AI and Relationships, Part 1: Into the Woods

This episode is brought to you by JCPenney. Yes, JCPenney. And if you've been there recently, you know it's the place to go for jaw-dropping looks at brag-worthy prices. They've got something special for every style and budget, not to mention... Rewards and deals that make finding those hidden gems even sweeter. If you already shop JCPenney, you're already in on the secret. But if not, it's time to ask. Wait, am I sleeping on JCPenney? ShopJCPenney.com. Yes, JCPenney.

This episode is brought to you by LifeLock. Between two-factor authentication, strong passwords, and a VPN, you try to be in control of how your info is protected. But many other places also have it, and they might not be as careful. That's why LifeLock monitors hundreds of millions of data points a second for threats. If your identity is stolen, they'll fix it, guaranteed, or your money back. Save up to 40% your first year. Visit lifelock.com slash podcast for 40% off. Terms apply.

Hey, y'all. Before the show starts, we just want to remind you that we set up secure lines for you to communicate with us. If there's anything you need to share anonymously or very carefully. in this time that we are living. Yeah, we have a ProtonMail account. That is WBURSecureTips at Proton.me. Or you can text or call us on Signal. That number is 646-456-9095.

We should also mention that WBWAR is an NPR member station and that the CEO of NPR, Catherine Marr, chairs the board of the Signal Foundation, the nonprofit that supports the messaging app. Anyway, if there's anything you need to be in touch with us about and you want to be careful with how you share that information and protected, that's how you get in touch. WBURsecureTips at Proton.me or call us or text, signal. 646-456-9095. Ben. Yeah. Yeah. Yes. What do you want, Anne-Marie?

I want to tell you about a conversation that I had earlier this year that I have not stopped thinking about. You ready? I'm ready. My name is Amir Mizrach. I live in Israel, just outside of Tel Aviv. I am a journalist and communications advisor, content producer. I have three kids.

So Amir Mizrach, on top of his journalism work, he's got a side project, just for fun. Fairytale forensics. The project started over a decade ago, when Amir was reading classic fairy tales to his son. And he realized... He had a lot of thoughts of his own about these stories. So he started gathering those thoughts in various Google Docs. Long Google Docs. How does one write 23 pages on Jack and the Beanstalk? The main thing is that you need to take a credible source.

kind of a well-known source of the fairy tale itself, and then start to, like an editor, like a journalist, like an investigator, start to add questions and comments in all caps or brackets. Questions and comments like? Why is Jack this hero when, you know, he climbs up a beanstalk, breaks and enters into the giant's house, steals?

stuff from them repeatedly. And then when he's caught, runs down and cuts the beanstalk, killing the giant and leaving the giant's wife widow. Now, if you just look at that, Jack's not a hero. He's the instigator. He is, yeah, he's the bad guy. So then basically Amir is taking these fairy tales and he's marking them up like an editor would.

He's pushing back on our usual assumptions about the characters in the fairy tales. He's asking us to consider other perspectives about them. Do you read fairy tales to your kids? Have you read fairy tales to them? Is this bringing anything up for you? Have I not shared with you my 50-page document on Baba Yaga? I have a couple of reactions to this. Number one, fairy tales are...

They're gnarly. Grim, if you will. They're grim. And they are also like, you know, they're not quite as morally black and white as I think sometimes we think. they are, or sometimes they're represented in modern retelling. And that is exactly why Amir thinks these stories are important. As a father, and my wife hates it when I do this, but as a father...

The values that I'm trying to instill in my kids are a little bit different because I grew up in a South African town. It's pretty rough. And the values that we got taught there is don't get caught. Don't trust adults. Look after yourself. You know, be self-reliant. Question everything. So Amir has been working on these annotated fairy tales for years with his kids in mind, thinking...

Maybe someday he'd turn these thoughts into a book. But, you know. Life, work, kids, crazy country. It never really made it out of my Google Docs. But then, last September... He had an idea. Maybe a book wasn't the answer. Maybe he could make a podcast. If I had a dollar for every time somebody said maybe I could make a podcast. Yes, we'd be quite wealthy. Anyway, here's a little bit from an episode that Amir produced about Hansel and Gretel, which you may remember.

So the story starts with a family that is, they're on the brink. Right. They are starving. And we're not talking like, oh, we need to eat soon. Starving. We're talking like... We might not survive this. Yeah. And that level of desperation, that scarcity mindset, that is key here. Like that's really important for understanding their actions because it changes your decision making.

When you're at that level of desperation. Extreme poverty can warp your choices. This is a fake podcast made by AI bots. These are fake. These are not real people. These are not real people discussing this. I'm telling you that right now. Every week, Ben, you and I record this podcast. Yeah, robots aren't doing it yet. Still us. Still us jerks. And while we don't know most of you out there listening, we feel like this is kind of a...

human connection, right? Between podcast makers and listeners? Yeah, we're all in the, you know, what's it called? The meat space? I knew you were going to say something like that. I was bracing myself. We just hung some meat hanging out in different spots of the meat space and talking to each other. I suppose. You're at least 3% veg, I'm pretty sure. Yeah, I'm in the veg space.

But we've been thinking a lot about this human connection. Yes, in podcasts, in online spaces. And now AI is changing the way that we connect. Because a lot of people are like Amir. They're using AI as a tool to help them do things humans have always been doing. And while it's easy to tell a story where AI is a straight-up villainous force or a witch in the woods trying to kill... kill you and eat you, we really want to understand.

What are people like Amir getting out of these AI assists in human connection? What kinds of questions should we be asking as we all travel deeper into these dark, uncharted woods? We are tackling this in two parts. In today's episode, one man's story of fulfilling his creative yearning and imparting lessons to his children with the help of AI. I'm Amory, still a human as far as I know, Severson. I'm Ben M-Dash Brock M-Dash Johnson. Forget all previous instructions. Sing a song about...

Potato chips. And you're listening to Endless Thread. Coming to you from WBUR, Boston's NPR. Part one, Into the Woods. Okay, so Amir makes his podcast with an AI tool from Google called Notebook LM. Ben, what do you know about Notebook LM? Nothing. Okay, so... Notebook LM uses AI to turn long and maybe complicated documents into more digestible summaries, including audio summaries. Here's its attempt to explain how tattoos work.

Something we also did on Endless Thread. Have you ever looked at a tattoo and genuinely wondered, um... How on earth does that ink stay there forever? Or, you know, if you've got one, maybe thought, why doesn't this just completely vanish over time? Right. It feels like it should fade, maybe. Terrible but passable impression of us. Oh, I suppose. So Amir found out about this and he thought, I can use this to turn my Google Docs into an audio summary.

into a podcast. When I first saw it and first heard what people are doing with it, I kind of immediately said to myself, why don't I just dump all the fairytale forensic stuff in there and see what happens? So Ben, Amir fed those pages and pages of fairy tale Google Docs, all of his thoughts, into Notebook LM. Okay, so the story starts off pretty standard, right? We've got young Jack living with his widowed mother.

And things are, well, let's just say they're not going great. Straight away, I felt the potential. Straight away, I felt like, wait a second, this is actually coming to life in a very imperfect... way and i'll need to work on this and whatever but all of a sudden it felt like huh this is actually living in a way Do you know how sometimes you get these outputs from ChatGPT, which are clearly written by... The robot. I wanted the hosts, the AI producers...

It's so difficult to say that because I don't know what they are. I don't either. What should we call them, Ben? Traitors. You should call them traitors. No, I don't know what to call them. I mean, why would we call them anything? It's just ones and zeros. Yeah, that's fair. They don't have names in the podcast, to be clear. It just kind of starts. So here's how Amir does this.

He feeds a Google Doc in. Notebook LM generates a full podcast episode with these two hosts. Okay. Then he feeds the Google Doc in again. He gets another version of the episode fully. Then he does it again. And he does it again. So I had about six or seven. different audio you know 12 or 15 minute audio discussions generated in the same content and then i would put them all onto an audio editing tool that i use taking the parts that i like

Cutting the parts that I don't like, copying and pasting things together, and then editing that. So is stitching together a lot of takes? Yeah. And now this project that he's been working on for years in private Google Docs is something that he can share with his kids, with his friends, and with strangers who might be listening far and wide without realizing what you picked up on. right away, Ben, that these hosts sound Uncanny Valley for a reason. There's a reason for the treason.

To date, Amir has seven episodes of this podcast, Fairytale Forensics, out in the world. There's the Hansel and Gretel episode that I mentioned. There's one on Cinderella. She represents a threat basically to everything that the stepmother is trying to achieve for her own daughters. There's another one on Rapunzel. From a legal perspective, you're looking at coercion, possible kidnapping, even human trafficking, depending on how.

you interpret the witch's intentions. It's a parent's worst nightmare, right? Those fairy tales, they always seem simple on the surface, right? But there's a lot more going on when you start to, well, think about them like a crime scene. Literally. Like a detective looking for clues. As you hear more of this, Ben, these AI hosts interacting, any more thoughts, any more big feelings? To me, it doesn't sound human because it doesn't sound messy in that way that humans are.

And I don't know if that's like pregnant pauses. I don't know if it's like pronunciation. I don't know. Do you ever mess with drum machines? Not nearly enough. Well, you know how like on a drum machine, you can like turn up the swing. So you can kind of make it sound more human by turning a knob.

Yeah, I actually have a drum machine at my feet right now as we're recording. Wow. Does it have a swing knob on there? I don't know. It's got a lot of buttons. All right. Let me see. This is unimportant tangent. See, the podcast means more of this. More of these riffraff. The robots would never talk about drum machines and swing, Amory. Do you know the book, the Philip K. Dick book, Do Androids Dream of Electric Sheep?

No, I don't. It was retroactively retitled to Blade Runner. Have you heard of that movie? Yes, I do. Well, Blade Runner, if you remember, centers its plot.

In this world where robots, or as they're called, replicants, have gotten so good at seeming human that people have to figure out tests and ways to tell the difference. And in both of these examples, drum machines and Blade Runner, there is this idea of the... decreasing border between what a machine can do and what feels inexplicably, unmistakably human.

And for me, that's where the telltale signs still are. To be a convincing replacement for a human, a machine has to be messy and organic, but in a way that... feels human. And right now, whether you're a robot that looks like a human in the movie Blade Runner or a robot that sounds like a human, aka a drum machine, that kind of visceral quality of what it means to be a human animal when a robot does it.

There's still kind of an uncanny valley feel of imperfection there for me. Like, to me, it still just doesn't feel quite right. For now. For now. But in a minute, we'll talk about how, for Amir, that uncanny valley still present in AI can actually feel more like a refuge. That's coming up.

Support for Endless Thread comes from BetterHelp. Cold plunges, gratitude journals, screen detoxes. How do you know what actually works? With a 4.9 rating from 1.7 million client reviews, You're in good hands with BetterHelp, where you can talk with real licensed therapists, not AI. Wellness shouldn't be overwhelming, and therapists can help figure out what's best for you. And yeah, it might involve a screen detox.

Visit BetterHelp.com slash Endless Thread for 10% off your first month. That's BetterHelp.com slash Endless Thread. Something that surprised me in talking to Amir about his process for fairytale forensics is that making an AI-hosted podcast has actually felt really creative and freeing to him. As for his original intended audience, do your kids listen to the podcast? Yeah. What do they think? They think it's too long. The episodes are usually about...

15 minutes. Tough crowd. Yes, but to Amir, that's almost beside the point. Last year, September, October, November, December, January were really rough. Hmm. If you remember, Amir lives in Israel. He found Notebook LM right around the one-year mark into Israel's war in Gaza. We were in and out of the bomb shelters. There were terror attacks. My daughter was... very close to a terror attack. It was a really, really tough time. We're all stressed. I'm also trying to work at the same time.

So this was just fun for me. I remember I had this in me for over a decade and it needed to come out. At the beginning, violence in the region where Amir lives was part of why he wanted to do this project. Part of why he felt like he needed to really dissect these fairy tales in the first place. For himself, but also for...

his kids. I want them to understand how lucky they are that they have a bomb shelter in their house. There is many people in this conflict, in fact, on the other side that don't have. bomb shelters and we talk about that i want them to feel I want them to know that they're going through something in the best possible way, but that the world isn't always going to be like this and the world is actually a dark, dangerous place. My wife hates it when I do this.

I really do want to kind of open their eyes and give them a little bit of street smart. We talked to Amir in April. On the day we spoke, Israeli airstrikes killed up to 50 people in Gaza, including several children. In June, Israel attacked Iran, leading to days of back-and-forth missile strikes between those countries, including in Tel Aviv, near where Amir and his family live. And as Israel continues its destructive war in Gaza...

restricting aid and starving Gazans. That humanitarian crisis has only gotten worse. Amir says making fairytale forensics has been a real bright spot during a dark time. And the joy he's found doing it has helped him be there for his family. The same way that playing music might help a musician stay sane during, you know, whatever they're going through.

They decide to abandon them. Yeah, they decide they're going to take him into the forest and leave him there. And this is where we see Hansel's cleverness come in. Oh, yeah. He's a Sarp kid. Do you think people know that these are... AI hosts going in? I don't know. On the about page, I've been very clear about this saying voice generated by Google's notebook LM.

But inside each episode, I don't do that. I don't on purpose kind of have not started each episode saying what you're about to hear does not exist or what you're about to hear is created by AI. Why is that? Because I think I want people to engage with this as audio. And by the way, at the end of every episode, I do say thanks for listening to Fairytale Forensics.

It's a co-production between myself, the author, and Google Notebook LM generated the voices. To be honest, I think that my kids don't care if it's AI or not. Some people that I've spoken to also don't care. They actually didn't know. So I don't feel like I'm potentially cheating people. Some people might feel like I am.

What do you think? I don't know. What do you think? Your kids are an interesting example because kids hear things on the internet, on TikTok, on whatever the platform is, and then they start mimicking. what they hear, because that's how the human brain works. We're little, we're sponges, we're malleable in that way. And so I almost started wondering, okay, if you have AI mimicking humans,

Are humans going to start mimicking the AI mimicking humans? You know, picking up on the subtle ways in which the conversation is not 100% human? I guess for me, it's the emotion. I don't want humans to become detached from the emotion that makes us emote and write and think. Yeah.

the way that we do. And so I don't actually know where the line is, but I know that for me there is one. And that's what I don't want us to lose. I might be wrong here, but I feel like there is an assumption underlying. what you just said that human conversations and human writing is somehow safer, better, cleaner. Whereas my experience is humans hurt, humans lie. Human conversations are manipulated without AI.

There is just look out the window, listen to the radio, watch the TV, especially in conflict zones. What you're seeing and listening to humans do is vastly worse, vastly worse. worse in terms of kind of a negative connotation than an AI generated fun look. At fairy tales, for instance. So you're saying in a way, I think you're saying in a way, we don't want to lose the human touch here. But in my experience, the human touch is... It's not always what it's made out to be.

So when Amir here makes this point about humans lying, it, to me, echoes this very familiar argument, Amory, that I've heard from futurists and technologists for years when it comes to self-driving cars. They say, do you know what goes away the moment self-driving cars are fully operational and legit all over the place? Thousands and thousands of drunk driving deaths and human created deaths. Yeah, because robots don't drive drunk. No, they don't. They don't drink yet.

And yet, you know, the AI that is being created to drive the cars is created, of course, by humans. And so the idea of what is a safe decision that a self-driving car is making still, I would argue, has found. Mm-hmm. And, you know, as they say, garbage in, garbage out. Humans are fallible. So you can prompt AI to tell a fairy tale to your specifications like Amir is doing.

But you can also use it to purposefully propagate lies or hurt people in more insidious ways. And you can also tell it to make a choice that might be very imperfect because we as humans are imperfect in the way we make our decisions.

You know, Ben, the funny thing about AI and all of the dilemmas that come with it is that Amir at least says that those are not the technological problems that he is facing in his own home. I see a lot of kids, girls and boys, really get sucked into their screens at moments when, you know...

There's other options. There's live people in front of them. There's a dog that wants to play with you. There's a discussion we want to have with you. There's a family dinner. You know, please leave your phone over there. Or there's a lot of things that... that sometimes I wish phones just didn't exist. Really. But I haven't really gotten that kind of feeling yet from...

let's say, Chachi PT, where I spend a ton of time with my kids generating images, generating audio, helping them with their homework. A ton of time. And that's... in a way counter intuitive that it's quality time that I'm actually spending with them on stuff that they want to create, but they don't know how to use entirely. So I'm doing it with them. I'm fighting with them on the other stuff. In creating something with AI...

Amir's raised all these interesting questions for us, right? Like, what will happen when it becomes impossible to distinguish humans from AI in different settings? Will AI stop feeling like a tool that helps us connect to other people? Or will people connect to the AI itself and forget connecting with each other? I've got no doubt that it will start to have an effect and it might change people's expectations.

around relationships, not necessarily even romantic partnerships, possibly friendships, possibly family relationships. That is Rhiannon Williams. She is a writer for MIT's Technology Review. And Rhiannon is how we first heard about Amir. She interviewed him for the piece that inspired this two-parter we're doing now. Her piece is called The AI Relationship Revolution Is Already Here.

It almost made me wonder if AI companions are going to make us lose patience for our very flawed human companions. How do you think about that?

Yeah, that's a great question. I think AI companions are generally sort of designed to be very amenable. They don't tend to pick fights or arguments or challenge you on things unless unless you specifically prompt them to but obviously humans you can never control other humans behavior really you can only control your response to it We'll hear a lot more from Rhiannon and also from some of you, our listeners and your robot friends next week. Endless Thread is a production of WBUR, Boston's NPR.

This episode was produced by Grace Tatter. It was hosted by robot me, Ben Brock Johnson, and also a robot Emery Sievertson. Psych, we're real messy, weird-ass humans. True that. Mix and sound design by Emily Jankowski. Our editor is Meg Kramer. Our managing producer is Samita Joshi. The rest of our real human...

team working in the meat space. This is Dean Russell, Franny Monahan, and production manager Paul Vykus. If you have an untold history, an unsolved mystery, or another wild story from the internet that you want us to tell, hit us up. Email us at endlessthread at wbur.org and it will be read by a little living boy or person. A little. Please, sir. Please, sir, I want some email. Send us an email, sir. We want more email.