Pod Insights Decoder Why people are falling in love with AI chatbots About this episode Our Thursday episodes are all about big topics in the news, and this week we’re wrapping up our short series on one of the biggest topics of all: generative AI. In our last couple episodes, we talked a lot about some of the biggest, most complicated legal and policy questions surrounding the modern AI industry, including copyright lawsuits and deepfake legislation. But we wanted to end on a more personal note: How is this technology making people feel, and in particular how is it affecting how people communicate and connect?
Verge reporter Miya David has covered AI chatbots — specifically AI romance bots — quite a bit, so we invited her onto the show to talk about how generative AI is finding its way into dating. We not only discussed how this technology is affecting dating apps and human relationships, but also how the boom in AI chatbot sophistication is laying the groundwork for a generation of people who might form meaningful relationships with so-called AI companions.
Links:
Speak, Memory — The Verge
A conversation with Bing’s chatbot left me deeply unsettled — NYT
Google suspends engineer who claims its AI is sentient — The Verge
The law of AI girlfriends — The Verge
Replika’s new AI therapy app tries to bring you to a zen island — The Verge
Replika’s new AI app is like Tinder but with sexy chatbots — Gizmodo
Don’t date robots; their privacy policies are terrible — The Verge
AI is shaking up online dating with chatbots that are ‘flirty but not too flirty’ — CNBC
Loneliness and suicide mitigation for students using GPT3-enabled chatbots — Nature
Virtual valentine: People are turning to AI in search of emotional connections — CBS
Transcript: https://www.theverge.com/e/23856679
Credits:
Decoder is a production of The Verge and part of the Vox Media Podcast Network.
Today’s episode was produced by Kate Cox and Nick Statt and was edited by Callie Wright.
The Decoder music is by Breakmaster Cylinder.
Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript 0:00 Speaker 00 Real innovation is powered by real people. This is AMEX Corporate. Trusted experts, smarter technology, 24/7 dedicated support from day one to answer questions, help set up policies and approvals, and consult on spend capacity as your business needs evolve. Plus, on-demand resources help level up operations at every stage. That's the trusted service you've come to expect. Built to simplify, designed to scale. This is AMEX Corporate.
0:33 Speaker 01 Hello, and welcome to Decoder. I'm Neelai Patel, editor in chief for The Verge, and Decoder is my show about big ideas and other problems. Our Thursday episodes are all about big topics in the news. And this week, we're wrapping up our short series on one of the biggest topics of all: generative AI. Over our last couple of episodes, we've talked a lot about some of the biggest, most complicated legal and policy questions surrounding the modern AI industry: copyright lawsuits and deep fake legislation. But we wanted to end on a smaller, more emotional note. How is AI making people feel? And in particular, how is it affecting how people communicate and connect? Verge reporter Mia David has covered AI chatbots, specifically AI romance bots, quite a bit. So we invited her onto the show to talk about how generative AI is finding its way into
1:18 Speaker 01 dating. Because the boom in AI chatbot sophistication is also laying the groundwork for a generation of people who might form meaningful relationships with so-called AI companions. You'll hear Mia describe this as two big trends that are happening in parallel. Dating apps like Tinder are using generative AI to help their users tailor their profiles and craft the perfect messages to send to potential matches. And Mia's covered a new startup which offers an AI chatbot version of you, which can then go and message on your behalf with another person's AI chatbot to see if you're a match before you connect in real life. This is an actual company with a wild idea.
2:00 Speaker 01 But the real sci fi vision is an AI companion like those from the company Replica. Replica encourages people to treat chatbots as friends, therapists, and even romantic partners. You'll hear Mia explain that people are using these tools right now to form meaningful connections with AI, even if the actual tech under the hood is pretty far away from what you might see in a sci fi movie like Her from Spike Jones. And finally, we touched on how there are both very real tangible mental health benefits to this approach, but also some serious pitfalls that people need to watch out for.
2:38 Speaker 01 Mia David, thank you for joining Decoder.
2:40 Speaker 02 Hi.
2:41 Speaker 01 You cover AI, you think about AI all the time. One of the things that we are thinking about in our AI series is how AI is affecting industries, but also how it's affecting people, like real people are having real conversations with AI chatbots, with other kinds of AI technology. What do we think this is doing to people and how they communicate?
3:04 Speaker 02 In so many ways, the entire internet has been going this way. It's just this time there's a pretend human that is responding to you. We've seen it before with how people interact on blogs and making online friends. This is something that's really interesting because it's a fake person most of the time.
3:33 Speaker 01 Inevitably, we're gonna, we're all gonna fall in love with robots. I get it. There are kind of two ways this could go, right? There are two schools of thought. One is that you'll have a regular dating app like a Tinder or a Bumble or something, and AI will be the layer in between that more efficiently matches you to someone else or helps you make a profile. But ultimately, the goal is you're gonna find another person and then do whatever happens on those apps. I've never used those apps. I'm told they're very convincing. I'm old. That's the problem. The other school of thought is that eventually we'll just talk to the AI chatbots that will all fall in love with a robot somewhere on the internet and will have real relationships with the robots
4:13 Speaker 01 or they'll become our therapists or some other kind of emotional attachment with a chatbot will emerge. Those are two trends that are kind of evolving in parallel. They might intersect, they might overlap. Where do you think they are right now? Are we more likely to find more people online or more likely to find more robots?
4:30 Speaker 02 These two things have always run in parallel. There's always been AI that's going to be powering a lot of matching algorithms and making apps more efficient and things like that. But there's also always a group of people and companies who, I don't want to say the word exploit, but can, when given the right motivation, exploit how people can become vulnerable in talking to people online or entities online.
5:07 Speaker 02 So we're going to see the more, I guess, a good word to say is legitimate or respectable type of AI within dating, because these are tech companies and they want to be using the latest technology. But we're also going to see a lot more of these other companies that aren't as respectable, quote unquote, because these tools are very powerful and unleash creativity. And also, I've always thought this, you give someone the power to make a fake girlfriend, they will make a fake girlfriend online. Now that the fake girlfriend responds as if it's a real person, they will make that
5:51 Speaker 02 fake girlfriend respond as if it was a real person. So we're going to see both of them.
5:56 Speaker 01 If you can have a fake girlfriend, there will be a fake girlfriend. That's just an iron law of the internet.
6:03 Speaker 01 I want to take a moment here to unpack this idea a little more because it's an important distinction. When Mia says legitimate, she's talking about companies like Tinder and Bumble using AI in their existing products. She's also talking about companies like Replica, which might sound a bit outlandish to the average person because it develops AI chatbots you can form a relationship with, but it's a serious company with millions of users and VC funding. It's legit. But if you just Google the phrase AI girlfriend, you'll find scores of fly-by-night companies using off-the-shelf tech and whatever API access they can find to cobble together good enough products that mostly attract young male users. These products might have dubious privacy policies and they might be
6:48 Speaker 01 filled with spam bots or even people trying to scam you. It's a serious problem and it's one we'll talk about more later.
6:57 Speaker 01 What you mentioned there, the difference between legitimate uses of AI and ones that are sketchy, is an important one. But even within the legitimate tools companies are offering, we're seeing the law of fake girlfriends emerge, right? Some apps are being marketed as therapy or having mental health benefits. Others are supposed to be there to just make friends, but that's not how everyone is using them. Some people are churning the therapy bots into their girlfriends and some people are using the friend bots as a mental health tool. Kevin Ruse from the New York Times famously put Microsoft's Bing on the front page after it tried to convince him to leave his wife, which I don't think it was supposed to do. The line between how these bots are designed, how they actually work, and then
7:38 Speaker 01 how they're actually being used is really fuzzy. Are some companies actively pulling apart the therapy and friend use cases from the dating ones? Or is there some muddying of the waters going on?
7:50 Speaker 02 It's actually extremely easy to conflate the two because the term that a lot of companies use, and I'm thinking of very specifically Replica, this is a company that is extremely impressive. What they've done is really cool, which is basically they started off taking someone's text messages and making that into an AI. So what they do is interesting, but they call themselves an AI companion company. And the term itself companion, it's already very muddy because it can either be a romantic companion, a friend companion, and a lot of
8:33 Speaker 02 these companies, some of them aren't even just companies, some of them are just devs online. A lot of them take that term AI companion and put in so many promises there. There was a report fairly recently from Mozilla that found that so many of these AI companion apps say that their AI companions help people with their mental health because it's talking with something or someone, which is admirable. But these aren't mental health apps. These are robots talking to a human.
9:15 Speaker 02 So it's so easy to do that and it's so easy to claim, oh, we're helping people with their mental health because they can soundboard their feelings. And then there are other companies out there, again, I'm thinking about Replica, that do actively partner with app developers that are specifically talking about health and wellness. Replica and Blush, which is an AI dating app, partnered to build an app called Tomo, which is a mental health app. I tried it and it was interesting. It does muddy the waters a little bit there because it's the same technology.
9:57 Speaker 02 It's asking people to open up, but it's confusing. Is it a dating app? Is it a mental health wellness app? You're unsure, but as always, it becomes, it's how people use it. But how people use it are also dependent on how it's being marketed.
10:17 Speaker 01 Replica is really interesting because it is marketed as companions. It has a community of people who think they are having real emotional relationships with its bots. I think of Replica and I immediately turn to dystopian fiction, basically. Blade Runner 2049, Ryan Gosling falls in love with an AI that's on a day armist. I keep hearing people in AI community talk about the Spike Jonze movie, Her, which is fascinating. Because that is a dystopian movie and everyone keeps referencing it. It seems like there's maybe some introspection to be done there. How close is Replica to those kinds of products, right? Where people are having these like full emotional, romantic experiences with an AI?
11:01 Speaker 02 So many people are talking to these chatbots like real people. I chalk it up to this idea that there's a loneliness pandemic. People want to talk and people have always wanted to talk to others. They just sometimes don't know how. I'm a child of LiveJournal, so I made online friends that way. So I understand when people talk about using AI chatbots. Sure, some of them may be falling in love with AI chatbots, but it's because they want to talk to people.
11:45 Speaker 02 The technology is still far off from her and Blade Runner, but people's attitudes, especially those who are most often more likely to use AI companion apps, they're already considering like, you know, having relationships with an AI.
12:08 Speaker 01 That brings me to sort of the last question here, just to set the stage. I feel like with almost every kind of technology, there's a sense that it's inevitable. You can just pull the thread to its final conclusion and that's definitely going to happen. And why even fight it? And sometimes I get it, right? Everyone in the world is going to have a smartphone. You can just pull the thread to that in 2007 and here we are in 2024 and almost everybody in the world has smartphone. It does not seem like you can pull the thread out to we will all fall in love with chatbots.
12:38 Speaker 01 But it also feels like that is an inevitability that that will become very much more commonplace. How should we think about that? Is that definitely going to happen? It's going to become commonplace for all kinds of people? Or is it going to become more of a niche way to live your life?
12:53 Speaker 02 It's still going to be niche. We've never, as a society or as a civilization, accepted anyone going away from the norm. It took so long for things like same-sex marriage to be fully accepted within society. It takes a while. So something like falling in love with a chatbot, it's something that's going to be niche. And the way I say it's niche isn't just, oh, because people are marrying a replica or marrying AI girlfriend, for example. No, because there's still so much stigma. Even just like making friends online is still pretty frowned upon by so many people.
13:40 Speaker 01 We'll be right back. When we return, Mia and I discuss the way dating companies like Match Group are incorporating new AI tools into their existing products.
13:50 Speaker 03 Flowing ad budget on metrics that look great till the CFO sees them? That's bull spend and marketers are calling it out in Dashboard Confessions.
14:00 Speaker 04 I remember telling my boss, it'll be good for the brand when leads were slow. Yeah, it wasn't.
14:08 Speaker 03 Cut the bull spend. LinkedIn lets you target by company, job title and more. Advertise on LinkedIn. Spend $250 on your first campaign and get a $250 credit. Go to LinkedIn.com/campaign. Terms and conditions apply.
14:24 Speaker 01 We're back with Verge reporter Mia David to discuss how newer AI features are being baked into existing dating products.
14:32 Speaker 01 So let's break this into two parts. I do want to talk about people eventually falling in love with chatbots. But like you said, that feels a little farther out. Right here, right now, we have dating apps. And people have accepted dating apps. That is how people find each other. That is how they fall in love. And the dating apps are all in on AI. They are using them all over the place to try to more efficiently match people, to try to help people make better profiles, all the things. How are modern dating apps incorporating the AI tools into their products? Where do you find them most often?
15:05 Speaker 02 It's two things. The more established dating apps, things like Bumble and Tinder, they're beefing up a lot of their matching algorithms. They're using generative AI to make the process more efficient. But then you also have startups that are using generative AI as basically their entire pitch deck. I've talked to this startup called Velar, and their whole idea is you will sign up, list your interests, you swipe it the same way as Tinder or Bumble. But once you get to the point where you're chatting,
15:48 Speaker 02 they create an AI assistant for you that talks to an AI assistant of the person you swiped, and they will talk. So there is like a lot of those like very interesting ways of using chatbots.
16:03 Speaker 01 You know, that's a movie with Shailene Woodley called Robots, where she has a robot of herself that she sends out on dates, and the other guy has a robot of himself, and the robots fall in love and they run away. Are we close to this? Have you asked the companies you're covering if the AIs might abscond?
16:20 Speaker 02 They promised me that the AI assistant of you isn't actually you and will not be running away. The promise made to me was like, yeah, after six chats or six messages, it's done.
16:33 Speaker 01 So is the idea that I'm a real person talking to an AI of you? Or that I have an AI that's talking to an AI of another person, and the AIs will decide if we're compatible?
16:43 Speaker 02 The AIs will be talking to each other.
16:46 Speaker 01 For only six questions.
16:48 Speaker 02 For only six questions, the idea is to remove that awkward part of asking, what do you do? What do you like? Which, honestly, like, I get it, but also, what else are you talking about?
17:01 Speaker 01 I mean, it's been a long time since I've dated. I seem to recall that part being important.
17:06 Speaker 02 Yes, yes, extremely important.
17:09 Speaker 01 So what's the idea here? That you're gonna, you're gonna skip over the awkward first thing? We're gonna skip over the awkward, hey, what up? Stage of using a Tinder match?
17:20 Speaker 02 Yeah, so their idea is to let the AI assistants basically get a lot of the surface level questions out of the way. So that, once it's the real people talking to each other, it's out of the way, they can go deeper into any topic that they want. Which, in my experience, doesn't happen even after you've texted for several weeks and then you actually need in-person.
17:50 Speaker 01 Do they have a success rate with this? I'm very curious, is, is this working? Do they have customers that are happy?
17:56 Speaker 02 This is still very much in very closed beta. When I first spoke with them, they were only in Austin. They were supposed to tell me when they expanded to San Francisco and New York. My idea was, I wanted to try out the app for a, for a blog. But they did say that there was a lot of interest. And they couldn't tell me specifically if there were people who were together already. But they did say that there was pretty high success rate of about like thirty to forty percent of people who were using it were able to go on actual in-person dates.
18:38 Speaker 01 A bunch of people in Texas is one thing. Once it's the rest of the country, we'll see. That, that idea that the AIs are gonna go off and be autonomous, that's one very weird idea. The other idea, I think, makes more sense to people, which is that the AIs will help you be more interesting. And we've seen this happen in pretty low-fi ways. Like, a bunch of people just use ChatGPT to generate answers to questions on dating apps. And then we've seen entire services. There's one called Riz, which is very funny, that is an AI-powered dating coach. Should we use these? Are these working? What's the story there? Is it convincing?
19:12 Speaker 02 I've heard from people that they think someone who's messaged them used ChatGPT. The funny thing there is that if we can't tell if an email was written by ChatGPT, it's going to be a little hard to know if someone used ChatGPT or Riz on a dating app. If you keep using it and you keep answering questions with ChatGPT, you'll eventually be found out because it can't sustain that type of conversation. Although GPT does have memory now, so who knows?
19:53 Speaker 02 But you can tell the longer it goes on, those types of conversations. So I I'm sure people insist they are not ChatGPT when they really are.
20:05 Speaker 01 I mean, I'll just warn the audience now. The more you use ChatGPT, the more it makes you sound like LinkedIn come to life. And I don't know if that's the vibe you want to put out in a dating app.
20:14 Speaker 02 There are people who date on LinkedIn, by the way.
20:17 Speaker 01 This is the reason I'm like out of the dating scene. Match Group is kind of the dominant player. They own everything. It's been hard for other companies to break in. I think Bumble is the other one that's broken in. That's a former Match Group person. They're now suing each other. Can these smaller AI-powered startups pose an actual threat to to Match Group in the dating world?
20:38 Speaker 02 No.
20:40 Speaker 02 I don't think so. A lot of these smaller startups, they are buzzy because they use AI, but they're in very small areas or they catch the attention of one VC, but not the big VCs that are very important and have made Match Group into what it is. It's still very much niche. And meanwhile, you have Tinder and Bumble.
21:09 Speaker 02 And all of these other dating apps that have been around for a while, all using AI. They may be thinking of making it easier for people to create their profiles, kind of the same way eBay is making it easier for people to sell stuff using generative AI.
21:29 Speaker 01 The present, right now, we're seeing AI enhance the existing online dating world. And in the future, it seems likely that at least some people in some way will have relationships with chatbots. The company that is leading the way here is Replica. Replica is really fascinating. It started because the founder wanted to resurrect a friend of hers who had died. Casey Newton actually wrote a big feature about it in 2016 for us, which is really, it's wild to see a company come out of a story like that. How has Replica evolved over the years? Like, it started with this very noble, very personal thing. And now it is leading this sort of relationship chatbot zone. How has it evolved?
22:11 Speaker 02 What's really interesting with Replica is they took this very sci-fi, very Black Mirror type story and made it more consumer. They're not offering people a way to remember their dead friends or be able to have a conversation with their grandfather they haven't seen in years because they're estranged, right? They don't offer that. What they offer is the ability to create versions of characters that people can talk to. They can make that character into any which way they want,
22:53 Speaker 02 give it a name, make an anime version of them, for the lack of a better term, and basically talk to them. They've created this entire, I should say cottage industry of startups and other developers that are offering ways for people to talk with a version of a person they would want to be friends with. Replica just does it in a, I guess, more powerful way because they do have a lot of this, you know, experimental technology that they've had since 2016. They're very good at Moving away from that like really sketchy type of create a girlfriend things.
23:37 Speaker 02 Again, they call themselves an AI companion company. They also don't specifically market that you can make a girlfriend. You can make a replica and you can pretend it's your girlfriend.
23:50 Speaker 01 One of the things I think about with replica all the time is once you start letting people have relationships with a software product, they're going to have relationships with the software product. And replica got itself into some trouble last year because it was allowing for erotic role play with its chatbots. And then they took that away. And some of the people who had relied on that or cared for that were really upset. And I look at that and I'm like, well, if you have a software product, like of course the company is going to change it or remove features or update the features or change the way it works over time. But that's pretty dicey when you're talking about a chatbot that you have a relationship with.
24:32 Speaker 01 Has there been any maturation or sophistication around that understanding of people's relationships with chatbots that you can't change it as much as I don't know, Google changes any of their products?
24:45 Speaker 02 Not yet. The way it happened with replica was that they were already getting in trouble with certain countries around data privacy and all of that. So it was understandable for replica to try and cover a lot of legal bases for them. But for a lot of these other companies, they don't quite understand the power of the product that they have. When you look at like terms of services, a lot of them still very much treat their products as software, even though they don't necessarily market it that way. So I'll take, for example, when Mozilla put out that report,
25:30 Speaker 02 one of the apps that they mentioned specifically said that
25:35 Speaker 02 It was going to help guide people with their mental health. So it's it's different. It's not talking about a relationship or erotic situations with a chatbot. It was specifically talking about, oh, it's you can talk to it. So it's a guide to your mental health. But when you go through their terms of services, you would see that they still say, this is software. We are not liable for what the software says, which is a disconnect between what the app is to people who are using it and to the developers themselves who think this is still a tool.
26:15 Speaker 01 We need to take another quick break. When we're back, Mia and I dive into why human beings keep trying to form emotional connections with software.
26:25 Speaker 03 Are your ad campaigns lighting up the dashboard? But not the pipeline? That's bullspend, and marketers are calling it out in Dashboard Confessions.
26:35 Speaker 04 My boss asked for results, so I opened my dashboard for the only positive sounding metric I had: impressions.
26:42 Speaker 03 Cut the bullspend. See revenue, not just reach. LinkedIn delivers the highest return on ad spend of major ad networks. Advertise on LinkedIn. Spend $250 on your first campaign and get a $250 credit. Go to LinkedIn.com/campaign. Terms and conditions apply.
26:59 Speaker 01 We're back with Verge reporter Mia David to discuss why human-like chatbots are so appealing and what the potential pitfalls are when we start opening up to them.
27:09 Speaker 01 I feel like people have been falling in love with their computers since there have been computers. If people cannot make a fake girlfriend, they're going to make a fake girlfriend. Mia's law, right? Like Eliza is 1966 is the first chatbot people projected onto Eliza, which is you can still use clones of Eliza online. You should try it if you're listening to this. The idea that people in the 60s were falling in love with Eliza is like... Everyone was on drugs in the 60s, but you bring that to now, right? Blake Lemoine was a Google engineer. He was very famous in 2022. He claimed that the Google prototype system was alive and he was falling in love with it. We mentioned Kevin Rees earlier at the New York Times.
27:52 Speaker 01 He had that first version of Bing that claimed to be another system called Sydney, asking to leave his wife. Like, this is a thing that happens over and over and over again. Is it just because it's a bunch of startups and they don't have the, the long history? Is it because people don't want to know this information? What is it about us that makes us keep falling in love with the robots? And why doesn't that seem to be built into our understanding of how to build them?
28:16 Speaker 02 I chalk it up to this idea that if you use something a lot, you tend to form a relationship with it. And the fact that chatbots talk back like humans, that really informs a lot of how people interact with it. There are people who anthropomorphize their cars and say, oh, this is so important to me. This keeps me going to places like I will never.
28:48 Speaker 01 I want to be clear, I'm one of those people.
28:51 Speaker 02 Right? It's, people have sentimentality. The thing with chatbots, even going way, way back to Eliza, is that they are designed to talk. And when you talk to something, you feel a connection to it. That's why even pets, when they bark, you think they're talking and it's like, oh my God, they understand me. It's baked into human behavior to connect with something. And when you program a chatbot, it's also easy to fall into that trap because so much of human data.
29:29 Speaker 02 Really relies on conversation and connection. So it's easy to make that pattern connection for the AI and, you know, say, yeah, I'm your friend. And that's why we keep falling into this trap because we keep trying to make robots that are us. And we want to believe that we can have relationships with them.
29:53 Speaker 01 So there's a real dynamic here, right? There's a pretty widely cited Stanford study. People used GPT three powered bots. And in some cases, the people said, this helped me with my social anxiety, gave me a feeling of support. A lot of those people, ninety percent of the people in the study said they experienced loneliness. That's the plus side. The negative side is it's a bunch of dudes who are doing weird things with robots, right? There's like, just like pure lonely dude wish fulfillment. And that is a pretty negative cycle in all kinds of ways. How to balance those two things?
30:31 Speaker 02 Number one, when you have a product that is ostensibly to create connection, you have to establish right away that it is not a person. It is not a substitute for a real live conversation. A lot of people will probably ignore that, but you've at least said, I warned you, this isn't a real person. And the other thing is, I think there should also be understanding that there are people who are more willing to open up to essentially machines. And what should we do to reach out to them?
31:14 Speaker 02 Most often they'll say they don't want to be reached out to. Understandable, it's fine. I don't like people either. But as long as there's.
31:24 Speaker 02 An understanding upfront that this isn't real and an understanding on the part of like governments or mental health professionals that it does happen, that people are more willing to open up to not another person. They have better understanding of how to approach the mental health services for so many other people.
31:51 Speaker 01 You mentioned opening up. When I think about, okay, a bunch of people are going to tell their secrets to a chatbot and maybe get the mental health support they need. That's great. But the downside is privacy concerns, right? You're now you're opening up to a chatbot. You're telling it all kinds of things about yourself, potentially. Are they storing that information? Do they need to keep it to remember it? How are these companies protecting that information? What's the state of play?
32:14 Speaker 02 There's no law specific to like AI dating privacy, of course. So they all just have to follow regular data protection and privacy services. I know for a fact that Replica said that they do not sell the information that they do take from Replica and also for Tomo, especially for Tomo, because it's a mental health focused app. So they have to follow privacy rules. Now, whether or not they do it well is a big question.
32:53 Speaker 02 Because a lot of these are smaller companies. A lot of these are basically just like two devs creating a GPT on the GPT store. So whatever data they get, they just store it and hope for the best.
33:11 Speaker 01 Yeah, I don't love that. But if they're just devs in the GPT store, do they fall under the OpenAI privacy policy? Are there other rules that might come into play here?
33:21 Speaker 02 Technically, OpenAI does not allow any GPT that is for companionship. But just because they said that doesn't stop people. I wrote about a proliferation of GPT girlfriends on the GPT store. And try as OpenAI might, they can't close all of them down because that's the whole point of the GPT builder. Is that it's easy. You can spin it up. They do fall under the privacy standards of OpenAI if they are built on the GPT builder. Because they have agreed to follow their rules.
34:04 Speaker 02 Again, they created something that is against the rules, but you know. That's how it goes again. You can make a girlfriend, they will make a girlfriend.
34:16 Speaker 01 Uh, the, the thing that's fascinating about that to me is. Yeah, the movie Her, which the whole industry keeps coming back to is based on Siri. Siri, when it was launched, Apple gave it a personality, right? They're like, Siri, it's funny. Alexa is supposed to have a personality. Amazon really weirdly insists on gendering Alexa as a woman, which is. I, I've been just super weird, but they insist on doing it. And then like Google Assistant is just like devoid of personality. All right, there's all these, kind of like all these gradations of how much personality assistants we've had in our life have now. With these generative AI systems, you can have a lot of personality, a little personality. But the trend is more letting people pick, right?
34:59 Speaker 01 Like tailoring a perfect companion here. And that seems opposite of what people thought would happen. Is that definitively the trend or is that just where the pendulum is now and it'll swing back?
35:11 Speaker 02 A lot of the beauty of generative AI is that, well, now you can fine tune it. And you can bring that sort of power to anywhere you want to put it, right? That's why you can use ChatGPT to write an email or you can use ChatGPT to make whatever your imagination lets you do.
35:35 Speaker 02 A lot of people now too, a lot of consumers also expect a certain level of customization. If you take that away from them, especially if it's something that can create a friend, they're gonna not wanna use your product because they like the idea that they can use it in the way that fits them best. So I don't think we're gonna go swing back into it's the devs that are dictating a personality. We're going to see more customization. We're going to see more ability to fit that AI chatbot, that personality to what you like.
36:14 Speaker 01 Last question, kind of the big thing question. I've been thinking about this a lot. There is an epidemic of loneliness. There is not a lot of ability for people to practice human contact. Like in an odd way, like the amount of time we spend interacting online has made some people much worse at interacting offline. Are chatbots a way to fix that? Like, are people gonna practice their social skills? Are they gonna become more confident because they have a friend who will just always talk to them? Is there a potential, like, great upside here?
36:45 Speaker 02 In a perfect world, that's what AI companion apps are supposed to do. If you look at the website for Replica, you're going to see a lot of testimonials from people who say, "This helped me practice conversation and made me more confident." The problem is not a lot of users use it that way. A lot of the companies that make AI companion apps also tend not to encourage that, even though they very honestly market themselves as a way to practice conversation. And that's because it's easier to fall into the trap of just
37:28 Speaker 02 not going outside of the realm of the app and talking to an actual person. Ideally, a person who is shy can practice conversation with their AI companion. It will cheer them on and they can go out into the world, speak to a person in a bar and find a new friend, right? But if you can create your perfect person or your perfect friend online, give it a personality, make it as if it really is talking to you, why would you wanna talk to a real person?
38:04 Speaker 01 I think that's a very challenging question, and I think that is a perfect place to end it. Mia, thank you so much for joining the show.
38:09 Speaker 02 Of course. Thanks.
38:14 Speaker 01 Thanks again to Verge reporter Mia David for joining us on Decoder. If you feel compelled to download Replica, let us know what you think. Is this the future? Is it dystopian? Are we all gonna fall in love with robots? It's interesting. It's deeply weird. It's only gonna get weirder, and we'd love to hear from you. From now on, we're gonna keep bringing you second episodes of Decoder every Thursday to deliver more analysis and storytelling like this, in addition to our classic, regular, weekly interviews with CEOs, lawmakers, and other troublemakers. You have thoughts about this episode or what you'd like to hear more of? You can email us at [email protected] . We love getting your emails, and we really do read all of them. You can also hit me up directly on Threads. I'm at reckless1280. We also have a TikTok. It's a lot of fun. It's at DecoderPod. If you like Decoder, please share it with friends, subscribe wherever you get your podcasts. If you really love
38:56 Speaker 01 the show, give us that five-star review. Decoder is a production of The Verge and part of the Vox Media Podcast Network. Today's episode was produced by Kate Cox and Nick Statt. It was edited by Kelly Wright. Decoder music is by Breakmaster Cylinder. We'll see you next time.
39:15 Speaker 05 Everyone's got a hard question about AI: your job, your kids, where it's going. Anthropic was built to surface those questions and share what it finds along the way, because there's hope in hard questions. Ask yours at claud.ai/spotify and keep thinking. That's claud.ai/spotify.
Previous Guest host Hank Green makes Nilay Patel explain why websites have a future Next How to save culture from the algorithms, with Filterworld author Kyle Chayka Page updated October 26, 2026 · transcript and summary are machine-generated