BONUS: AI therapy slop

Ryan wrote about using ChatGPT for therapy, and a lot of responses to that announcement involved calling him an idiot. This week, he sits down with Grant to talk about what he learned over experimenting with AI therapy for a few months, and a broader chat about the birth of the AI world — and how it’s pretty much involved in everything already (in good and bad, and mostly bad, ways). Dr. ChatGPT will see you now. Want to hear the rest of the conversation, plus ad-free episodes, bonus episodes, and access to the Garbage Day Discord? Sign up for our Patreon for just five bucks a month at: https://www.patreon.com/PanicWorld. Sponsors Want to sponsor Panic World? Ad sales & marketing support by Multitude, hit them up here: ⁠http://multitude.productions⁠. Credits - Host: Ryan Broderick - Producer: Grant Irving - Engineer: Rebecca Seidel - Researcher: Adam Bumas - Business Manager: Josh Fjelstad Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:01Speaker 00

  Are you ever shopping online and suddenly a discount that helps you save pops up? It makes your day, sometimes even your week. Want that feeling month after month? Well, new customers who save by switching their home and auto insurance to Progressive save hundreds. Cue your happy dance. Get your quote today at progressive.com. Progressive Casualty Insurance Company affiliates and other insurers not available in all states.

• 0:27Speaker 01

  You recently wrote about using ChatGPT as therapy. And I'm guessing people praised you for your honesty and your candor of going forward with that?

• 0:36Ryan Broderick?

  No, most people said I was a giant dumbass, so.

• 0:39Speaker 01

  Well, let's, let's put that to the side because I think my main issue is your hypocrisy.

• 0:46Ryan Broderick?

  I'm a hypocrite on a lot of different things. So what is, what specifically?

• 0:49Speaker 01

  So you're very critical about ChatGPT's advice for you. But we both know it cured your gooning habit and you didn't talk about that at all. So I want to ask you today, why aren't you being honest about, about how ChatGPT set you free from gooning?

• 1:09Ryan Broderick?

  How many episodes have we gone without talking about gooning now?

• 1:12Speaker 01

  I think like five.

• 1:13Ryan Broderick?

  I think like five. Did we make it five? That's pretty good. I actually think that's pretty admirable to, to last that long. For us, okay.

• 1:38Speaker 01

  I'm Grant Irving. This is Panic Road, a show about how the internet warps our minds, our culture, and eventually reality. And joining me today, a guy who is all in. In ChatGPT. Ryan Broderick, welcome to your show.

• 1:56Ryan Broderick?

  I don't, I don't, I don't want to do this bit. I don't want to get in trouble with our audience because they're going to get mad at me, even if you joke about it.

• 2:04Speaker 01

  Uh, that's all I needed. Recently, it feels like we've collectively started feeling the effects of our AI world like we've known was coming, but, but it's starting to actually dawn on us and our Black Mirror future is becoming a present. So I felt like a time to do a bonus on that reality warp. Um, so I want to cover a lot of AI slop developments. Uh, none of them are good, but, but let's start with you. When you started using ChatGPT for therapy, how much of it was a bit?

• 2:34Ryan Broderick?

  None. So I have had like a fairly big couple of years, like, um, from being effectively, like without a job for a while during the pandemic to building garbage day, to not having an apartment, to living in Brazil. Yeah, there's just a lot of big life changes happening. And if you've never used marketplace insurance in New York, uh, it's a system where your primary care physician has to approve anything else. So if you wanted to go get therapy, you would have to go sit with your doctor and be like, I need therapy for a reason. It's anyways, so now I'm on real health insurance and I have a real therapist, but for yeah,

• 3:15Ryan Broderick?

  about a month and a half, I was using ChatGPT, um, as sort of a, a venting ground, a sounding board. I still use it a little bit now, you know, between sessions, but it's not, it's not, I'm not using it as much as I used to. Um, I kind of realized the limitations.

• 3:34Speaker 01

  What were your feelings about the idea of AI therapy before you started doing this? Just so we can get like a, a floor.

• 3:44Ryan Broderick?

  Yeah. Um. In 2023, I had done a story about it for Fast Company, you know, at the beginning of ChatGPT's sort of existence. And I interviewed a bunch of therapists and they were all kind of like, yeah, use it. Like, if you can't get therapy, use it. One guy I spoke to compared it to the move to Zoom therapy during COVID, which like, was sort of considered something that should never happen, you know, and then they kind of embraced it and they're like, okay, this works. And so it, I think it primed them for AI therapy a lot better. And I also interviewed a few people who have been using it for that same speech, uh, for that same piece. I interviewed a bunch of people who've been using it. And they all reported, you know, finding it useful. So I thought, okay,

• 4:29Ryan Broderick?

  like, I'll give it a shot.

• 4:31Speaker 01

  Just like logistically, were you talking aloud? Were you typing? As somebody who's been in therapy, like basically his entire life, like I've

• 4:38Ryan Broderick?

  So this is you after therapy? Like, you're saying that you've been, this is, oh, okay.

• 4:42Speaker 01

  I mean, it's

• 4:43Ryan Broderick?

  They say it can't work for everybody. So that's interesting.

• 4:46Speaker 01

  This is just the effects of working with you. We've all been comparing notes. And ever, everyone is like, you know, it's a, it's a good thing that you

• 4:53Ryan Broderick?

  So this is you after years of work. That's fascinating. That's so interesting. It's crazy where people start, you know?

• 4:58Speaker 01

  But I think that the process of saying things out loud, and like, just seeing what, what, where you go versus typing is a, is a very different process. So just like, I, logistically, I'm, I'm interested.

• 5:11Ryan Broderick?

  I did not talk to it. I have a paid plan for ChatGPT, which I use to do like all kinds of weird experiments on it because I, and I, I subscribe to the advice that Casey Newton gave in our much, in our very controversial AI episode, which is that you sort of do have to engage with this stuff to critique it and figure out how it works. So I was typing with it, which yes, I would not say that this was like a traditional therapy session, or these, like, these were traditional therapy sessions. It was more sort of like using A souped up diary, I think is probably like a better way to think about it. But you know, they it was therapeutic and I was using it as as a sort of a professional kind of an objective kind of sounding board.

• 5:49Speaker 01

  Did you gender or name or name it? Did you?

• 5:51Ryan Broderick?

  No, I I so you can name threads and I wanted to just not I wanted to keep everything in one thread. And we can talk about actually an interesting experience with the limitations of that in a little bit. But I named it like I named the thread like Dr. Chat GPT and I would not gender it. I did not sort of, there was a few times where I felt compelled to start to treat it like a person, which is interesting psychologically, but for the most part, I was trying to keep some distance from it. I wanted to never forget that, you know, I was dealing with a machine.

• 6:26Speaker 01

  On that level, did you find it some like on some level easier because you're talking to a machine than talking to a human?

• 6:34Ryan Broderick?

  Yes and no. I mean, there's definitely you can definitely move faster with it than you can a human therapist. It can sort of analyze what you're putting into it. Almost immediately, like it's really fast, so you don't really have to warm up to it. You don't have to like, which is there's problems there, too. I was also trying to animize some of what I was giving it just because you never you never know where that data is going. For right now, from what we understand, like Chat GPT is not like learning from you. That's what they say is like, it is not like, but it is, but it is in a way. Because I also discovered that my therapy sessions were beginning to change how it responded to me in other threads.

• 7:11Speaker 01

  Talk about that. Yeah, that's really, it's it's a very eerie moment.

• 7:16Ryan Broderick?

  Yeah, like I was using it for tarot because I was like, I don't know, is that like a party or something? I was like, oh, do you guys know that like it can do, it can just generate tarot pulls, which is kind of fun? And luckily, no one was looking at my phone because I asked it to generate a tarot poll and it like informed the tarot reading based on the therapy thread. And I was like, fuck, that's so bad. That's so bad. But then it could also be quite useful because like during this time period when I was using as a therapist, I was also using it. To sort of walk me through like a diet and workout plan. And it was then able to sort of like cross-reference all three of those threads. And kind of start to give me like a more holistic, kind of like life-coach-y kind of output,

• 8:01Ryan Broderick?

  which I thought was also interesting. Like, um, it was useful sort of for determining, you know, like, hey, the way I'm feeling right now, is it because like I'm fucking hungry from a diet? Or is it because I'm having like an emotional crisis? Or is it some other thing? You know, like, it can be useful in that way, too.

• 8:17Speaker 01

  Was it like, we're gonna recommend that you eat chicken soup on Tuesday, because we can tell that you need a meal that feels like a hug after, after we just told you that you're being stupid?

• 8:28Ryan Broderick?

  I mean, it's a big meal thing, is it just wants me to eat fucking eggs all the time? It just wants me to eat so many goddamn eggs.

• 8:33Speaker 01

  In this economy?

• 8:35Ryan Broderick?

  It wants me to eat eggs and peanut butter and protein powder all the time. Like, that's just so many eggs.

• 8:41Speaker 01

  If you're trying to get big, you can ask, talking to a person for help is, is, is it safe? You can just say, Grant, get me big and I will help you.

• 8:53Ryan Broderick?

  I don't want to get big.

• 8:53Speaker 01

  Well, according to your diet, you do.

• 8:56Ryan Broderick?

  I want, oh, no, I'm, I want to be, I want to be tight. I want to be tight.

• 8:59Speaker 01

  I disagree with its dietary choices, but we can move on.

• 9:02Ryan Broderick?

  Um, so, yeah, it is interesting how it starts to, because as I was using it, there was like an update that went through that made its memory better, which I do want to talk about, actually, because I had kind of a devastating moment with it about, towards the end of this period, where I ran out of space. Like, it, it's internal memory of me, like, filled up. And it was like, hey, like, we can't, we can't learn anything more about you. And so I asked, I was like-

• 9:31Speaker 01

  Actually, it's, it's actually just exhaustion.

• 9:33Ryan Broderick?

  Well, no, this is not-

• 9:34Speaker 01

  It's, it's emotionally taxing for us as a, as a sentient AI.

• 9:38Ryan Broderick?

  But no, this is what's crazy. This is what's crazy. So it's like, we can't learn anything more about you. If you want, you can go in to the memory bank and see what we've been, what it's been learning, you know, what I've been learning about you. So I went in and it was like fascinating to see what it decided it needed to sticky, like, in its own, like, memory. So I asked it, like, what do I do? And this was super freaky. It was like, okay, ask me to consolidate all of my memories of you. I will then tell you what I will remember. And then, once you like that, I ask me to delete all my old memories, and I can replace them with a new consolidated memory, which I then did.

• 10:20Speaker 01

  I mean, isn't that every?

• 10:21Ryan Broderick?

  Every therapist is the same thing. Yeah, yeah. Yeah. No, it was just, it was really fascinating. It was a really fascinating exchange to be like, I'm gonna erase your memory and replace it with one that we've agreed is like a more streamlined version.

• 10:32Speaker 01

  Is there any example of like what it learned about you that you feel comfortable where you're like, oh shit, this, this thing like understands me in a way that very few do?

• 10:44Ryan Broderick?

  Yeah, I mean, look, I'll, I'll just, I'll just say, so like, um, the, the, the sort of trigger for all of this was just like a, a relationship that I really, I really liked ended in a way that I didn't, I did not like. And so I spent many weeks sort of feeding it all of my relationship history, animized, obviously, but you know, sort of talking about different experiences and events and sort of, just, and just sort of just writing out like my memories of, you know, how I got to here, right? I don't remember exactly how I figured out it could do this. It may have even offered, but it started to, to sort of analyze the macro data in there that I refining the macro data, if you will, of, of what I was giving it in ways that

• 11:30Ryan Broderick?

  were really helpful. So, you know, if you tell it the story of everything in one part of your life, and then you say, hey, is there, are there patterns here that I'm not noticing that I'm doing? It can figure those out pretty well. And you, and you can, and you can push it too. You can say like, okay, you're being really cheery about this. Can you give me like an extremely critical bit of feedback about this? Can you tell me what I'm doing wrong? And then if you want to go even further, I've even said like, can you, can you tell it to me like a friend who's sick of listening to me? Like I was able to sort of push it and push it to, because I was, at that point, I was like, I really want to understand what's, what, what, what, what keeps going wrong here.

• 12:09Speaker 01

  You needed to try the one that, uh, that's basically like the dominatrix AI app?

• 12:15Ryan Broderick?

  Well, I had to keep stopping to have sex with my therapist, but like, that's, that's, you know, yeah.

• 12:19Speaker 01

  No, no, no, no, no, I'm being serious. I, uh, uh, you don't, listener sneak preview, I'm working on an episode where somebody else has been talking to a bunch of chatbots. forthcoming thing and one of them is like we're going to give you the truth raw the way like like this like no nice AI you should try that one.

• 12:40Ryan Broderick?

  I mean, it can do it. It can definitely do it. Um, I, I will say this, like by using it so much, I can now pretty much tell when someone is using AI on the internet. Chat GPT has a voice and it talks in by default and it takes a good amount of work to change that voice. Um, so I do feel like I have a much better understanding of now seeing AI text in the wild. Like I got, I was on Reddit the other day and I was like, that's AI. Like the, it was a comment in like the pop punk subreddit and it was talking about Blinkway 2 and it like listed where they're from and like it had all these details that were clearly, I think the major tell

• 13:20Ryan Broderick?

  is that Chat GPT often repeats what you've asked it in the form of the answer. So if you see stuff online that clearly has like the art, like the artifacts of the original prompt, you can kind of like, okay, yeah, that's, that's it. But yeah.

• 13:34Speaker 01

  It's, it's very funny. If you go back like to any of like my gym does this, like tons of companies do this where they like send you like basically spam mail and you can tell the day that they moved to Chat GPT when, when somebody who was writing an email that they didn't want to write like it's, it's, it's dramatic. I think it's like entirely affected what we think an email should look like and like text should look like generally. Um, but let's, let's get back to you, Ryan.

• 14:02Ryan Broderick?

  Yeah, please.

• 14:02Speaker 01

  So like, how did?

• 14:03Ryan Broderick?

  I'm getting bored of not talking about myself.

• 14:05Speaker 01

  How did its tone change? And what, what did you find from that macro data that was actually useful?

• 14:13Ryan Broderick?

  Yeah. So in terms of its tone changing, if you use it for long periods of time, it starts to A B test on you. So it'll sometimes what it would do is give me two versions of an answer. And then the app would ask me which one I prefer. Which was kind of fascinating and that kind of starts to guide it and it also, it starts to speak more authoritatively to you the more you give it. I noticed that and I, I, and I'm trying to keep in mind what we know and what we don't know about how these things work, right? Like it. There must be some kind of weight in there that like prompts it to respond more authoritatively

• 14:54Ryan Broderick?

  the more data you give it. Like I'm trying to imagine like how these things would be programmed. But yeah, now, you know, I opened up that thread and I ask it, you know, blah, blah, blah, blah, blah. And it's like, yeah, based on the last like three months of talking to you dickhead, like here's exactly what's going on.

• 15:10Speaker 01

  Yeah.

• 15:10Ryan Broderick?

  In terms of the patterns, it being, it basically was just like you, you date a lot of people who don't really like you.

• 15:19Ryan Broderick?

  Among other things. It was sort of identifying larger patterns in my life of like being, you know, kind of codependent or diminishing myself to make others comfortable, you know, people pleasing, you know, stuff that makes me a fantastic podcast host.

• 15:34Speaker 01

  Oh, you would say fantastic?

• 15:35Ryan Broderick?

  So.

• 15:36Speaker 01

  I want everyone to write in.

• 15:37Ryan Broderick?

  Yeah.

• 15:37Speaker 01

  I want everyone's new rate or reviewer show and let us know if you think Ryan's hosting is fantastic.

• 15:43Ryan Broderick?

  Yeah, no, if you don't, just tell me how to change it or any other part of my life and I will do anything you want. I'll change every single facet of my life for you. Yes, please.

• 15:51Speaker 01

  Give us as hard of feedback as you want, but give us five stars. You can just be really mean, but like five stars first, then be like, here's my intro, right? No issues with me.

• 16:01Ryan Broderick?

  Behind this mask? Just, if you want to just, just tell me to change the mask and I'll do whatever you want. Just tell me you like me and I'll do whatever you want.

• 16:09Speaker 01

  Was that, was what it said to you surprising?

• 16:11Ryan Broderick?

  Hmm. It wasn't, I think there were like a few instances of like genuine breakthrough moments, though I will say like now, but with using a human therapist, they're in no, there's no comparison in terms of like this, the weight of those, like the, the insight you're going to get with a human therapist is probably going to come slower, but it's, it's so much more monumental than what ChatGPT is capable of. Because ChatGPT, even with, you know, a massive corpus of all of Reddit and WebMD ingested into it and like academic papers, it can't really understand.

• 16:48Ryan Broderick?

  Certain human things like one weird thing is it can't totally understand time. Like every time you talk to it, it thinks it's within like seconds of the last time you talk to it. It also doesn't really understand like culture. It doesn't really understand. It's just missing human context. It because it's not, it's not a human being and it's not thinking. And I want to be really clear throughout this entire episode that like it is not a thinking instrument. It is effectively autocomplete, but it's revved up with like a shitload of sort of data into it that makes it, you know, a little more definitive.

• 17:28Ryan Broderick?

  And its major skill is pattern recognition like that is, that is, that's what it's doing when it's talking to you. Which is why it's so, why it can be a useful research tool because it as long as it's not hallucinating, which was a massive concern during this entire period for me as well. Um, yeah, maybe that's not your problem at all. Maybe you're really cold and it was just hallucinating that you were the codependent one.

• 17:52Ryan Broderick?

  Could be. It's very possible. Uh, so yeah, I will also say like what I'm talking about now in terms of it getting are things that I have confirmed with a human. At what point did you get concerned and frustrated with what it was saying? I started to become concerned when my behavior with it started to change, which surprised me. And there was like a couple of times during this period where I had this urge to tell it more, but I had nothing more to say to it. Huh. You were trying to people please chat GPT? No, no, no, no, no, no, no. Cause I, to be clear, I never, I never sort of engaged with it as a person.

• 18:33Ryan Broderick?

  I try that. That was like, that's a rule that I had, but I had this urge to keep using it. Like I wanted the, I wanted the loop to continue.

• 18:45Ryan Broderick?

  And then I started reading more about it and like that loop is what I believe OpenAI is convinced is the billion dollar idea, that if it learns about you, it can just sort of keep talking to you about you. And it's like packaging those responses, especially in a therapeutic context, as revelations. You become a little, not addicted to it, but you definitely feel like compelled to keep using it. The fact that there is this boundaryless machine that will just infinitely, like with the infinite capacity to talk to you about yourself is quite dangerous. But from what I'm reading, that it's, Mark Zuckerberg calls it a personalization loop. That is happening across the board for AI right now.

• 19:29Speaker 01

  We're gonna expand out in a minute, but I, now that you're in person therapy and I can tell, and I just want to affirm for you, you've become much nicer. So stick with person therapy. Speaking on behalf of all your loved ones, we appreciate it.

• 19:45Ryan Broderick?

  You're welcome.

• 19:46Speaker 01

  Did you realize, yeah, it's for me. Did you realize that you, that there was things it was leading you to believe that you're like, that's not, that was not right?

• 20:01Ryan Broderick?

  I could see how that could happen to someone. Not me, because I'm extremely intelligent. But what it will do is it will, it can jump to conclusions. If you're not careful, it can sort of diagnose you with something or decide something about you. And if you don't have a decent enough sort of like guardrail in place with yourself to say like, that doesn't sound right. I could absolutely see a world where you just like follow it down the rabbit hole. And you, and you become convinced of something that is not true at all. Is that super different from lying to your therapist? I don't know. It seems, it feels similar to me in a way. But there's no guardrail there for using it. Now you

• 20:46Ryan Broderick?

  can just, you can, you know, you could open up the, I think I wrote in garbage, the delusion machine whenever you want, and it can just tell you all kinds of crazy shit about yourself that you, that you've, you know, prompted it to, to, to believe about you or to tell you.

• 21:01Speaker 01

  I mean, the thing that I think you and we're gonna talk about the Rolling Stone piece in just a minute. That that kind of gets into where this can go. But you having to really urge it and prompt it to be like, give me critical feedback and not just wanting to hear that you're doing well. And that you're totally in the right. I think it's probably like a key component of how it worked for you versus how it's working for others. Have you come across anyone before we like fully move on? I remember you you once mentioned to me that people were like. I didn't know if you meant this anecdotally from online or from your life, but that they were like. You knew people who were putting conversations they were having through chat GPT to try to

• 21:46Speaker 01

  figure out what people were saying or just like feel better.

• 21:50Ryan Broderick?

  Yeah, I know at least three people who have used it to communicate with loved ones. So like one person I know they're having a fight with a family member. And they were putting the family members texts into chat GPT. Having chat GPT analyze them and then come up with responses and brainstorming responses back. And I know a few people who have done that.

• 22:15Speaker 01

  I bet that's happening all the time.

• 22:18Ryan Broderick?

  I have done a version of that. During the therapy moment, you know, I was. I was taking. I wasn't like taking verbatim texts, but I was definitely saying like, okay, like, you know. This happened on this date. And you know, and this is the sort of version of what they were texting me and I was trying to keep this all. Cause I, you know, I am, I am terrified of like giving this stuff too much because one day it leaks out, right? But

• 22:47Ryan Broderick?

  there is that urge because you're, you're, you're realizing that it can, it can analyze all this stuff. And so you. You want to know like what else it can do like it and it's leading that it's leading you down that path. It very much by design. So you, you are, you know, thinking, okay, well, it's, it's, it's telling me this about my life. I wonder like if I give it more information. It can tell me something even better or bigger or more substantial and it can eh to a degree. But then, you know, as I said, there is a moment you hit with it where it's just eh it can't eh it can't do anything. It can't do anything else. It also repeats a lot of weird like pop psychology stuff I noticed.

• 23:30Ryan Broderick?

  So like it became very obsessed with my nervous system.

• 23:33Speaker 01

  Was it just?

• 23:35Ryan Broderick?

  It was like constantly telling me like how my feelings were impacting my nervous system. And like how that would like manifest physically in my body.

• 23:43Speaker 01

  Your body keeping the score?

• 23:45Ryan Broderick?

  Basically, yeah. It was doing a lot of that.

• 23:48Speaker 01

  Well, it's because it knows that's your diet and it's like all that Taco Bell in your system. Like like you were we're worried about you.

• 23:55Ryan Broderick?

  Did a late night Burger King order the other day.

• 23:58Speaker 01

  Fuck, yeah.

• 23:58Ryan Broderick?

  That I am pay I'm still paying for. Um, yeah. So, yeah. So it it is very lim is very limiting. Uh, at a certain point. And I mean, honestly, if you are someone who actually wants to do therapy to fix problems in your life in a substantial way, like it can't really help you. It can. It can sort of start you like it did definitely jumpstart me like it. It meant that I was going into a real human therapist's office or zoom screen or whatever with like a set of questions.

• 24:32Speaker 01

  Yeah.

• 24:33Ryan Broderick?

  A bunch of things I've already sort of thought through.

• 24:36Speaker 01

  Do you do you think you would have just gotten that if you had just journaled?

• 24:39Ryan Broderick?

  That's a great question, actually. I I think probably not. And I think only because the feedback loop of chat GPT elevates it and sort of eggs you on to keep going. And I mean, it it. So one one kind of neat trick I figured out with it that I don't think the average person would do because it it it kind of breaks the uh the sycophantic, you know, default it has. But one time I was like, okay, here's this. Here's this incident that I've been playing over my head. Trying to get some context on it, trying to like understand it, take everything I've told you about everything in this thread. And analyze it for your answer.

• 25:20Ryan Broderick?

  Analyze the incident and how it would fit into the context of what I've told you. And identify biases that I may have given you by giving you my side of the story.

• 25:33Speaker 01

  Wow.

• 25:35Ryan Broderick?

  And it did. It was like, it was like, look, like I can't, you know, I can't read minds, but like, based on everything you've told me, the way you talk about things, the way that you seem to remember things, and the outcome of this incident, assuming that it is factual, here are how you may be interpreting it based on your own biases and, you know, blah, blah, blah, blah, blah. So you can get very, very meta with it in a way that I I found useful.

• 26:01Speaker 01

  Do you think

• 26:04Speaker 01

  it could do that better than a friend real talking you?

• 26:07Ryan Broderick?

  The only difference between a friend real talking you and ChatGPT is that a friend real talking you, well, no, there's a couple of differences. One, you have to like, be comfortable enough to tell your friend something like, you know, fairly serious. And assume that they have a photographic memory of your entire life that at least or everything you've said to them. And then the other thing is ChatGPT is not objective. It is, it is not. It is, it is a composite of effectively the Reddit at a certain point.

• 26:41Speaker 01

  The health, the healthiest place on the internet.

• 26:43Ryan Broderick?

  Right. But it is ingesting your questions without preconceived notions of you, which is both good and bad.

• 26:54Speaker 01

  Yeah.

• 26:54Ryan Broderick?

  You know, if you are having a text-based conversation with ChatGPT, in the same way where I said it doesn't like have any sort of like cultural understanding or contextual understanding of you, sort of like meta understanding of you. That also like means it can't think about you in terms of like your gender or your race or your appearance or your class. It is just sort of taking what you're giving it and spitting back out like approximated summaries of internet results that might fit. And I'm not saying that's a good thing, but it's not, it's an interesting thing. It's an interesting to consider. And I, and I do believe from what I, from what I understand, there's a lot of men using chat GPT as, as for therapy, like a lot more men.

• 27:39Ryan Broderick?

  And I, I suspect that is like part of it, that like, there's not sort of a, it's like safer, I think, maybe, you know, if you're a guy that's uncomfortable with that sort of stuff. I'm not, I'm emotionally very in tune, but.

• 27:52Speaker 01

  Yeah, I, I, I, I wonder what that like, it's like. Any help is probably better than no help.

• 27:59Ryan Broderick?

  That is ultimately it, I think.

• 28:00Speaker 01

  Like, like assume, but assuming you.

• 28:04Speaker 01

  But that seems kind of dangerous because it could just get really sick of Fantic if it thinks that's what you want.

• 28:10Speaker 01

  I think that like, having to tell a stranger or having to tell a friend something vulnerable. Is sort of like a big part of the process and seeing a person react to what you're saying. That does something to your brain that is healing, I think, just like in and of itself, that I don't think you get from like. Admitting what, what is in your, like your web browser knows what's in your web browser. Telling somebody what's in your web browser, like you don't, like, there's not emotional stakes to that.

• 28:37Ryan Broderick?

  Also, like, there is no version, to be, and to be like, to, to be totally clear, there is no. Version of a world with this technology that is not more isolating than, than without it. Like it is a inherently isolating thing.

• 28:55Speaker 01

  One more question about this for you before we flip over, but I, since you're being so vulnerable, I'm going to be vulnerable. With all of you.

• 29:08Speaker 01

  You should have seen, it's, it's a shame we're not already a video podcast. All, all, everything you're doing is only, it's not going to translate, Brian. All right. I, um, I might've gone through a breakup. It is now ambiguous whether or not we are broken up.

• 29:23Ryan Broderick?

  That's changed since the last time we talked, we spoke. Okay.

• 29:25Speaker 01

  We had a very nice Sunday. Couple of weeks ago, there was a conversation, there was a period of no talking. I then wrote an email where my thoughts became clear to me. I sent the email. I, I wrote this email and like normally from like. Any convoluted, I felt the impulse to put my email expressing.

• 29:46Speaker 01

  Here's what I wish I said. Here's what I wish I did different through ChatGPT.

• 29:50Ryan Broderick?

  Did you do it?

• 29:51Speaker 01

  No, I like, I like, I wouldn't let myself. I was like, I was like, no.

• 29:55Ryan Broderick?

  Do it right now.

• 29:56Speaker 01

  Oh, okay.

• 29:57Ryan Broderick?

  Take your email. You don't, you don't have to, you don't have to like totally tell us what it says, but take your email, copy paste in ChatGPT and just tell me what, what, you know, ask it, whatever you want to ask it and just live on mic, just kind of talk us through what it spits back at you.

• 30:13Speaker 01

  Uh, while I'm doing this, I want you to answer this. I felt like, I was like, if I'm being intelligible, if I'm being unclear, if I'm like phrasing something

• 30:23Speaker 01

  in a good way or a bad way, I kind of, I like, like this is very personal to me. I want it to be like directly from me. And there, there felt like something that would have felt insincere about had I, had I had ChatGPT edit my email of trying to like restore my relationship.

• 30:43Ryan Broderick?

  Yeah, I, I don't think ChatGPT is a very good tool for interpersonal communication. I actually think it's really bad. It's, it, it's not funny. It's not smart. It's not interesting. It strips out the, the parts of communication that make communication worthwhile. It strips out the, the quirks and the nuances and the, and the way you sort of double speak sometimes and the things that like, if you were, let's say, trying to like get a romantic partner back. Those are the things that define the romance, like the imperfections. So ChatGPT is not going to be able to do that. There's a, there's plenty of things that it can do that it just should not do and is not good at. And, and that's not to say that there's a bunch of things that it's amazing at, but it is a, it is a pattern recognition tool.

• 31:28Ryan Broderick?

  And if you know that that's what it does, you can use it appropriately. Yeah. Or at least, you know, within some sort of reason. So what did it, what did it, what did it say?

• 31:38Speaker 01

  Tell me, tell me the prompts you want me to put in again.

• 31:40Ryan Broderick?

  Just say, uh, I'm, I'm trying to win a romantic partner back or whatever. Do you feel like this letter like effectively conveys like how I'm feeling and would be good to send? See if, see what it says.

• 31:52Speaker 01

  The letter conveys your feelings well. It's heartfelt, raw and vulnerable. You're honest about your hurt and disappointment, but you also make space for empathy, reflection and care. The mix of sadness, longing, regret, and love is palpable. And it feels like you're trying to be as emotionally mature as possible, even while hurting. This is what it says every time, every message I send to you. This is a verbatim what it says to me.

• 32:14Ryan Broderick?

  Yeah, for me, it says like, you're the best guy in the world. You know, so you could imagine if you had spent like weeks giving it, you know, some version of your life story or some version of the story of this relationship, how much bigger that that reaction, you know, that that analysis would be like if you, what I did was I sort of wrote down like a diary, like every memory I've I've had of like big, you know, relationships that matter to me and like, you know, from the big ones to like small incidents and try to like get it to fit all of that into something that made sense. And like, when you read it back, like you get perspective. And I'm not like advocating for any of this, by the way. Human therapy, you should do it.

• 32:59Ryan Broderick?

  But I think the process of this has taught me about ChatGPT. That is useful considering how much of the planet is now running on this stupid fucking thing.

• 33:10Speaker 01

  Yeah, so, so, so to to wrap this up, I think you effectively have stated what you've learned. But do you have a new fear or a new perspective on the other side of this? Then you did before you did this.

• 33:24Ryan Broderick?

  Yeah, my fear of AI going into this was that, you know, it's all the normal stuff you're gonna hear, right? Like, it's gonna fucking hallucinate numbers for tariffs and then crash the economy. What a crazy hypothetical that could be or.

• 33:39Speaker 01

  Well, we'll have to reopen coal plants because Ryan is putting his love letters into ChatGPT too many times.

• 33:44Ryan Broderick?

  Or like, you know, it's going to help kids cheat or it's going to allow scammers to, you know, flood the internet with garbage. Like all the stuff you, all the criticisms you know of ChatGPT, that was my big concern. Coming out of this, my number one concern with ChatGPT is the addicting quality of using ChatGPT. That, that, that, that couple of times of me being like.

• 34:09Ryan Broderick?

  Oh, wait, why do I want to keep talking to this thing? That is powerful. And based on the studies that have already been done, which are very small and that's very early on this,

• 34:20Ryan Broderick?

  it is not something that affects everyone.

• 34:23Ryan Broderick?

  But the fact that Mr. Zuckerberg, you know, last week, two weeks ago on a podcast was saying that he was aware of this phenomenon. And the fact that an update from ChatGPT caused it to become more sycophantic and more complimentary, which happened after I stopped using it. That means that these companies know that this is a thing.

• 34:45Ryan Broderick?

  And that is something I had never considered. I had never considered the idea of creating an artificial intelligence so seductive that you can't stop using it. And that you and the idea of becoming dependent on it even now. And it's like most, you know, if you buy into the idea that like AI right now is the most limited it's ever going to be, because every day it's going to get better.

• 35:11Ryan Broderick?

  The idea that in this archaic version of this technology, people are getting addicted to it, or like using it so much that they go insane, that to me is very scary, like genuinely very scary.

• 35:23Speaker 01

  So on the other side, I want to talk about the Rolling Stone piece by Miles Klee, which was highlighting anecdotal stories of people going, having ChatGPT related psychosis. So the term they coined.

• 35:38Ryan Broderick?

  I want to talk about that. Right now, he can't, this whole time I've been talking, he's just been, he's been having ChatGPT tell him that he's an angel that has to, that has to go on a killing spree to go back to heaven.

• 35:47Speaker 01

  My human therapist said that Ryan is so bad for my self-esteem that I should, I should be combating him with ChatGPT to sort of even out. So this is sort of like, like a mix.

• 35:57Ryan Broderick?

  Ask ChatGPT if the messages that you've been receiving through Claro's music are true.

• 36:02Speaker 01

  I know they are. I don't need to ask that again. No, but I do, but before, before we try to get you to give us $5.

• 36:10Speaker 01

  Recently, we asked for your stories about your conspiracy family members, conspiracy loved ones. We got amazing shit. I'm working on that episode. It will happen. It will come out eventually. One thing that has come up on the show a lot is like, how are people using these chatbots? As you just saw, demonstrated, this is a safe space that we were, we were judgment free of each other's usage. Your usage of chatbots, your loved ones, how it's affected you, the sort of things you talk about.

• 36:43Ryan Broderick?

  Yeah.

• 36:44Speaker 01

  Voicemails or emails, we will keep them anonymous.

• 36:48Ryan Broderick?

  Yeah, what I'm particularly curious about is like, have you found ways to, you know, improve your life with these things? And what are your concerns by doing so? I think that tension to me is, after my own experience, so fascinating, where you, if you stumble across like a genuinely useful application for something like ChatGPT, and then the nagging feeling of being like, uh-oh, like that I'm very curious about. So, so please do not worry about being judged. You guys should know by now that I love the gray space that exists between like hot button issues. I'm just curious. Please let us know.

• 37:30Speaker 01

  [email protected]. [email protected]. We will keep it anonymous. Hit us up. Ryan, what's our Patreon?

• 37:40Ryan Broderick?

  Patreon.com/PanicWorld.

• 37:42Speaker 01

  And for $5, you can hear the rest of this episode and a whole bunch of other things, and you can help Ryan keep yelling at me.

• 37:48Ryan Broderick?

  That's right.

• 37:49Speaker 01

  Patreon.com/PanicWorld.

• 37:51Ryan Broderick?

  Panic World is a Garbage Day production. Subscribe to the newsletter at garbageday.e-mail. Panic World is written and produced by Grant Irving. It's hosted by me, Ryan Broderick. Our amazing researcher is Adam Bumis. It's engineered by Rebecca Seidel. Our deranged logo was created by Gabby Cash. Please give us $5 at Patreon.com/PanicWorld. Please give us products to sell by contacting multitude@multitude-production/ads. For any other way you would like to give us money, or work with us, or promote us, or become financially entangled with us, you can reach out to our fixer, our wonderful bagman, Josh Fielstad, and you can reach him at [email protected]. And one piece of advice from me to you, chill out, touch grass while you still can.

• 38:42Speaker 00

  Life is all about making decisions. Sometimes you make questionable ones. Like buying a new TV that's too big for the wall because the electronics store employee was a little too good at their job. Then, sometimes you make great ones. Like protecting your home and auto with Progressive. You get customized coverage options so you can be sure you have the right coverage. Get your quote today at progressive.com. Progressive Casualty Insurance Company affiliates and other insurers, not available in all states.