Detector Shows Basically Everyone's Using AI

0:00Parenting unlocks some pretty impressive skills, like carrying grocery bags, a soccer ball, and a sleepy kid all at once. Good thing the newly designed Chrysler Pacifica has a few tricks of its own, hands-free power-sliding doors and lift gate, rear seat entertainment, and still-en-go seating and storage that adapts faster than your family's weekend plans. Visit Chrysler.com to learn more. Chrysler is a registered trademark of FCAUS LLC. When fall comes back around, there's a sense of a reset. And new pieces from Lulu Lemons fall job are an inspiration to push your goals and elevate your routines. As your calendar fills, easy to layer styles meet the demands of your day, and as you challenge yourself to try even more, versatile sets transition with you. The Lulu Lemons fall lineup brings new levels of comfort and confidence, whether you're starting something fresh or starting something over. Check out the latest arrivals in store and on LuluLemon.com now.

1:04Upgrading your dinner party conversation, one show at a time. The last show with David Cooper. Welcome in! Here's what's coming up on the last show tonight. Getting older, your hair goes gray, your priorities change, and suddenly Bob from accounting is looking pretty good. Halfway through the hour, aging gracefully. What actually gets better with age? And why some long marriages and right around retirement? It's a phenomenon called gray divorce. Then after that, the most important award there is. Not the Nobel for physics, but People Magazine's Sexiest Man Alive Award. Yeah, we'll get into that stupidity next hour. Okay, lots to cover. Let's dive in. You've read an email, even a text from someone you're dating and thought, hang on, did this person actually write this? What about LinkedIn posts? My goodness, come on. That strange feeling that you get when you're not sure the person you're talking to is

2:06all there because they took that shortcut and asked Chad GPT to write back to you. But it's bigger than that. What about when authors lie? Blogs that we read. What about when their AI generated op-eds? Is anyone writing anything on their own anymore? Well, I am here with someone who used an AI detector for a month and what they found out from their peers' dates, even their heroes, horrified them. I'm here with Aki Ito, a chief correspondent at Business Insider. Aki, welcome to the show. Hey, thanks so much for having me again. This is such a great topic. What possessed you to basically snoop on everyone you know and all of their writing to see if they were using AI behind the scenes? You know, honestly, I think it was just this like almost a compulsive sense of curiosity. I just found myself asking myself pretty much all the time like, did AI write this as I was just going about in my life? And you know, I found out about this tool called PanGram, which is the leading AI detection

3:10tool on the market. And these tools were pretty bad for a long time. And so I guess I hadn't realized how good they've become. PanGram claims to be 99.7% accurate against all of the major models. And so I was like, okay, like, maybe I can come up with a story that will allow me to expense my $20 per month subscription to PanGram. That was the real inspiration. Can I get a free subscription from my employer? No, no. Before you did this, did you think you were good at spotting? AI writing? Because I think I am. I know that I'm probably not. But there's some tells for me the dreaded M- is one of them. For sure. The phrase quietly or the like quietly reshaping. I feel like every AI article quietly reshapes everything. But there's these little tells that we think mean someone use AI. But in fact, it might not be true. Did you think you were good at it? I did think I was good at it.

4:11And I will continue to claim that I'm good at it, even though most studies show that people are really bad at spotting AI. It does seem that there are some exceptions. Like, I think there was one study where if you are really familiar with these LLMs and like what they produce, then in that case, like you might be a little bit better than chance at spotting AI. But for the most part, people are really bad spotters. And it's funny because when you look at a lot of these studies, people often say that they prefer the AI's version over the human version if they don't know which is which, if they're given this blind taste test. But as soon as they find out it's AI, they immediately judge their peer, even though they might be using it themselves, I think is what one study study said as well. Okay. You thought you knew the tells. You're armed with this tool that claims to have a very, very good ability to detect this

5:12stuff. When you started applying it to the writing around you, what did you end up finding? Yeah. I mean, I guess the thing I found was that a lot of people use AI for their writing now, which maybe wasn't that much of a surprise. But I guess like just being able to see it so quickly as I was going about on the internet was really interesting. For example, pangram has this Chrome plugin where if you download it and you're on a social media site like LinkedIn, every post shows up with this label of like AI, human or mixed, mixed is when there's like a combination of human written and AI generated text. So that was just super interesting, like scrolling through my feed and then going to like different people's profiles and having pangram and just all of their posts and seeing, you know, who was a heavy user and who was not a user.

6:12And I found myself judging a lot of these AI users for sure. Yeah. Absolutely. So before AI like YouTube, comments in YouTube for whatever reason attracted the worst of the worst comments on Reddit wouldn't be quite as bad. What is it with certain platforms? In this case, LinkedIn attracting the AI slop of AI slop. Like the slop apocalypse seems to have come to LinkedIn posts first. Yeah. Yeah. I mean, I think like part of it is because LinkedIn is this like professional networking site and we're already using AI so much in our work. So I think that's part of it. I think another part is, you know, a lot of people post on LinkedIn not because they want to necessarily, but because they're trying to get a job or they're trying to like, you know, promote their own, you know, personal quote unquote brand of themselves as a, let's say like a software engineer or a marketer or something. And so there's kind of this like already like performative, you know, an authentic aspect

7:17to it. So maybe like that idea of like using AI to do this thing like nobody really wants to do anyway, makes a little bit more sense than on a site like Reddit, you know, where you're just like genuinely engaging in things. Fair enough. You're not really there to promote yourself. During this experiment, were there places where people use AI that surprised you? And yes, I'm team you up to answer a question about a breakup text that you analyzed. Yeah. So like, you know, as I was going about my experiment, at first it was just kind of fun. Like, you know, I compare it to waking up with a new superpower that could tell me who had gotten Botox. So it's kind of this like caddy, judgy thing. That's also really fun and gives me this new way of seeing the world. But then after a little bit, I remembered this, you know, breakup text that I received from somebody I was very briefly dating at the end of last year. And when I'd receive that text even in the moment, I had this inkling of like, oh, did

8:20this person use AI for this? There were two M dashes in the text and, you know, it just didn't sound like anything like her. And so, you know, I kind of, I guess I had that suspicion, but I didn't really have a way to check it. And now that I was going about using pain gram for about a month, I suddenly had this way to check it. So I put the text into pain gram and it came out as 100% AI generated. Very quickly, what went through your head when you thought that were you offended? Oh, absolutely. I mean, I was like, this is somebody, you know, I shared a connection with and the fact that they didn't even bother to, you know, sweat through the discomfort of rejecting me, that hit me hard. And like a true journalist, you actually contacted this individual and I want to get to what happened with that after the break. I'm here with Aki Ido, a chief correspondent at Business Insider, who experimented by using an AI detector on all the writing around her and the results were discussing.

9:24Hey, listen up, I want to talk about something. The phone rings, it's your bank or you think it is big mistake. That wasn't your bank. It was a scammer that bought your info from a data broker. There are hundreds of them that collect your information and sell it to scammers. That's why I use incogni. Incogni is an online service that sends removal requests to more than 420 data brokers and people search sites. Incogni gets it done and then watches to make sure they don't add you back in. Now incogni's claims are independently verified by Deloitte, the first data removal service to do that. You can't stop scammers from calling, but you can make it a lot harder for them. Incogni's already removed my info from 265 databases. Family plans cover up to five people. Now if you go to incogni.com slash Cooper, you can get 60% off. That's I-N-C-O-G-N-I dot com slash Cooper. The less they know, the less they can use against you.

10:25Incogni.com slash Cooper. Coming up after the break, why do we get so upset when we find out someone wrote something with AI? Instagram teen accounts come with automatic protections that limit who can contact teens. The content they can see and the time they spend on the app. Learn more about teen accounts and Instagram's ongoing work to protect teens online at Instagram dot com slash teen accounts. Here's something I can't stop thinking about. If a piece of writing moves you, informs you, makes you laugh, affects you deeply, and then you discover somebody used AI to generate it. Has the writing actually changed? Is there anything wrong with that? Well I'm here with someone who for an experiment used an AI writing detector on all the text around them. The results were discussing. I'm here with Aki Ito, a chief correspondent at Business Insider. Aki, welcome back to the show. Thank you so much for having me. Okay. You use this AI detector on a breakup text you received in the past and it came back saying

11:30that it was 100% AI. Very quickly you were offended by this. What was going through your head when you received that? I mean, I just thought like wow, like this person didn't even take the time to reject me personally. She just completely outsourced this to a machine. It felt terrible. I was a product she was putting back on the shelf. Then you, like a true journalist, I admire you for this. And fronted this ex about using AI. What happened then? Well, it was, if it were up to me, I don't know if I would have wanted to go back to her, but my editor told me I have to. You know, so she, to her credit, she responded right away and she was presumably honest. Did you use the AI detector on her response? I did. And it was 100% human. I had to. You know, she said she was really anxious about like, you know, these like five different

12:32initial versions she had drafted. And when she went to her co-workers to ask for their advice, they said like just plug it into AI. You don't owe anyone anything. And so she said that's what she did. So I mean, that must have hurt hearing the phrase, I don't know anyone, anything. When you're dating someone, you kind of do owe them treating them kindly. Yeah. I mean, yeah, it's, it was very telling. And the interesting thing I think about this is, you know, had I received like some version of this text when she broke up with me, I mean, maybe I shouldn't even say break up. Like we only went on three dates, you know, so it was a very, it wasn't a very involved relationship. You know, when, if she had sent me this like human text from the beginning, I think that would have given me so much more information than the, you know, chatchipity written text

13:34that she had sent me. The chatchipity written text had zero information in it. It was very vague. And also it said nothing about her because it was literally not written by her. The version, the human version of this text that she sent me recently for the story contains so much more information about what was going on in her head and about her as a person. And, you know, and so I think, you know, I was right to think I would have, you know, infinitely preferred to get the blunt human written text over the polite one from a bot. Now, your journalist, are there people you look up to that you ran this AI detection tool on and you found that the writing was written by chatchipity? Maybe it was like an op ed from a writer you like, and if you did, did that surprise you? I don't know if I found any like op eds myself. Um, the one, some that I did find were just linked in posts.

14:34Um, you know, from some, some people in my network, I do respect and, um, you know, in those cases, I did feel a little bit judgy, but as I write in my story, I found one of my linkedin posts got flagged for being about 50% AI, um, because I used AI to edit that particular linkedin post that I had initially drafted. And so, you know, I don't know if I was like really right to feel that judgey in that moment because, you know, it's not like I don't use AI at all in my own life. No, I remember talking to a researcher a few months ago about how people who use AI still judge their co workers when they, when they find out the co workers have used it, which is the crux of what I want to discuss it in our last, last few moments here. Um, like, and there's that study you mentioned before the break about people preferring AI written copy until they're told AI wrote it. Why do you think we judge other people so harshly for this?

15:38Why do we think we want a real human on the other end of the text, so to speak so badly? Yeah. I mean, that's the question that I wanted to answer when I started reporting out this essay, um, you know, scholars call this the AI disclosure penalty. And at first, I thought it was just this like irrational human bias, like this, like prejudice that we have against machines. Um, but I think like the more that I thought about it, the more I was into this experiment, and especially after I spoke to this, um, philosopher named Lindsey Brainerd, who's been writing a lot about AI and creativity, I started to think like actually, maybe we're on to something, you know, when we have this like instinctual distaste for AI generated writing. And it's because like, you know, writing serves a bunch of different purposes. It like provides people information, it entertains, but it's also this vehicle for us to feel connected to another person.

16:42Um, you know, we're trying to understand another person. And the way we understand that person is by reading what they choose to disclose to us in that writing. Once you have AI doing that, you know, like, AI has nothing to disclose because it doesn't have a soul. It doesn't have feelings. It doesn't have any life experiences to draw from. And so, you know, you're kind of like expecting there to be a human behind it. But when you realize that there isn't, that feels really hollow. You feel kind of duped. And I think that's what we're feeling when we realize that something was written by AI and we feel really disappointed. You know, it's the same kind of feeling when let's say a celebrity we love, we find out they have a ghost written memoir. We might feel a little like, ah, seems empty or hollow. And it's just so much worse because at least that was another person. You know, do you think this bias, this kind of negative view of people who use AI and writing once we found find out it's generated by AI?

17:45Do you think that'll go away because everyone's kind of using it? Even you yourself mentioned that you've used it to kind of edit or correct grammar and LinkedIn posts. And yet you judge other people's LinkedIn posts. Like do you think this is going to go away? This is a temporary thing or you think this will always be there. I don't know. I'm not entirely sure. If it were to completely go away, I would think that like the stigma, the bias would have lessened a lot more by now because, you know, we've had generative AI in our lives for almost four years now. And the, you know, the uproar over AI generated writing, I feel like the anger about that has only intensified over the last maybe six months or so. So I'm not convinced by the argument that this is just a temporary phenomenon. I think there's, you know, something real that we're hanging on to that said, I also hear from a lot of people that they just assume everything is written by AI now.

18:48Even the stuff that is actually written by humans. And that's a kind of cynicism that makes me really sad because that means that, you know, we're probably not like, you know, engaging with writing anymore. You know, we're not really like, you know, meaningfully engaging with people's communication. And that is this like miracle of a tool that we've had for millennia, right? This tool that we call language, it's like the oldest technology we have to communicate what's going on inside our brains to another person. And, you know, this kind of feeling that we're not really, you know, using that anymore, that we don't assume that that miracle is happening every single day. That makes me a little sad. That was really well said. I've had a lot of fun talking to you about your experiment. We've covered some fun stuff and some deep stuff too. I really appreciate you being on the show, Aki. Oh, thanks so much. It's always so nice to be on your show. You can read more about this topic in business insider.

19:51The article is called My Strange Horrifying Month Using an AI Detector on Everything I Read. And that's written by Chief Correspondent there, Aki. Aki, thank you again. I appreciate your time. Thanks so much. When fall comes back around, there's a sense of a reset. And new pieces from Lula Lemons fall drop are an inspiration to push your goals and elevate your routines. As your calendar fills, easy to layer styles meet the demands of your day. And as you challenge yourself to try even more, versatile sets transition with you. The Lula Lemons fall lineup brings new levels of comfort and confidence, whether you're starting something fresh or starting something over. Check out the latest arrivals in store and on LulaLemon.com now.