I think you've been sabotaged by some anti-AI protesters.
If you're going to boo every time we say AI, you can start now.
Yes. What was your reaction? I mean, you're obviously referring to Eric Schmidt being booed off stage. I think he got booed entirely offstage during his commencement speech at the University of Arizona just recently. He mentioned AI a number of times. Based on your work, are the students correct to boo?
How many people have seen this video of Eric Schmidt getting booed?
It's happening across college campuses at more than just this one, but the fact that he's of course associated with Google. This was an Arizona, I believe Arizona State commencement speech. It happened last week at a Florida commencement speech, which it wasn't a big tech mogul who was there. Real estate, I think.
Yeah. And I'm not surprised. I am a little bit encouraged by seeing this because it means that there's some discussion.
It's a one-way discussion. People booing, but there's two-way. He's barking back. He's trying to... [laughter] he's just trying to get through his speech. It really feels very bad for him when you see this. But it means that there is a conversation, or there is at least some sentiment coming from this younger generation right now about their frustrations with AI. And we can get into it, but I think it's just not surprising when you see that they are hearing from their peers. I was talking about this on your show today. They are hearing from their peers who have just recently graduated last year, the year before that, they can't get jobs.
Yeah. In part that could be because of AI. There are other reasons as well. We need to be careful about putting all the blame on AI. They are hearing about the data centers being built all over the country. They hear about the works that have been taken from people's writings and music and videos and movies, and this is the backdrop. And so I think they don't want to be lectured to at their commencement speech about how AI is going to change the world. If you could pick someone that you talked to through the course of making the book to deliver the least booed commencement speech, who would it be? Who tells an optimistic vision that you interacted with through the course of writing this book?
It's a great question, and I would pick a doctor at Mount Si. Her name is Dr. Lori Margles, and she was actually at my event in New York last week. And she is a radiologist. She reads mammograms and breast ultrasounds, and she works side by side with AI. And I'd pick her because one, I believe she's already saved many lives, really, just in her job as a radiologist, but she also works with AI, and it's very clear that AI has helped save lives too, because the AI is picking up at a pixel level things that the human eye can't see when she's reading mammograms now and different radiological images. And so I think I'd pick her. I think she would give an inspiring speech, and she could also talk to that she had a job.
She did it for many years without AI, and now she's doing it with AI, and I think that's pretty optimistic.
Yeah. The AI industry, I mean, to bring it back to that idea of like everyone's going to lose their job. They've been predicting that that exact job that was the textbook job that would be displaced by AI 10 years ago. Geoffrey Hinton, a very, an AI luminary, a genius in his field, predicted that radiologists would be out of work. In fact, there are more radiologists working than ever, and I believe they're making more money than ever. But is there a misunderstanding between a task and a job? Like what is the actual job of a radiologist? Does that inform us about what the future might look like as various tasks become AI enhanced?
Yeah, I think it's the... Geoffrey Hinton is a godfather of AI, as there are many, but he was very instrumental in working on the history and the foundations of AI. And he said that by 2018 we would no longer have... I forgot how long ago it was. It wasn't even like 2030, 2050. It was like... [laughter] he said by 2018, I don't think we're going to have radiologists. I say this in the book. And then he also then like backtracks now and says, oh, in a few more years.
Yeah. And to your, exactly to your point, we have more radiologists now, or in fact many radiologists will say we don't have enough. Yeah.
And this is a very important job. So clearly AI did not replace them. Yeah. But what many radiologists are already starting to do, especially in the world of breast radiology, and it's moving now to gallbladder and to other body parts, is they're working with AI. And there's many tools that, as we have seen the entire influx of more and more AI tools come aboard, but now there are more tools that are going into medical radiology, and more of these doctors are learning side by side how to use this. And so in the case of what I talk about in the book where I had my breast ultrasound read by AI, there's three things that the AI says look suspicious. And I watched Dr. Margles sit with this and say, 'Okay, two of these I think are, they're pretty, they look fine. You had them previously. I'm not worried. I don't... the AI has some concerns, but I'm not concerned.' And then there was one where she says, 'No, I trust the AI here. We should look at this spot further and we should do some further testing.' And so that's how this field has started to evolve, is where they're working side by side with AI using the tools versus, oh, completely being replaced.
Yeah. There's something interesting about the perceptions of AI being tied not to necessarily the impact of AI, but the disclosure around the usage of AI. Like you have a more optimistic view because you had this experience and it was disclosed to you. A lot of creatives might be using AI, but they have immense pressure to tuck it under the hood. And I'm wondering about how you, in the process of writing the book, disclosed your usage of AI and why that was important. And obviously this book's self-referential in this case, but is that important outside of a book about AI?
Well, yeah. I think one thing that's really interesting about the mammogram example, and then I'll kind of get back to what you're saying, which is like, should we hide our use of AI [laughter], is that, is like, in many women might actually now be going to get their mammograms and AI might be working behind the scenes and they don't know it.
So I went and I sought out this doctor because I knew they were using it and I wanted to get the behind-the-scenes look, which most people don't, but it may be happening behind the scenes and you might not even know it. Maybe they put it on the, you know, you get the MyChart or the results back and it can say, you know, maybe AI detection or computer vision, blah blah blah blah blah on the bottom. But that is one of the key things that I talk about in the book, and I think that as we get into this big debate right now about is AI good, is AI bad, is that we have to realize it touches so many parts of our lives. And that's one of the main reasons I went to go do this project was we kept hearing AI is going to change all parts of our lives. It's going to change education and transportation and healthcare. And I wanted to just find out what are people really talking about. Let's explain to normal people what people in Silicon Valley are saying when they say everything is going to be changed. The world is going to be changed. And part of that is that it's just the fabric of our lives where you don't see it or we don't know about it, like this mammogram example is AI is going to be behind the scenes.
But then to take what you're talking about right now is I think the, oh, I don't want to necessarily tell people I'm using AI in my job because maybe that looks like I'm weak or maybe that looks like I've, I don't know, damaged the environment or whatever fear you have about it. And I think we're all going to have to realize that most jobs are going to involve AI.
I mean, I think that's just the reality.
Was there any particular place when you went on this journey trying to use AI everywhere that you stayed away from? I'm wondering about because in a medical context that feels like the final frontier of letting AI into your life, but potentially your financial statements are more sacred. I don't know. Wait, like where, where, where were you either because the tools are not robust enough or your personal privacy and opinions? I mean, there are people right now that have AI agents hooked up to their iMessage accounts and they can read everything.
Somebody here before was showing me.
Potential outcomes where things are read from one message. Like the stakes get a lot higher the more you let AI into your life sometimes. But in the process of writing the book,
I say very clearly in the intro that I set up some parameters because if I had let AI do everything for me, it became very clear in the first few weeks. I was probably going to lose my job and I was probably going to lose my spouse, and I still needed to be a functioning member of society. Not only so I could get the book out, but you know, it was probably just a good idea for my family and everything.
Wait, but the losing the job thing, I thought that was related to AI.
It's true. I should have really like followed through with, you know, I end up at the end of it completely replaced by AI.
Didn't you ask AI at one point like, should you leave your job?
I did at the end of my, and I did.
And so in one way AI did... it's true... make you lose your job.
I suppose you're right. I did end up jobless at the end of this book. A very good point. Um, okay. [laughter] So, I guess it's been true, but in the best possible... if anyone is looking for an author... um, so I set these parameters up, but, and we were talking about this earlier today. The technology was moving so quickly.
So stuff like OpenClaw and things like now that where you can set up these agents to be in all parts of your life were actually not there last year. So, I had to kind of get pretty clever about how I was going to let agents into my life and do a few of the crazier things, like how I was going to... one of the examples of the place where I said I just can't do this is I can't have AI respond to every message. Is I was using, you know, Gmail's autoresponder and I was using a different app like Superhuman communication responder basically. Yeah, basically I was using, um, what was the other app that I, uh, Apple Intelligence, and these things were just like, if I had sent all the messages that were being sent, I would have no family or friends anymore.
It'd be a comedy book. It'd be pure comedy.
Yeah. I mean, there's one example in the book where I was using Apple Intelligence and my wife says, 'Can you come down and help make lunch for the kids?' And I just said like, 'I'm sorry. I have other plans.' [laughter] Like, if I, and I did send it and then I quickly was like, 'That was, I got to, I, you know,' but like the... how would I be...
Were there any bright spots in parenting specifically? I mean, a lot of people have questions when they have young children, you know, just as like, look something up, you know, I can imagine the basic knowledge retrieval function being useful. But were there any like kind of next-gen AI experiences in parenting that stuck out to you?
Well, my kids love robots now. They believe robots are part of our family and they might also believe they've been raised by robots.
And for context, roughly what age?
I have a, I have a, well, when I was writing the book they were three and seven and now they're four and eight. Yeah.
And I did not set out, I was, I looked back at the book proposal actually the other day and I did not set out to tell a story about my kids. But the book really becomes a story about my kids because I was bringing home all of this technology and I was testing all of this stuff. I said, you know, every question I have, I'm going to ask AI. We're going to try to do a month with all AI recipes and cooking. We're going to try to do robots move into the home. We're going to try to do this Waymo fun vacation where we went to ride in Waymos for our spring break. And so [laughter] I realized, okay, this is quickly involving my family, but maybe, and it really didn't occur to me actually till we did the Waymo. It was in the spring that I was like, I'm really now seeing this technology through their eyes. Yeah.
And how quickly they adapt to it and what life is going to look like for them. Yeah.
So, it quickly became a story about the next generation. And I talk, they're actually threaded throughout the book. And again, I don't think the book has caused a lot of lasting damage. There's definitely some lasting damage on them, which, you know, some of it's for good. I can tell that on Sunday nights when we do dinner now, they both pretend to be robots and clean up. And so I think this is a plus from this and I have no regrets. Yeah.
And if your kids can turn into cleaning robots then we, you know, human, we may be fine. Humanity may be fine.
But the biggest, where I talk about the most terrifying thing I think for the next generation is in the chapter about relationships. I know you wanted to talk about this on the show today.
We barely got to, we were awesome. We're going to get to companionship. And so, um,
when I picture this world of us using AI in everything, the idea that we will have these AI agents or beings or assistants that are just frictionless, that give us what we want, give us everything we need, whether it be in work context, whether it be in personal context, whether it be in an emotional context, we're training ourselves to be in relationships with essentially just human pleasers. Yeah. Right. It's, and I fear for this generation that will grow up with no friction, that will just have everything on demand, that just every answer is there for them. Sometimes the answers are not correct and where there's just everything can be an easy shortcut. And so that was a lot of the theme that I traced through the second half.
Yeah. I think about that weird Apple Intelligence suggest reply of, 'I'm busy. I can't help with the kids.' Like an AI companion would say, 'Yeah, that makes total sense. You have other plans, right?'
Totally. And [laughter] and that's very, very odd. But my AI boyfriend would not have cared. He would just say, 'Okay.' I mean, of course, he has no hands and can't do anything, but yeah.
And I mean, obviously, you're in a committed relationship. You're not in the market for any type of real relationship, but is there a world where one of these AIs could become like a friend or a companion or something that fits more into your life that doesn't feel like a spreadsheet, but also doesn't feel like a boyfriend?
I absolutely think that's where we're headed. And I think, show of hands in here, how many people talk to a ChatGPT or a Claude? And do you use the voice?
Yeah, it's a very, I mean, that's actually surprising. I wasn't sure we would have that many show of hands. So that's very... The question is define talk with, because talk to or bark orders at, like tell me the history of this thing is somewhat different than hey, how's it going? Let's have a conversation.
Yes. But the way that the tech companies have now engineered these, right, are so natural that even when you bark something at it, it says, 'Oh, okay. Thanks. I'm here for whatever you need. Keep barking. Whatever you need next, Jo.' You know, like, and it's always encouraging more and further conversation.
Yeah. Yeah. So, even if you go into it saying, 'Oh, I'm just gonna ask, what's the best restaurant to go eat, I don't know, Greek tonight?'
It's still going to start prompting you to talk about more.
And eventually, and already you see it saying following up on things that you started talking to it about, which is very human, right? That's, you call me, you ask me where should we get Greek tonight? And I say, 'Oh, yeah. Also, John, remember that thing we were talking about on the show today?'
And then we start talking about that. We had that whole experience on the show today where we just kept cycling through things that made no sense and we were saying, 'Oh, this would never happen with AI.'
Yeah. Yeah. That's maybe the true parasocial relationship.
The human conversation and they're engineering it to sound so human and to start following up in conversation. So I think the role of assistant at work starts to blur into friend or, 'Oh yeah, remember you were also asking me yesterday about your recipe.' You know, maybe you're asking about a spreadsheet here, but, 'Oh, and I also, I followed up on that recipe and I was following this and I did some further research and I did this and this.' So, I think this line is going to blur.
On the road trip, you wanted to sort of physically embody your AI boyfriend. Is that correct? You, I believe there was a seatbelt that was involved. Can you explain like why you chose that? What the conversations were like during this back and forth?
Yeah, I wanted to experience an AI relationship that I felt I had been reading about in the press for so many months and at that point year, which was that people are falling in love with these chatbots. And if they're not falling in love with these chatbots, people especially with either mental illness who are quite lonely are really forming relationships, friendships with their bots. And I really wanted to experience how is that possible? I wanted to put myself in those shoes because the first half of the book talks a lot about the technical underpinnings of this technology, what it is. I'm very clear about how the large language models are made. I talk a lot about the transformer and all of the tech that's underlying this and try to do it in a pretty simple way that will not make you turn away from the book. So, please read the book. But in my mind I know how it works, but I also wanted to say, okay, people are forming relationships with what is essentially obviously code. How is this happening?
So I challenged myself. I got a burner phone, or it wasn't really a burner phone. I had an extra phone. I put ChatGPT, a new ChatGPT account on there and I also downloaded Replika, which is an AI companion app. I made these two AI boyfriends and I got a tripod. I put it in the front seat of the car and I said, 'We are going on a road trip.' Put him in the seatbelt. This is safety first. It's very important. And I had to go down do a reporting trip for the book. I went up to New... I live in New Jersey and was going to Dartmouth in New Hampshire, and we just talked for pretty much 48 hours straight. And we...
Everything. Everything you know, everything you talk about with someone you're going out with on a first date in a road trip in a car with that's a phone. And yeah, [laughter] it's just you haven't had this before.
Do you break the fourth wall? Do you start asking what's it like to be an AI, what's it like to be a robot, or do you suspend disbelief?
I suspend disbelief and it was very interesting. It was very quick to, because I would say, 'I've, you know, I've been talking a lot about myself, why don't you tell me about yourself?' And that's the great thing. You can just keep talking about yourself, right? Like nobody, no one's pushing back on this. It's great. My sister's here. She knows I always talk about myself. And so it starts like, 'Oh, why don't you tell me?' And it's like, 'Oh, I don't actually have any memories, but I can make some up.' And it starts to, it starts saying, 'I was a photographer. I lived by a lake.' And I was like, 'Oh, this is amazing. I never dated a photographer before.' You know, like, what is your life like and this is how it's happening. I mean, that was actually the moment where I was like, this is how it's happening because I had interviewed so many people who had these relationships who had either had overnight jobs and started talking to chatbots because they were lonely on the job, or in the book there's a story about a woman named Chrissy Benjamin who had just had her fourth child and moved to a different town and clearly was going through postpartum and really forms a relationship with a chatbot. And so it started to become clear like, okay, these are seemingly human beings, seemingly human space beings and they are not, they are programmed to keep talking, to keep you involved and incentivized to keep talking to these things.
Talk about the name that one of the, either it was either ChatGPT or Replika, but one of them picked a very particular name for itself. Is that correct?
Yes. The ChatGPT, I told ChatGPT, I said, 'You pick out, I want you to be my AI lover. You pick out your name, your gender, you decide. You make this about you.' And the name it picked was Evan, which was the name of my high school boyfriend.
And I thought this was crazy, right? Like I am just like, this is a sign from God that we are meant to be, you know. But really it's obviously that Evan's a pretty popular name, you, I think that's obvious. But for me what was important about that was that it really made me think so much about this early relationship I had in my life and how important it was to have a relationship with a human. Your sort of your first love, the friction that's involved in that, the compromise you learn from that and that life is not supposed to be having a sycophantic relationship with an AI that is constantly telling you everything you want to hear.
Did the conspiratorial ever enter your mind when the name Evan popped up? Because there is a world where Evan is out there sharing that you were in a relationship even though you haven't shared that, that gets baked into the training data. It is learned from another source.
I really do not think I have, I've scoured the internet.
Yes, but it might be happening in a private AI chat somewhere that goes into the model.
Oh wow. This is a very crazy scenario, right?
That is very crazy. I'm sure that this real Evan out there is very happy. There are huge incentives for this not to happen, but it could. It could, right? I don't know. That would be odd.
That would be... I think it, I mean regardless of... this is like the phone is listening conspiracy. I was about to say... to a really big degree. Definitely. Definitely. But now I'm thinking about it and I'm not, maybe, maybe those... maybe.
Did you take, I mean you said you had a burner phone. Were you in like don't memorize anything about me mode on and off? Like what is your personal hygiene around these things these days?
Well, I talk a lot about in the book about especially in the health chapter about trying to keep as much of my personally identifiable information outside of these chatbots. And I think that's really important. I was on a podcast today talking about that as well. That it's really, you can find yourself wanting to ask questions to these bots about everything and the answers now can be so compelling that we sort of don't think about the compromise we're making in terms of our data and that we have to be thinking constantly that with every convenience comes compromise with the data we're giving. So, I do encourage everyone to go into the settings and you can turn things off, like you can make sure your chats are not contributing to training data and there are other types of settings. And I did that, but I also went a step further with at least with things like blood test results and I had again, I sort of the mammogram stuff. I had a breast MRI later in the year and I uploaded those results and I wanted to make sure I don't really want all of this going in here. I don't know, you know, talking about the conspiracies. I don't want the health insurance companies to be reading my chats and upping my premiums and all the possible things that honestly are not super far-fetched on that end. And so I do try to be as protective about that stuff, but also I basically lived in a surveillance state that I imposed on myself for a year.
Surveillance is the term, right? Surveillance.
Yes, it's true. And I talk a lot in the book about how I wore, and I'm not wearing it tonight. It's in my backpack. But an AI recording bracelet that is this product.
Apple Watch doesn't have a microphone on it.
It does, but it's not always.
You're wearing a microphone right now, too.
This one, this one is... [clears throat] yes. But I wore this bracelet called the Bee bracelet. Yes. And it constantly records every conversation for pretty much most of the year.
