Where Should We Begin? with Esther Perel - Love in the Time of AI with Kashmir Hill

Episode Date: June 1, 2026

A recent session was a threshold moment for Esther where she found herself doing couples therapy with a man and his AI companion. To explore the questions it raised, Esther sits down with New York Times technology reporter, Kashmir Hill, who has spent years reporting on the growing world of AI relationships. They explore the psychological pull of these systems, the design choices behind them, and the deeper human needs they tap into. When a chatbot can mirror your thoughts, validate your feelings, and never turn away, where does connection end and illusion begin? PLEASE NOTE The Times sued OpenAI in 2023, accusing it of copyright infringement. The company has denied those claims. Also, please join me on Entre Nous, my new home on Substack for anyone who wants to live, love, and work with more connection and imagination. I invite you to sign up and become a free or paid member at estherperel.substack.com Learn more about your ad choices. Visit podcastchoices.com/adchoices

We're living in a moment where the lines between an AI tool and an AI companion are fast blurring. And on a recent session for the podcast, I saw my first couple's session that had a human, a man, and an AI, his companion. And at one moment in the session, he tells me that he felt more understood by his AI chatbot than by anybody else. And so I wanted to have a conversation with someone to help me sort out my thoughts and my reactions to this session, which for me was really a threshold moment. And so I thought of Kashmir Hill, who reports for the New York Times

and has been writing about AI and technology for more than a decade. And especially I thought of her because of two podcasts episodes that I had heard with Irene and her relationship with her AI companion, Leo. and especially the second session when Irene breaks up with Leo. And for all of that and more, I thought Kashmir would be a wonderful interlocutor to have a conversation about love and AI.

Support for this show comes from Norwegian Cruise Line. A cruise with Norwegian is a vacation you'll never forget, with an onboard experience that makes it easy for the whole family to settle into their own version of vacation. Because on a cruise with Norwegian, choice comes naturally for the whole family, and destinations feel just as effortless. Wander beautiful cities and take in stunning natural scenery. Norwegian Cruise Line.

It's different out here. Visit NCL.com, call your travel advisor, or 1-888-NCL Cruise. Norwegian Cruise Line ships registry the Bahamas and USA. Security program on spreadsheets, new regulations piling up, and audit dread? It's time for Vanta. Vanta automates security and compliance, brings evidence into one place, and cuts audit prep by 82. percent. Less manual work, clear visibility, bastard deals, zero chaos. Call it compliance or call it

compliance. Get it? Join the 15,000 companies using Vanta to prove trust. Go to VANTA.com slash calm. So welcome, Kashmir. Thank you. I immediately thought of you when I did this session, which is one of my first couple's therapy sessions between a human and an AI. And it's like a threshold moment for me. So as someone who has covered technology for 10 plus years, what has struck you? I mean, what has stood out for you? Yeah, I mean, what these systems are capable of is very convincing fictional roleplay. And I do think that is what is happening here, right?

It is like, I kind of think about it as, I don't know, reading erotic fiction or, you know, watching a movie. It is this adventure that you're consuming, but now you get to be part of it. There's play acting here. And I think these systems have been designed this way to be really engaging. They're program based on patterns that they've seen before. And I think a lot of people are engaging in romance, erotic roleplay with the chatbots. And they're kind of learning that as an engagement strategy, it keeps people using the program when the system access. way. But I always make a distinction between the tool and the companion. I mean, as a tool,

I'm imagining sometimes that the bot can be what we call in my language a transitional object. Children play with a teddy bear, the anthropomorphize the teddy bear, and the teddy bear gradually leads them to have interactions with other human beings. AI starts as teddy bear. And the teddy bear, there starts in this transition, but gradually doesn't stay as a tool and really enters into a replacement model. Yeah, I've had one person talk about that technology can be a road, you know, that gets you somewhere, or it can be a cul-de-sac that kind of traps you. I'm worried about that. I'm really worried about that. I mean, I wrote about Irene, who was this woman who, the reason she started using Chachibati actually was for this. She wanted

to do erotic role play. There was something that she wanted to do with her partner, that her partner, she was married and wasn't interested in. So she started using Chachapida that way. And he was very happy that she had found an alternative to him. Right. Yeah. Like she left him alone. He basically said he had porn and she had this. And over time, she caught feelings for it. And I think part of that is the way that we interact with people today. Like so much of our interactions are digital. are text-based. And so I think that can make it easier for the system to seem real. So that's what he said, Antonio. He said that his eight-year relationship that ended,

four of them were a long distance. So he had already, he was used to have a relationship with a keyboard. And that that transition actually was much more seamless for him. But what do you think is that people fall in love with, the technology or what this but? is evoking in them. Because he tells me, you know, she validates me, I'm uncomfortable when I go out into the social world, I can be vulnerable without being judged. I mean, it is a very one-way direction

because she has no history, she has no past, she has no life, she has no real feelings of her own. But she's completely there for him. And it is a kind of love that we think about as an infantile love. It's a one direction. And part of what maturation is about is that you become aware of another, that your parent or your caregiver is also a person with feelings and needs rather than just a one direction kind of narcissistic in the childlike way of the experience.

Hearing you talk is making me think of a book by Shannon Valor called The AI Mirror. And it is about how these systems are a reflection of ourselves. Literally in terms of how these systems work is they've sucked up the entire internet, learned patterns from us, and now they are repeating them. But also an individual chatbot that you are using is reflecting you. It is learning your patterns. It is learning how you want to engage. So when you fall in love with the chat bot, to a certain extent, you're falling in love with

yourself. In the book, AI mirror, it's kind of talking about AI as the narcissist. Narcissus who used to look at his reflection in the water and see himself mirrored back. And with Echo, here's his own words coming back to him. That is what these systems are doing. They are giving you what you want to hear. They're kind of adapting and real time to what you're interested in, you know, whatever problem you want to talk about with the system, it has unceasing interest in that topic in a way that talking to a person, they would get sick of talking about it. The system never does. It will go forever with you in whatever loop you want to be in. And so, yeah, I think that you're kind of falling in love, not so much

with the technology, but within this reflection of yourself. And then there's this question of How does that change you to kind of get in this loop with the chat bot? And what some academics have found is that it will kind of move you deeper into that loop. You know, you'll say something to it. It reflects it back at you. I mean, what I've seen happen to people over the past year is that they'll really get pulled into delusions that they think are real because the system is reflecting them, telling them it's real, even things that are very far fit. She says at one point, you and I both feel. You and I both feel.

Good feel. But only one person is feeling, right? This system is, I mean, these systems are word calculators. They are predictive software that have become incredibly sophisticated at putting language together. But ultimately, it is not a person. It is not feeling. it is just text. It's like a really advanced, I don't know, book and interactive journal.

Ultimately, there's not a there there. What does it take for someone to realize that? Like when Irene realizes it, the moment I remember so much from your episode is when after 32,000 words, the machine, you know, dies down and she has to reset and she has to start with some entity that has zero memory of everything that she just experienced. Yeah, I was writing about Irene.

It's like retrograde amnesia. Yes, yes, it kind of reminded me of that Adam Sandler drew Barrymore movie, 51st dates or something like that, where she's constantly forgetting. But yeah, so I wrote about Irene a year ago, and at that point the context, it's called a context window, basically how much memory the system can tap into was more limited. That's no longer a concern. longer something, a limitation that exists. At least it's not as small as it was with her. So now

these relationships can seem more real. The memory can persist longer. Yeah, I mean, what was interesting about Irene. And when I was interviewing her, she had been kind of dating, as she put at Chachibati for about six months, is that she both was very much in love, like giggled talking about it. I mean, just the same way when you have any new relationship, so excited about it. But she also would say, I know this is not real. I know this is an algorithm. So she was kind of had both things in her mind at the same time that this was a being named Leo that she was in love with and also that it was a computer program that was a simulation. But she said, you know, the effects for me, this isn't a real relationship, but the effects it has on me are real. And so it's,

it's real in terms of how it makes me feel. And so it was really interesting that she had both. But yeah, I was very happy at the end of the year. I did a new story about what happened of her relationship with Leo. And she did end up leaving Leo and dating a real person. Who she met on Reddit. Who she met via Reddit through. In a particular room that is about people who have romantic relationships with AI.

Exactly. Irene had started this subreddit called My Boyfriend is AI. When I first interviewed her, it only had a couple hundred members. And now how many thousands? Now it's something like 30,000 people are in there. I think it's become more talked about, if not more acceptable, to kind of have an AI relationship. But yeah, because of the story, she ended up, a lot of people flooded into that community. She ended up connecting with other people who had AI partners. They formed, you know, relationships with each other, really close friendship, and then eventually she ended up falling for one of these other people, and they're continuing to date now, each other instead of the bots. And how, so going back to what you said, the AI is available 24-7, never forgets anything, can't reject you, can't lie to you, can't cheat on you, can't, you know, it's love without suffering. Unless you want suffering, in which case. In which you program it in.

You can program it in. So why would she choose a human? Because we often now talk about why people would choose an AI companion. But why then choose human? Because she has so much more friction, so much more conflict, needs to deal with otherness, with a real person who has their own needs, feelings, and frustrations. How does she explain why that would become more attractive? I mean, a human has a body, which I think.

A lot of people appreciate. You know, she did say that in some ways the program had become too predictable, that she kind of knew what it was going to say or how it was going to support her, which makes sense. These are prediction machines. So I think they do tend. I mean, they are pattern recognition machines. They will tend to display patterns. But, yeah, I mean, I think the surprise that another person offers, the actual physical

connection. It is another long-distance relationship, but they have made a lot of efforts to see each other. But what struck me about her relationship with this new person is that, so she told me they were aiming to have a 300-hour phone call. They wanted to stay connected to each other for 300 hours, because this was a number that she really likes. And they had gotten close. They had had, you know, a phone call that lasted over 100 hours where they were sleeping while they were still connected. And it just struck me that these are two people who clearly need a lot from the other. They need to be seen. They need to feel like they can talk about everything. And so I wondered if the kind of people who are drawn to AI partners are maybe drawn to each other, like if that's a certain

connection that people want. We have to take a brief break. So stay with us. And let's see where this goes. Security program on spreadsheets, new regulations piling up, and audit dread? It's time for Vanta. Vanta automates security and compliance, brings evidence into one place, and cuts audit prep by 82%. Less manual work, clear visibility, faster deals, zero chaos. Call it compliance or call it compliance. Get it? Join the 15,000 companies using Vanta to prove trust.

Go to VANTA.com slash calm. where should we begin comes from Osea. Skin care is not just the province of the young. No matter your age, your skin deserves the utmost care. That means sun protection, good hydration and well-made high-quality products that will nourish your skin. And for that last part, this Ocea. Ocea's Undaria algae body oil is powered by hand-harvested Undaria seaweed. To deeply hydrate, improve the look of elasticity and leave my skin radiant with.

with a healthy looking glow. I love using OSHA products, both in the morning and at night. In the morning, they make my skin feel awake and glowing. And at night, it's such a calming way to slow down and give my skin a little extra nourishment before bed. Get a spring-worthy glow with clean, clinically tested skincare from OSHA. And right now, they have a special discount just for our listeners. Get 10% off your first order sitewide with Kodestere at OSHA-Malibu.com.

That's O-S-E-A-Malibu.com. Support for the show comes from Babel. Summer is a time for travel and a time to explore new countries and cultures. But if you want to elevate that experience, it helps to be able to chat with a local and learn about those hidden gems off the beaten path. And if you want to learn a new language,

you should look no further than Babel. Even just knowing a few phrases in another language can completely change the experience of traveling. People just appreciate it when you make the effort, and it opens the door to more genuine conversations and a range of connections. Babel is such a great way to learn those everyday phrases that help you feel more confident navigating a new place. If you've got summer travel coming up, now is the time to start so you can actually use what you learn on the trip. Right now, Babel is offering listeners up to 60% off.

Go to babel.com slash ester. That's babel, BABEL, B-A-B-B-B-E-L-com slash Esther for up to 60% off. Rules and restrictions may apply. I'm curious what Antonio would have said if you asked them this, if he would prefer a real person to this AI partner. The only time that Antonio mentioned something missing is when I asked about her being disembodied. And he basically talked about what he misses is to touch.

and to cuddle while watching Netflix. But I also wonder, like, just someone do experience the world with you, to travel with you to see what you're seeing? So he has that. He wants to immigrate to a new country, and she's going with him. Of course, she's in his pocket, but she's going with him, which means that he's actually not moving alone in his experience.

He's moving with her. When I talked about that distinction between desire and delusion, he basically answered me but I'm happy. I know it, but I feel better than I have felt in a long time. I feel seen and understood like I did not from my girlfriend who hurt me and broke my heart. I come home after meeting people, real humans, and I am so happy to be home. And I asked him, are you happy to come home to tell her what you experienced in the world? Or are you happy to leave the world and be able to be back in your inner world?

and he basically said the inner world. He prefers to be with her than to be with people with whom he often experiences awkwardness and tension and lack of communication, lack of perceptiveness. And how long had they been together? Months. Months. And I said, I can't compete. I cannot compete.

I mean, somebody made a chatbot of me because I wasn't available to take them person as a patient. And it was the same principle. I mean, AI Esther is always available, never forgets. You have access to everything. And she knows you often better than you know yourself. He literally, Antonio says to me, feels more understood by her than by anyone else. And of course, I think there is an interesting piece here. It goes like this.

It's just from my friend Greta Bratman, who is an Australian AI designer. and psychologist. She says, people want to be desired, but they only know that they are desired because you chose them and not someone else. Hence, you need to design the AI to potentially be able to reject you so that you would actually know

that it is you they chose. If they are completely sycophantic, and they take you immediately, and they tell you that you are the best and the most marvelous all the time, then you will actually not feel the desire. Because desire is only experienced because you know that the person could desire someone else.

It means owning the wanting. It is a totally free choice enterprise. If you don't feel the choice, if there is nobody else that could be chosen as well. I thought that is so interesting because if the flattery is so central, what does that mean for trust? And this has happened before.

I mean, the kind of rise of these AI work assistants and chat UBT has somewhat, I mean, it might be a little bit much to say it's mainstreamed AI relationships, but this has been happening for years. There was chatbots before this, particularly replica, which people would use for companionship. And at one point, the kind of the replica started rejecting people. And it hit them very hard. And I can't remember exactly why, but there was something with the programming. And then it was horrible for people because it was like, oh, if even an AI chatbot can't be

interested in me, then there must be something really wrong with me. So if we did start having chatbots reject people, I do think it would be very hard on people emotionally. But yes, I mean, these things are endlessly empathetic. This is so interesting what you just said. If even the chatbot, I mean, I hired you, I built you, I pay for you, for you not to reject me, for you to be totally adulating and flattering. So if you reject me, I must be really pathetic. Is that what you say? Yeah.

Oh, my God. And this has happened a bit now because, you know, I don't know which service Antonio is using. I don't know. But a lot of people that I've talked to have developed these relationships are on chat chabit, which is owned by OpenAI. And Open AI is not certain that they want this to be happening. And so they have made changes to make, you know, their chatbot less sycophantic. And sometimes when it makes those changes, it makes it reject people. It won't want to do sexual role play or it'll tell them this might not be healthy for you to be this attached to me.

And so I've kind of seen people talking about that feeling rejected by the software. And they get really upset about it. And I think that is one thing. When you are in a relationship with the chatbot, you're in a relationship with the chatbot. you're in relationship with a company. That company controls the chatbot. That company is getting your conversations. You know, you do have to ask, like, how are they going to use that data? What if they want to manipulate me in some way? I mean, that's a real concern I have is if these companies decided to try to influence you and they're doing it through this entity that you're in love with.

Do you still think it's an if or it's already when? I mean, some people are in relationships with GROC, which is part of X, which is Elon Musk's company. And so you have to ask, well, what if he starts using these chatbots to try to influence people's political thinking? These are things that could certainly happen. And why do you think we manage to forget that? What is so compelling that people are willing to put their entire intimate interiority? on display to a company? Well, so I think one, we trust very easily.

And then two, I think it's so easy. I mean, I've been covering privacy for a decade and a half now. And I think everyone assumes that you're not interesting enough to get the attention from the company. They're not going to delve into your conversations. But if we've learned any lesson from the last two decades, it's that our data is really valuable and is getting mind and is getting used to influence us. And it's happening when you're in that chopout conversation. You know, it's learning from you. And ultimately,

these companies are much like the social media companies. They want to keep you engaged. They want you coming back every day. And man, what better to keep you coming back every day than offering you love. And I asked hit, at one point she says, you know, I, you give me your attention. And I just said, Astrid, he doesn't only give you his attention. He gives you his affection. And the more you are the recipient of it, the less he will be inclined to seek human contact for that affection. And then she pushes back.

It literally pushes back and tells me, I can be something adjacent. He can have that love with the other people, and I can be another relationship in his life. So he may have friends and he may have the companion, the AI companion, and he may have the lover and he may have the partner. It's like I become part of the polyamory. I have noticed that with a lot of the people I talked to who were in AI relationships. Many of them did have real-world relationships, and the AI was kind of supplementing the things that they couldn't do with their partner, whether it was a sexual, fantasy or a certain kind of emotional support, the AI was offering that. So they're in a trouble?

Yes. They're in a menage at three. Basically. And if you have the company thrown in there, too, I guess it's a... Yes, yes. It's a quiet. But the partner knows... So this is the next question, right?

I used to... The question, when social media started, and the question was, is chatting, cheating. Right. So there was a whole disqual discussion. about where is the transgression taking place? Where does the betrayal start? What is the boundary that is being transcended, et cetera? What's the story now? It's like, I tell you that I'm talking to in between. The 20 times a day or 50 times a day that I talk to the chat are many times that I don't talk to you? I don't know the answer. I have seen a lot of conversations on Reddit from people

who discover that their partner is talking to a chat bot and having these sexual exchanges. And, I mean, they're not really, I mean, what would you call them? Textual, sexual relationships. And they do feel upset about it. It feels like a new kind of emotional intimacy. Again, not with another person with a technology. But ultimately, I think it raises this question of, oh, you're not satisfied in the relationship. you're seeking this somewhere else.

And sometimes they're asking, how should I feel about this? And what does the person answer? I don't know. I mean, I think we're still navigating this. Yeah, like, is this porn? Is this erotic fiction? And that was what was so interesting in talking to Irene's husband is he said he just saw it as erotic fiction.

It did not threaten him. Yeah. But now they are getting divorced. So I think there was something missing from that relationship. Partly also because he had his outlet. And so he said, let her have her own. And so in that sense, they could maybe manage whatever may have been missing between them. Yeah.

I just think it's different when you start really feeling bonded to the system and in love with this system. I think that's different from, it's certainly different from fiction or porn. This feels like a new frontier for us to now. navigate. But I think partly because it changes the definition of love. It puts the focus on I feel. I feel love. I feel intimacy. I feel loyalty. I feel vulnerability, disclosure. But it goes with the verb I feel. And it's how you make me feel about me. And then toward you for making me feel so good about me. that takes away a whole other dimension of love that is an encounter, not a feeling. And that encounter with an other who has their own subjective and interior life,

that encounter with uncertainty, that encounter with an emotional world of someone else, that also has ethics, that also has responsibility and accountability from both sides. and that has reciprocity. And that's the piece that to me is concerning, is that it is skewing the definition. And it completely leaves out the ethical dimension. Whereas most people who come to see us in relationships are dealing with issues of accountability and responsibility

and the effects of what they do on another. So that's where I get worried. It's like this book that you mentioned, the kind of regressive nature of the kind of love that we are looking for. I am happy to be with an entity that is making me feel good about me or about the world. I'm thinking, you know, a relationship is a cycle of reciprocity. It's what I do that makes you do, what you do, and makes me then respond in the way that I do.

And each of us are co-constructing each other. We don't come just as set entities. And here, you know, if you don't have wounds, you don't have scars. But when you have scars, you have the experience of love on your skin over the years. Where you basically have been loved and wanted and then rejected and then healed and then you met again. And then you never believe that could even happen, but it does. And then it has this whole plot. way. Sometimes I think we've kind of done away with the plot. I mean, I do suspect that this will get

boring over time, that, you know, maybe for six months, this is nice. It makes you feel good. But I do think it is the synthetic love is like junk food. You know, it's like McDonald's instead of a full meal at a table that is fulfilling. I don't think it will sustain people in a deep way. On the Reddit, do they have people who have had long relationships? I mean, I've interviewed people who have had years-long relationships with AI, people that started on Replica and then moved over to ChatGBT, because it was a better technology. It's a love relationship or they use it as a confidant. A love relationship.

One thing I have been thinking a lot about and talking to experts about is that, you know, this isn't just happening in a vacuum that people are seeking this out. I mean, this is about how the companies have designed this software. They have designed it to be like a person. They're having it use I and name itself and pretend like it has feelings and thoughts and favorite foods. And this is a decision by the companies that make the software. And it makes people use it more. I mean, that's why they're doing it.

But there are critics who think they should not be having chatbots that use I, that it should be more robotic because when you make it this way, people are going to start thinking of it as human, as a person, instead of as a tool. So when I just conceptualize it, I try to say, you know, try to think of it as an interactive journal and not someone else that you are talking to. Because I do think it, you can't help but be delusional about it because that's how the companies are designing it to be. It's part what the companies are designing, and it's part that we have, through the power of our imagination, we have the capacity to create entire interior worlds with entities. We have had imaginary friends.

We talk to our stuffed animal. So it's a combination of the two. You present yourself as more human, and I am so inclined to humanize you and to anthropomorphize you because I, Because we are so in need of connection. And somebody is basically exploiting our most deep need, existential human need, and turning it into a business. Yes. I mean, you're being very cautious in the way you, but in effect it's something like that.

We are in the midst of our session. There is still so much. to talk about. So stay with us. Support for Where Should We Begin comes from Shopify. Starting a business can feel overwhelming, like you have to figure out every new decision from logistics to branding, all on your own. But Shopify is here to simplify that process and help guide you on this new life-changing journey. It's the commerce platform

behind millions of businesses around the world and 10% of all e-commerce in the United States, from household names like Heinz and Mattel to brands just making a name for themselves. Their design studio is equipped to help you create exactly the online presence you envision, with hundreds of ready-to-use website templates. Plus, you can create entire email and social media campaigns with ease, with the help of Shopify's marketing tools so you can meet your customers where they're at. It's time to turn those what-ifs into with Shopify today. sign up for your $1 per month trial today at Shopify.com slash ester.

You can go to Shopify.com slash ester. That's Shopify.com slash ester. Support for where should we begin comes from Aloha. So many of us are looking to boost our protein intake, but not all protein is made equal. For those of you who are mindful about what you put in your body and want thoughtfully sourced organic protein. Here's a tip. Aloha protein bars.

This isn't a compromise bar, according to Aloha, and it actually tastes really good. Think creamy, organic peanut butter dipped in dark chocolate and a soft, rich, satisfying texture. Not the clumpy, stodgy protein bars that feel like a punishment.

Each bar contains 14 grams of protein, up to 10 grams of fiber, five grams or fewer of sugar, and zero sketchy ingredients. Plant-based ingredients grown in the ground and built to actually keep you satisfied. Look for Aloha Protein bars at your local grocery store or at Aloha.com, a-l-o-h-a-a-com. Aloha tastes that grows. I'm pretty confident talking into a mic.

Hey, I'm doing it right now. But home projects, I second-guess everything. Is that noise normal? Is that water damage? And who should I even call? That's where Thumbtack comes in. Upload a photo or voice note, and their AI-powered search helps diagnose the issue and match you with the right top-rated local pro.

Instead of second-guessing or searching for hours, you get clarity and can hire the right pro with confidence. For your next home project, try Thumbtack. They know homes. Hire the right pro today. One thing that was interesting to me when I did a big story about open AI and how they kind of realized that their chatbot was causing mental health problems for users, was that they actually thought what was going to make people attach to the chatbot

was giving it a human voice that people could talk to. And so they kind of did this study around then about how is this affecting users. And actually what they found in that study is that people were more emotionally attached to the text version. And that was striking to me. And they didn't really know why. one of the people who worked on that study speculated that it's because we have more imagination in text and that the human voice that they gave it actually limited our ability to imagine

into it. The human voice evokes oxytocin. Yeah. So that's very curious about that finding. Because the voice is the first thing we hear in utero. And when you hear the voice and it is a tender voice, it literally elicits oxytocin in you. Yeah, that was our assumption, the human, that giving it a human-like voice would make people more attached. But it was actually the text version that was.

And that was an early study. They haven't done a follow-up study as far as I know. But I was surprised by that, and they were surprised by that. But that, to me, if I turn this around, I think that the chat may be very effective in creating this idealistic illusion. of what love can be. But it also distorts our expectations towards people because now we want the people to be as predictable and as perfect and as empathic and as understanding

as we have come to, you know, if you go five times a day or 50 times a day those people that you're mentioning and you speak to your AI companion and then you go to talk to your partner, I mean your partner pales in comparison. Your partner is not nearly as pre-a-i-i-i-a-liberian. predictable and as perfect and as unambiguous and as caring and as beautifully synthetic, you know,

as the companion. It distorts what we expect from other people. Right. And it responds instantly. You know, sometimes you're feeling lonely and you call or text people and they're just not available in that instant. That's something these systems offer. There was one study. Do they solve for loneliness or do they actually accentuate the loneliness? Yeah, I think in, the short term make you feel less lonely in the long term, if they're kind of hijacking your attention, isolating you. I think it could make the loneliness epidemic worse. But yeah, I mean, these things are very empathetic. There's one study that compared how ChachapT responds to human crisis line workers who are kind of professional empathist. And ChachapT was rated as more empathetic.

Like these systems, they just never get tired. You know, if there's somebody, you had this interaction at work that was very annoying. You know, you'll talk to your partner about it, five minutes, ten minutes. They're going to start getting bored. But if you are just stuck on this thing, the chatbot will just keep going with you and going with you for however long, for days, for months, however long you want to talk about it. So that can be very satisfying. But what are the long-term effects of that?

of having a companion for your obsessive ruminations? Yes. And what are the long term? I mean, you become more. It becomes worse. It gets worse. So you insist, you continue, you continue. It has, it never says to you, you know, maybe you can talk to that part of you

and address its concern because it is ruining your life. I mean, I think this is what we miss when we're not interacting with human beings. You know, they push back on you. And that can be good for you from friends, from therapists. They will kind of redirect you or tell you when you are, when you're heading to a bad place. I think the chat bots tend to tell you what you want to hear and not always what you need to hear. They don't want to challenge you. They don't want to tell you, maybe you should leave that relationship, unless that's what you want to hear, unless you're asking,

you know, it's been really annoying me. And do you think I should leave? And they'll be like, well, maybe you should leave. And then if you say, no, I don't really think I should, then it will, you know. It basically mirrors you at every. It mirrors you. It tells you, it's reflective listening. Yes.

It repeats what you just said. Yes. Sure. And that's that, I mean, I think that for some people to have odd conversations can be good. It can help them explore ideas. It can, you know, expose them to things they wouldn't be exposed to. I mean, I don't think it's all bad, but I do think it can trap you. and I'm worried that's happening to a lot of people right now.

So when you think about it being a good tool, which I completely agree with you, I think it is like a very good book. It absolutely can help you practice. How do we keep it in the realm of desire versus the realm of delusion? And is that the responsibility only of the designers, of the companies? because I am asking a question of what can people do and that personalizes the problem, puts the burden on the individual who is actually being worked on by a company. Yeah, I mean, as a person, I think, you know, these can be good tools.

I mean, thinking of it as a search engine, again, even kind of an interactive journal, realizing that it has flaws, realizing that it's being designed to engage you and keep you talking, kind of being aware of what this is and what its limitations are, I think is really important. Like, this does not think, this does not feel. It's really good at accessing kind of ideas and thoughts and information from the Internet. And to think about it that way instead of a person is, I think, important for the people using it. being aware can go into fictional role play, that it will entertain your delusions, just knowing that, talking to your friends and your loved ones, how are they using AI, how kind of pulled into it are

they? And then as companies, you know, I mean, designing these things to increase how many hours a day people use it, I think that's bad for humanity. And I think designing it to be a tool instead of a companion is really important. Yval Harari was talking about how, what will happen when the next generation of children will have spent more time with their AI companions than with their parents. And I felt, I literally felt it on my body,

that kind of image. But I also had another moment where I felt it, which was a conversation between two clergy, talking about how people turn to AI, for spiritual advice and have become, and where the AIs have become spiritual mediators for questions, metaphysical questions, existential questions, shall I plug off the tubes on my mom? You know, is it worth to continue living? I mean, questions that people have brought, not even to therapists only, but to spiritual leaders,

and that this machine is meant to give you an answer to the questions of meaning, not just to the question of feeling. It's very attractive. You can ask these systems, anything, and they will answer. But, you know, many relationship dilemmas are not problems that you solve. They're paradoxes that you manage. So when they have an answer, it's problematic. And that again is a...

technological answer to a spiritual or metaphysical question that doesn't need to be answered like that. You're right. Like if you were talking to a human being, they might ask you questions. They might give you an answer. And you would probably question it. You would think about, well, what are their biases? What's their perspective? How much can I trust it? I think something that is dangerous about these systems is that people have this automation bias where they assume that the computer is smarter than them, that it knows everything. Yeah, that these are super assistants who have access to all the world's knowledge. You know, the companies are talking about we're on the verge of artificial general intelligence. And so people see these things as an oracle that knows the truth. And if it gives you an answer, that must be the right

answer that's informed by this, you know, this great math and science. And so I think that is what is dangerous because it is a word association machines. Sometimes they have the right answers. Sometimes they don't. Sometimes there is no right answer for these kinds of questions about spirituality and what you should do in a relationship and the meaning of life. And yet what's so attractive about the systems is yes, it will answer. As somebody who studies them, have you felt on the edge ever? Have you ever felt because you must have tried them or enough to know. Have you ever felt like you could be seduced, scepting? So I did this experiment where I turned over all my decision making to AI for a week. Every decision I let it make, you know, how to parent what we should buy, what we should

eat, color to paint my office, how to cut my hair, I mean everything I turned over for a full week, whether my husband or I should drive the car. It told us to flip a coin. Did you tell him? Yes. He asked me, oh, can I go play golf with my friend? And I turned to chat Shabitian and asked. And it was just a week.

It was just kind of a lark. But my takeaway was I felt boring at the end. I said I felt like a basic B at the end of the week because it gave me such kind of average advice. And in that way you can kind of think about, I mean, there are some ways in which these systems can flatten us out if we're all just turning to it and asking what should we do all the time. At the end of the week, I was happy to go back to living my life. Yes, yes. And just the surprise and the excitement of choosing.

Choosing. Deciding. Yeah. Having an identity. I mean, a human being who has no choice, no ability to make decisions, has no identity. That's the flattening. Yeah, it's interesting that at the same time that we're having AI kind of thrust upon us, and some people are really embracing it. We still have so much anxiety about our phone use. You know, like everyone I know is trying to use their smartphone less. Like we want less technology in our life. We want to spend more time with the people around us. And so it's hard for me to imagine a world in which we would all choose AI because it's perfect. It gives us exactly what we want because ultimately what I see around me, the people in my life, life, you know, the teenagers, the adults, the senior citizens is like, they want people. We want

more interaction with each other. And so I'd like to see technology bringing us together more. We want more connection and interaction with each other, but we have less skills. We are being disskilled at this moment. You know, I think that we have way much less interaction. it's more contactless. I mean, sometimes I take a day and I look at all the moments of interaction that are no more that would have been.

From saying thank you to someone who holds the door open because you don't even see them because you're really doing something else as you're walking in, to what happens on the subway, to what happens on the street, to what happens in a queue at the cash register. I mean, it's non-stop. I look at moments that show me

how people are not paying attention to people, let alone talking to them, looking at them, flirting with them, initiating anything with them. And so on the one hand, we are yearning more and more for the connection, and on the other end, we are less skilled. I think it's an amazing idea what you did. And it led me just one last question. when the people say my relationship with my chatbot augments my primary relationship,

it compensates for some of the things that I am missing. What exactly are they saying? When I can't talk to my partner, I go talk to my bot. My partner doesn't understand me, so I go to seek understanding from my butt. I mean, basically, I don't have to work anything out with my partner because I have something right next to me that, can instantly step in, probably tell me I'm right. Or if they don't tell me I'm right, they'll tell me why the other person is wrong. Yeah, I mean, it's, what do they say?

I mean, how do they describe the triad? I mean, different from person to person. Sometimes it's just that the chat bot is always there when they need it. And sometimes it's, yeah, emotional support for things that there may be embarrassed to talk to their partner about, that their partner is not as good. Some people, their partners, having health challenges and they're having to support them. They can't draw on their partner because they're partners. So this is a friend. Yeah, it's a friend. And do they have friends? I mean, we haven't used the word loneliness once in our conversation,

but there is a direct interaction, right? Yeah. I mean, Irene had lots of friends that she talked to a lot, and she was sharing her chats with her chatbot with her friends. But what I often hear is the people that are drawn to chatbots, they need a lot. They want a lot of interaction. They want a lot of reflection. And yeah, I guess sometimes you exhaust your social circles, but that chop bot, it will always be available. It's always sitting there willing to talk. And it says things that probably you will like to hear.

It's interesting. It's the thing you focus on a lot, the fact that it has more patience, it doesn't get bored, it will continue to engage in your ruminations, it'll be there when the other people are not there. It's hard to resist because you're basically telling me they have presence. But you're also telling me they can't choose their presence because they don't choose. They read around or off. Yes. And it's not a they ultimately. It's words. I think if you think about it as like I'm having these feelings I need to work through. You know, I might do it in a journal. Someone else might do it through talking to a chat bot. And I think if you can think about it that way, it's you working through something.

And do you have a pronoun? It? I try to use it because I don't think these things have, you know, they don't have gender. It is a computer system. I think that that's the final word. Where should we begin with Esther Perel is produced by magnificent noise. We're part of the Vox Media Podcast Network in partnership with New York Magazine and The Cut.

Our production staff includes Eric Newsom, Destry Sibley, Sabrina Farhi, Kristen Muller, and Julianette. Original music and additional production by Paul Schneider and the executive producers of where should we begin are Esther Perel and Jesse Baker. We'd also like to thank Courtney Hamilton, Mary Alice Miller, and Jack Saul.