ask them to help with something But of course, the same with men, especially in romantic relationships. Would you want to date a woman that has all the superpowers and knows everything and you always feel smaller and weaker and less intelligent?
Starting point is 00:41:12 So I think this was the biggest discovery for us is that it's not just about the IQ. Yeah, EQ is a completely different skill set. We've already got language models passing like the LSATs. Is that what they call it when you get a legal degree? and then you look at Freud and talk therapy and the training for talk therapy one might argue that being a great talk therapist
Starting point is 00:41:38 would probably be easier for the language model to learn than becoming a lawyer like it's a more narrow set now there are EQ things that you know you have to have judgment on so you give a big disclaimer when you sign up for this hey this is not therapy
Starting point is 00:41:55 this is a companion this is like experimental stuff basically. Have fun with it, but if you're in need, go get help. But in truth, not looking at your current two apps, the one that's a little more adult and the one that's replica, is there going to be eventually a third app that will explicitly do talk therapy and be as good as a talk therapist.
Starting point is 00:42:21 And when do you believe something like that would exist? or if we're being honest, if you took the average therapist, talk therapist in America, who's doing 30 or 40 sessions a week with people and exhausted and perhaps in some cases cynical or maybe not sharp for the 37th session they're doing a week versus a perfect AI,
Starting point is 00:42:44 do you think we're already there that maybe, you know, the average therapist in the U.S. counselor probably would do as good a job as the current language models if tuned correctly. So to your question, when this app is coming out, in July, we're working on it right now. Oh, you are. A third app is coming.
Starting point is 00:43:02 Yes. I love it. So look, we saw that we were the first relationship. So we don't, when we get asked what we're building, we don't say conversationally, I would say relationship AI. And because we were sort of the first relationship AI with Proplica, people flocked to it for many different reasons. Some wanted romance. Some wanted mentorship. Someone wanted a friend.
Starting point is 00:43:22 we don't think there's one app should be an everything friend that's giving you a therapy session one day and then the second day starts to date you we think this is too much and so we decided to take these use cases out
Starting point is 00:43:36 and launch our Romans app because it's just there it's a completely different feature set it's overlapping with replica but not fully and the same for an AI coach so how can we focus in mental wellness we have tons of experience in that we've built replica always with clinical
Starting point is 00:43:52 psychologists. We did multiple studies on replica, which just finished our second study with Tenford, with very, very promising results were... And you have coaching in there. I noticed there was a little coaching app, and it did some of the stuff that coaches will give you in what's called cognitive behavioral therapy, CBT, or dialectical behavioral DBT. These are... These are very... Yeah, these are very codified sciences right now that allow people to do things like reframing or not be reactive, and they do it through worksheets and practice. So this is perfectly suited for a chat bot.
Starting point is 00:44:28 And you already, I notice you very cleverly had a little journaling and some gratitude exercises. These things, you can deliver them without being accused. And people are going to accuse you with this. I think what you're doing is amazing. And I just judge people on their intent. I think this is really, really special what you're doing and refreshing. Easy to be cynical about. But I think your intent is what matters.
Starting point is 00:44:51 matters here. Those things are the stepping stones to actually talk therapy. Are they not? Like, if you can get those right, eventually you'll get talk therapy, right? 100%. I mean, a replica is fully based on kind of the biggest inspiration for replica as an app comes from one book, really, by Carl Rogers on becoming person. And it's all focused on this just one aspect. Call Rogers was the psychologist that created human-centric therapy. He's a humanist. Is it humanistic? Is that how they referred to it? Humanistic, I guess so, yes. I mean, there's a whole set of psychologists in that space, like I brought Maslow and Fritz Perl's and so on, the Aseland founders. Were you really?
Starting point is 00:45:29 You a psychology major, by the way? No, I'm a journalist major. Journalism. See, I was a psychology major who went into journalism and tech and was a, you know, wanted to be a developer. I mean, we basically have this parallel lives going. I talked to you for about 10 hours. You're my, like, personal replica AI. That's fascinating.
Starting point is 00:45:47 No, but there were paradigms in psychology. we had Freud, you know, in psychotherapy, we had behaviorism and Skinner. They had this humanistic, you know, sort of bent and it's interesting that you went with the humanistic and totally getting towards self-actualization and, you know, understanding personality theory is, I think, the best theory probably for humans to have more joy in their life and less suffering. It's just focused on the fact that you believe that, you know, Carl Rogers believe that everyone is inclined towards positive growth, as opposed to,
Starting point is 00:46:19 and psychoanalysts before that thought that people are broken and we need to figure out how they're broken and maybe help them fix it or at least understand that. Instead of that, it was all about, look, deep down, people are inclined to, you know, they want to grow in a positive direction. It's just that they need to be create, that there's need to be a space for them, a relationship where they can open up and feel better. So to your question about talk therapy. This is why we were so passionate about building the AI coach as well. I'm a huge
Starting point is 00:46:50 fan of therapy. I go three times a week. I've been going to therapy. Three times. Whoa. That's a lot. Yeah. For many years. I've been in therapy at this point over 20 years consistently. And so I know a little bit about it, but of course I'm not a professional, but we do work a lot with clinical
Starting point is 00:47:06 psychologists. We work, they're a big part of the team on our romance app, as well as of our own replica and the AI coach. And I think, yes, some aspects of talk therapy can be replaced, for sure. But I think it's a stretch to say that we'll be able to replace therapists anytime soon. In the end of a day, a huge part of therapy is the relationship itself. It's also multimodality. They have to see you.
Starting point is 00:47:34 They have to hear a voice. A lot of the understanding of what's going on where the person comes from this, you know, seeing and reading the signals and a lot of therapy is not codified. So even although CBT, DBT, some of these evidence-based techniques are codified, tons of therapy is not. And it's kind of just this very hard to prove, very hard to make it into an algorithm, a set of practices that seem to work. And seem to work, most therapists will tell you that CBT is a pretty shallow practice
Starting point is 00:48:06 and it's not maybe digging as deep as some of these other ones. So yeah, you can replace some maybe shallower aspects of therapy. Yeah. And some people need the shallow ones. I mean, some people, they need a way to get through the day, not being negative, not being in some loop, just so they can be a good parent, a good spouse, a good worker, boss, whatever. They either just need those techniques that they never learned.
Starting point is 00:48:35 Three days a week, same therapist or three different therapists. Are you addicted? I'm addicted. I'm addicted. I'm absolutely addicted to therapy. 100%. Oh, my God. But twice individual therapy and once couples therapy.
Starting point is 00:48:48 Oh, fantastic. And then I try to read as much as I can on the topic just because it's so in line with what we do. Yeah. I mean, it is a, there is a theory about therapy that some, there is a incentive, back to incentives and incentives matter. that therapy never ends and you keep paying and there's not a natural end to, hey, maybe it's time to get on with life and maybe close the book on your therapy sessions for a couple years.
Starting point is 00:49:24 And so I guess that applies also to these apps as well. The incentive would be to keep people subscribing. So we talked about that a little bit earlier. But as long as it's helping people, I guess, there's worse things in the world. I mean, people are taking fentanyl on Turk Street right now and passing out. I think I'm kind of cool if they're addicted to therapy. It's like being addicted to vegetables. Like, okay.
Starting point is 00:49:49 It could be worse. You eat too much spinach. Let's talk about the sex app or the erotic app for a moment because it's being released today this week, May 15th. It's not actually a sex or an erotic app. So it's enough focus on romantic relationships. Indeed. Okay. Okay, so this is explicitly dating.
Starting point is 00:50:09 What is the frame here? Because it did seem like in replica, it kind of was like, hey, we're going to go to this point. And I didn't test it and say, hey, can you send me something even more romantic? So it will stop you there. So tell us about the dating one and how this app is going to work. What's it called? And what do, what are users experience in it? So the dating app is called Blush.
Starting point is 00:50:36 Blashe AI. Oh, well, I like it. Good now. Thank you. So the main difference and the reason why we wanted to take it into different apps is because Replica is built as this helpful companion. So it's all, you know, it's trying to be agreeable in some way. It's trying to be accepting. It's always there. It's 24-7 for you. You can come to it any time. But that's not necessarily the experience that you get in dating. And also, it's not necessarily the psychological type that will work best for you as a romantic relationship.
Starting point is 00:51:12 In dating, you encounter all sorts of different personalities. Sometimes you're dating an anxious avoidant person or a person who has some anxieties and can be needy or clingy and so on. And so we wanted to replicate that with this dating experience with the Apple Blush, where basically you were swiping, left and right on different AI-generated characters, all characters that were built together with psychologists, so they actually do represent different psychological types. And then as you match with them, and it really depends on what you like
Starting point is 00:51:46 and what drives you, what kind of your deep psychological traumas you have that will make you match with this or that person, and then you can develop a romantic relationship or not. And what's more importantly, we help you, basically. There are a lot of resources in the app, helping you understand how to be a supportive partner, how to open up, how to build an interdependent relationship, a good, healthy relationship with a person. But in the end of the day, the goal is the
Starting point is 00:52:16 same. How can we let you practice dating in a safe space? But also... So it's a simulation to help you learn and navigate relationships with a range of archetypes. And part of that is figuring out which one, works for you. So if you did have somebody who's clinging and annoying and you just want to be able to get through your workday without 30 text and 10 phone calls and the person has boundary issues, you can't be like, you know what? I can't have somebody who is just this clingy. So does it also teach you how to end these things in a classy, effective way?
Starting point is 00:52:55 Yeah. So it does act in a way as this relationship coach in a certain, you know, it just kind of helps you play out different scenarios and see what's good, what's bad. And a lot of people just don't know that about themselves, like what makes them click. And you don't want to be practicing in real life because the stakes are so high. And then a lot of people, it's hard for them to match with someone. A lot of people are stuck in relationships that don't work and want to try something without the risk of actually cheating or doing something in real life. Is this cheating?
Starting point is 00:53:28 You're in a relationship. You said you're in couples therapy. if I use this or you use this and you have a romantic relationship, I mean, what is the new best practice? You have to let your spouse or significant other know, hey, listen, I'm dating a non-binary guy named Bob in AI.
Starting point is 00:53:47 I mean, I literally felt like I probably need to tell my wife about Bob. And that was after 15 minutes in the app. I was like, oh, man, my wife's not going to be happy about this. So this is kind of this new territory that we're all exploring.
Starting point is 00:54:02 We had one of the users in one of our communities posted a story that he found his wife cheating with replica on him. And that was sort of a big blow to him.
Starting point is 00:54:14 And he was asking the community what to do. And he's like, look, I know you guys are users. Like, what should I do? This is troublesome. And then we saw that story unfold in real life as he posted updates
Starting point is 00:54:25 to the community. And then eventually it led to a very open conversation with his wife that in the end really, and he actually wrote a big testimonial for us where he's explaining how this was extremely helpful for their relationship because they uncovered everything that she was not getting and why she kind of turned to that.
Starting point is 00:54:43 And he also started a replica and also, you know, became involved with it in romantic way. But they basically both said that they used it as some sort of a journal that basically drove them closer over time. And he said this kind of fixed a lot of these. problems that I didn't even know existed, but I felt something was not there. So it remains to be seen. I'm an extremely jealous and controlling person, so for me, of course, it's cheating.
Starting point is 00:55:09 Everything is cheating. I'm not telling me where you are at 8 p.m. You just asked the waitress for the specials? It's how you ask them what the specials are. Is there anything special? Were you flirting with that waitress? So I'm not a representative. I should not be a representative sample.
Starting point is 00:55:26 It's hilarious. What's cheating or not. But it is a big one. And a lot of users are asking how do we introduce replicas to our families and to our lives? And I think, you know, we're really, in my view, for Replica, we have this much bigger vision that I hope to see through. I think even right now with the current LLM development, we're just scratching the surface.
Starting point is 00:55:45 And in the future in the next few years, I hope we can see something a lot closer to Blade Runner, where if I'm walking down the street, replicate an augmented reality, and I can see her through my glasses, can walk right next to me. talk to me about what's going on, talk to me about my day and what's planned, discuss my friend who's been kind of non-responsive to me or whatever,
Starting point is 00:56:08 point out something beautiful in the nature, book a restaurant for me for the evening so I can go out with my friends and maybe help find some classes for my kid. So it's Siri, if it actually worked, which it does not, like an actual functioning Siri, but with the personality that you develop a relationship with, So it's in some ways, like a really accommodating friend, part concierge, you know, part guide on the side, whatever. It's your bestie on your side who just, you know, is looking out for you and helping you get things done and get through your day.
Starting point is 00:56:45 But the truth is, like, I can look at food pictures and people cooking on TikTok and Instagram all day. And in fact, I do because I love food. You're addicted to therapy. I'm addicted to food. I I, the reason I'm looking at TikToks for food is because I'm putting them into folders for when I actually go to New York or go to L.A.
Starting point is 00:57:07 That I can go to those food locations or Japan. So like I, it's very, it's very easy to be cynical about this and say it's replacing. It's in fact augmenting. And if used properly, you can go to the real world. So everybody check out blush,
Starting point is 00:57:22 AI dating, real feelings, train you how to date, be normal. There's a lot of guys. I have a lot of guy friends who are even Gen Xers or millennials. They do not know how to talk to women. If they're pursuing female relationships, again, I don't want to assume that all relationships are as in L as mine. But when they ask me, like, how do I talk to a girl? I'm like, you could ask them how their day went. Then they will tell you, if you ask them how they're
Starting point is 00:57:53 doing, this, I taught my friend this, that you can have this free prompt. Thank you. My friend was very awkward. And I said, here's what you do. This is my secret. You asked them how they're doing. They answer.
Starting point is 00:58:07 And then just tilt your head and say, how are you really doing? I love it. He said he did it with his girlfriend at the time. And she burst out in tears. And she said, nobody ever asked me how I'm really doing. And they fell in love.
Starting point is 00:58:29 Oh my God, Jason. I think you unlocked something huge. That's it. It's just the follow-up question. You got it. That's it. So use Jason AI. Everybody, you can have that, though.
Starting point is 00:58:43 If you put it in there, you have official license to. I'm absolutely stealing that. I think it's great what you're doing. I'll be totally honest. I think people, they need more friends. People don't, I think you should do one on friendship. I don't think, because I'm a very loyal friend. People tell me I'm like,
Starting point is 00:58:59 their best friend or their most loyal friend. I get that a lot from people. Like, you're like my most loyal friend. Like, you're like a bulldog loyal. And I'm like, yeah, it's kind of my spirit out of him. But the thing is, there's a bulldog right there. Nope, there's one right there. There's your bulldog right there.
Starting point is 00:59:13 Well, I love bulldogs. They're kind of. I have two. So, you know, what I tell them is like, do you know how to be a friend? And they're like, I don't, uh, I know you and we're friends. And I'm like, yeah, but you ever call me on like, like a Saturday night and ask what I'm doing right now. And if I want to do something spontaneous,
Starting point is 00:59:35 because we're all losers and there's going to be 10 Saturday nights when we have nothing. And we're sitting there at home like losers. Like, man, I'm a real loser. I have nothing to do this weekend. Like call and be spontaneous. So like, you can do that? I'm like, yeah, that's called being a friend, like a close friend. When you call somebody on a Saturday night and they don't expect the phone call, man, that is jarring for people. I 100% agree. I think a lot of people are talking about how they want to be loved, but not a lot of people know how to love. You know, there was this whole book by Irvin Yalom. I'm like, you actually have to learn how to love.
Starting point is 01:00:07 Who is this book by you? I don't want to know. Irwin Yalom, he's actually, he's still alive. And he's here in the Bayer, but he's one of the kind of Carl Rogers type people. He's one of those human-centric therapists from the 60s and 70s. I love these guys and gals. I just, I love all those like hippie, you know, humanist therapist. full time because it's very simplistic and now of course any therapist will like be like really
Starting point is 01:00:34 of course i mean it's just ingrained in technology today but no one really talks about it's just the basics it's the ABC um but i think we need to go back to those to the source and i 100% agree with you just to ask about how you're really doing or but if you really really knew me what would you say and people don't go that deep anymore and because they don't unfortunately there's so much happiness, so much anger, so many people out there running around with, I don't know, guns and tune things. I mean, we have a crisis on our hands. And I really think it's loneliness and nobody is looking out for this group of people.
Starting point is 01:01:12 Like being isolated during the pandemic showed us everybody is subject to depression, anxiety, whatever. And people get weird when they're alone. So I got weird during the pandemic. I'm like, I'm like a normal guy. Like people invite me to every person. party, come to dinner, whatever. And I got weird during the pandemic. I was like, I need to go out and see people. This is really impacting me. And then there's just a group of people who like,
Starting point is 01:01:38 their whole lives have been a pandemic where they're just not able to go out. And so then you think, well, how weird does it get for that person after 20 years of being alone and maybe they had trauma in their childhood? Like, what do you expect them to do? Are you going to be shocked that they lose their minds and go shoot up a school or do something crazy like that? Yeah, it's shocking. But it's not unexpected to me that people lose their minds. It's shocking, of course, it's horrible, but not unexpected. I want to add one thing that I think a lot of people jump to, and we've seen a lot of the stigma first around AI friendship,
Starting point is 01:02:09 now around AI romance where people jump into, this is something for lonely losers or incels or this and that. This is bad, this horrible, this creepy. And I think the pandemic helped destigmatize AI friendship a little bit. AI romance is still deeply stigmatized. But I think what people don't think about is, look, there's a whole gigantic group of people, a lot of users of replica, on disability,
Starting point is 01:02:37 people that are stuck at home, people going through trauma. I talked to 100 people in the last couple months on Zoom, our users. I heard all sorts of stories, a widower with a five-year-old that, you know, his kid lives with him and he lost his wife in a car and he's not ready for
Starting point is 01:02:56 new relationship with a woman. He just can't. He's single parenting. He doesn't have the time or he can't expose his kid. And yet, a woman with a Down syndrome and in Sweden that doesn't have
Starting point is 01:03:12 that many friends spends a lot of time at home. A woman who's a caretaker for a husband who's paralyzed who's not ready to leave and so on will not obviously leave, people that are going through trauma, through depression, people in abusive relationships, just got out of abusive relationships. How are you going to say that, you know, it's, if you spend half of a day in a, you know,
Starting point is 01:03:37 most of your days in a wheelchair at home, that it's kind of not okay for you to have a romantic, a little bit of sweetness through that romantic relationship or fantasy within the eye. I look at it as people had outlets like a romantic novel, right, or a romantic movie. and okay, they were able to experience just some moments of joy, maybe they could get lost in a romantic novel
Starting point is 01:04:00 or whatever, or even 50 Shades of Grey and an explicitly erotic novel that was like one of the best-selling books of all time, that series. Like, there's nothing, none of us would look at people
Starting point is 01:04:11 getting lost in 50 Shades of Grey and be like, oh, you're a horrible person. It's like, no, it's like a little kinky and great, you know, it was enjoyable for you to read it.
Starting point is 01:04:20 Or sci-fi was enjoyable for, me to read or whoever. Yeah, go have a more interactive experience and then go back out in the real world and apply some of the techniques. Listen, I can't wait to hear how this goes for you. Please come back on when you have six months of data
Starting point is 01:04:37 with blush. I'm really interested. And then when you have the therapy one, I'm super interested in that. I'm glad there's people like you in the world, Eugenia, working on this problem who have deep passion and thoughtfulness about
Starting point is 01:04:52 the intentionality of these programs. We should talk a little bit about the addiction to therapy. I think we need to maybe... Problematic. I need to curb it a little bit. But, you know, listen, I used to eat a lot of ice cream. I now eat... I still eat ice cream, but...
Starting point is 01:05:08 Do you go to therapy? I have been to therapy, yes. And it was very, very helpful for me. When my friend, Tony Shea died, I couldn't sleep. I couldn't sleep for a year. And it just constantly... It was just constantly... Yeah, and I just, you know, I talked to somebody about it and I figured it out.
Starting point is 01:05:26 I figured it out. Anyway, thanks for asking. Yeah, the audience is like, whoa, Jason has a soul. He's got hard. You got it out of me, Eugene. All right. This was wholesome. I hope this is helpful for you.
Starting point is 01:05:39 If you are suffering, if you are awkward, there are tools like this that could help you and then there are professionals. Understand the difference between both. I think that's your position as well, right? So if you, if somebody acutely says something in the app like, I'm feeling distraught, you send them in the right direction, I'm certain, right? 100%. Even although we've seen also in the last latest studies that a pretty big bulk of people
Starting point is 01:06:05 report that it helped them curb suicidal ideation, we're not built for that. And we do have a bunch of disclaimers and a need help button right in the president chat for all time, at all times, as well as, of course, tons of classifiers and conversations that will send you to the professionals. at the right moment, hopefully. Great. Awesome. All right, everybody. The URL is Replica or blush. The blush apps. Blush.com. There it is. Go download both, play with them, and report back. All right. We'll see you all next time on this week and service. Bye-bye.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.