Jesse (1:18)

Wait, hold on. I'm setting the my days since horse joke thing up to like I think like 700 and now I have to set it back to zero and done.

Katie (1:28)

Okay. So Charles's first cinematic masterpiece was this film Zoo, about men who like to get penetrated by horses. And there's this very famous story. Charles did some reporting on this for the Stranger years ago about a man who got to death by a horse in Washington. If you ever visit. Definitely this is the thing to ask locals about. This is what they want to be known for anyway. So Charles is an experienced filmmaker and my expectations, I have to say, were sort of low for this film. Not because of Charles, but because of Aham. Everything we know about him. But. And I was pleasantly surprised. I think it was actually significantly better than the book. Adult Braces. There's some like funny laugh aloud lines. Are my enemy. Ijeoma Aham's sister actually plays herself in the movie. So despite this, despite the fact that my own personal bully was a starring player in this movie, I have to admit I actually sort of enjoyed it. But there were.

Jesse (2:21)

Wait, what's it called again?

Katie (2:22)

Thin Skin. How could you forget that?

Jesse (2:24)

Oh, Thin Skin we talked about 30 seconds ago. My brain.

Katie (2:27)

It's available for free on tubi and. But There were a couple of moments in it that. So it's the story of. It's autobiographical. It's based on this like one man show that Aham did. He was also on this American Life a few years ago telling his, his sort of memoir, his. His life story. And it's a pretty interesting story.

Jesse (2:46)

Can I just say it's really impressive to me that despite being neurodivergent, um,

Katie (2:53)

and, and not a man, that he's fathered two children.

Jesse (2:55)

Well, that's a, that's a separate issue. But he's neurodivergent and yet he's. Yeah, I'll just leave it at that. It's cool, it's inspiring because like he, you know, he must have to try.

Katie (3:08)

We love to see autist do well in the world.

Jesse (3:10)

Yeah, it must be really hard for him to do his numerous public appearances and speaking engagements and so on despite being.

Katie (3:18)

But does he make eye contact?

Jesse (3:19)

I. That's a good question. That could be it.

Katie (3:21)

Okay, so the movie, it follows this like pretty brief period of his life where he's like working some soulless corporate job. He actually did have a job at one point. And he's bitter because he wants to be just a full time musician. And his father, who abandoned his mom, married a Nigerian academic. They were together for like, married for like seven years and then he abandoned the family. It was an Obama situation. And so Aham and his sister Ijama were raised by this white woman in Seattle. And so the film just is about this like brief period of his life when he talks to his father for the first time. And it's kind of a shame because if you listen to the this American Life story, it's actually like the story of what happens is actually pretty interesting. It turns out that his father, surprise, surprise, had many other children in Nigeria. And one of them ends up coming to Seattle and I think living with Jama for a while. That's not in the story at all. But they do become close to this brother. The father actually dies before they can meet him. Sorry, spoiler alert. Let me do that again. Spoiler alert. The father dies before they can meet him. And so it only covers this one little brief period. But there's this moment in the story where Aham is talking to his father on the phone back in his. He's in Nigeria, Homs in Seattle. He's talking to him for the first time on the phone and he says, I'm a musician, that's my job. And the dad is like, that's not a Job, That's a hobby. Go get a job. And I look, I'm like, psychoanalyze. I've never met the guy. I'm totally psychoanalyzing from a distance here. But that to me seems like everything that we know about this, like, go back and listen to our story.

Jesse (4:54)

He's trying to, He's. He's trying to please his dead dad.

Katie (4:56)

No, not even that. It's like this. So if people haven't listened to our prima episode on what happened when. Aha. Sort of. What's the word? Lashed out at a reporter who, who, who wrote this Slate profile of Lindy West. Go back and listen to that. But he, he writes this email to this reporter and he basically is like, you denigrated my career. I'm a serious trumpeter. I have four. I'm playing four nights at the Beacon in Boston or the Paramount or whatever. And that seems to be this recurring theme where he. The fact that his talents as a musician. As a musician are not recognized or not lucrative seems to have created this, this like, anger on his part. You know what I'm saying?

Jesse (5:42)

Thin skin. He has like, thin skin.

Katie (5:43)

Thin skin. That's a way of putting it, yes.

Jesse (5:45)

Can I say that I have not heard any of his trumpeting, his trumpetering. I just think that if I did hear it, I would come away unimpressed. That's my prediction.

Katie (5:55)

You know, I think that it's like a little flat.

Jesse (5:58)

You know, he.

Katie (5:59)

Look, there's. I can totally. After seeing the movie, I can totally get why Lindy is obsessed with this man. Because there's nothing like seeing a man dump a spit valve out on a stage to get the blood rushing. It's just so hard.

Jesse (6:13)

I do think if Eddie, if there's like, look, I'm thin skinned in my own ways, but I actually think based on what we've seen of this guy, I'm gonna say guy. I don't. This is. If you emailed him and he were just. And you were just like, lowest possible denominator bait. And you're like, I've never heard your music, but I bet it sucks. I bet you'd get like a 10 paragraph response.

Katie (6:33)

But like, yeah, he would sign it with his full name and initial. The mom in the show, in the movie is a great character. She's this very quirky, irresponsible white lady who collects snakes and seems to fetishize Nigerians as this, like this beautiful, honorable people, even though her Nigerian husband abandoned her and their two kids. That, that character is really good. I'm curious what people think about this movie. Maybe I'm being too generous because I like Charles Mande Day and he made the movie. But yeah, we'll put a link to it in the show notes, people. If you watch it, let us know.

Jesse (7:04)

I just did. I literally just did. Great minds think like.

Katie (7:08)

All right, so that has been our Lindy west update. Jesse, anything you want to add about Lindy West?

Jesse (7:15)

She should break up with a hop. Someone did a really good version of, you know, the white hand and the black hand clasping hands in agreement. Yeah, someone did a version of that with like eight hands of every conceivable political affiliation all saying Lindy west should leave a hop.

Katie (7:35)

The only other thing to update is that Lindy did cancel two of her or postpone two of the stops on her book tour, which of course sent me into an absolute fever thinking that she was, you know, like she was canceling the entire book tour because of the backlash to the book. But she, her Seattle, her Seattle appearance is ongoing and I think she did one in la, so maybe she just had a cold.

Jesse (7:57)

Well, we'll keep. We'll keep tracking the situation for years.

Katie (8:00)

Yeah. If anybody, if anybody who listening to this goes to any of her book events, please, please, please record it. I need to know if she talks about the backlash.

Jesse (8:11)

Do you think she does the baby voice the whole time?

Katie (8:13)

She might.

Jesse (8:14)

I'm going to read you a wittle excerpt from my big book. Okay, this is enough. This is already, you know, even I am starting to retiring. It might be time to move on. Yeah. What are we talking about today, Jesse?

Katie (8:24)

Today we've got another AI episode and there's been a lot of chatter over the past couple weeks about, you know, I'm not even quite sure what to call it. Is it AI plagiarism?

Jesse (8:33)

I mean, if it's passing AI's work off as your own, that's plagiarism.

Katie (8:37)

I think it's plagiarism. I think that that counts. So we're going to discuss a few scandals there, as well as a controversy currently roiling media Twitter about how to use AI in our jobs or how not to. So, Jesse, let's start with the scandals. There have been a few writers who have recently been busted publishing AI generated text. Although I'm sure the number of people actually doing it who haven't been busted is much, much higher. The biggest scandal has to do with a book called Shy Girl. This is by an author named Mia Ballard. And I'm sure you've heard of this, right?

Jesse (9:06)

This was the self published one.

Katie (9:08)

Yeah, it was initially self published. This is Ballard's second novel. She self published this one in February 2025. It's a story about a woman who meets a man on a sugar dating website, then discovers that he's really into pet play, which I assume is belly rubs and fetch.

Jesse (9:24)

Okay, wait, pet play? Like what is pet? So dressing her up like a belly

Katie (9:29)

rubs and fetch, maybe some hide and seek every once in a while.

Jesse (9:32)

Okay, well now that I've, it's been explained to me, we can move on because that makes sense.

Katie (9:37)

So this book. So she self published it and it got big on TikTok, on Booktok. So it got around 5,000 reviews on Goodreads and the popular. And because it was popular, it managed to catch the eye of traditional publishers. This is of course the dream for self published authors. And I think, honestly I think this isn't a bad way to go for anyone who wants to write a book. Like I've talked to a few hopeful authors recently and you know, they say like, should I write a book? Can I get a publish? And my answer is typically no and probably not. It is very hard to find a publisher these days unless you, you have an existing audience. And even if you do get a publisher, you're probably be going, you're probably going to be doing most of the marketing yourself. But Ballard hit the jackpot. Self published to a big five Publisher Hachette UK bought the rights to publish her book and it did in the UK in November 2025. That's a very fast turnaround. It usually takes a year from the time a manuscript is accepted for it to come out. And this may indicate that they didn't do much at the publishing house in terms of, you know, editing the book or vetting the book, which is very

Jesse (10:36)

scary by the way. If you have an editor for a long and they're like, it's great, I just made a couple tiny changes.

Katie (10:42)

I love that.

Jesse (10:43)

No, you love that. That's terrifying.

Katie (10:44)

It's a dream.

Jesse (10:45)

Oh my God, that's such a bad sign because there's no way anything I write is good. So if they don't have thoughts, it just means like they're too busy.

Katie (10:52)

That's what I want. I want an editor to be too busy to make even a minor change. I want like the substack model is my perfect model because I don't like being edited. Do you like being edited?

Jesse (11:02)

This is long. Well, yeah, if it's good editing. But this has always plagued our dynamic as partners because I'm a perfectionist. I need every little thing to be perfect and I'm super organized. And you, on the other hand, just wing it and it just doesn't work for us.

Katie (11:15)

Yeah, that's, that's, that's the problem. Anyway, so this was set to be published in the US this spring, although there might have been some red flags from the beginning. So when Mia Ballard self published Shy Girl, the COVID art was stolen from an artist named Win Lewis, who paints whippet. So those are the dogs, not the high school club drug. She says that she just found this on Pinterest, which is, you know, this is something that people do. They see art on the Internet and they take it. Don't do that. As we have learned. Don't do that. It's illegal.

Jesse (11:48)

You would think that even someone just self publishing a book would know not to do that.

Katie (11:53)

Yes, but yes. Alas, unfortunately, there would be no happy ending for Shy Girl. So in early January this year, this is around the time when advanced reader copies were being sent out to the US a post started circulating on the subreddit Horror Lit. The headline was Shy Girl by Mia Ballard. Does anyone else think this was written by chat GPT? And the poster claimed to be a book editor of 12 years. And they start out by explaining the chat GPT tells. And we're not just talking about AM dashes, but bigger picture stuff. So the Reddit post, this redditor includes the opening passage of Shy Girl. So, Jesse, why don't you read this?

Jesse (12:25)

I wear a pink dress, the kind that promises softness and delivers none. It's. What's that word?

Katie (12:31)

Tulle.

Jesse (12:32)

T U L L E. It's like

Katie (12:33)

what's a tool is like what a what a what A tutu is made out of tool. It's like a plastic fabric, kind of. Plasticky.

Jesse (12:41)

I didn't know that word. Its tool is brittle and sharp, brushing against my fur like a thousand tiny teeth. A cruel lover that bites with every move. Every scratch keeps me in place, a reminder of what I am, a pet, a thing shaped for looking, for praise, for command. The bows on my pigtails pull too tight, yanking the skin and stretching my head into something neat, into something pleasing, a quiet violence made. Beautiful white socks climb my legs, their frills delicate, A whisper of innocence over the bruises beneath the ones he says shouldn't happen if the socks are there. But they always do this like. I would not have pegged this as. There's like. Like whisper of innocence is a bit. I Don't know, maybe hackish. But I, I'm not sure I would read this and that it would jump out as being a ish.

Katie (13:22)

What about you absolutely would not have pegged this as AI. I mean there, there are some tells with Chad, like EM dashes most obviously. And I was an EM dash user and abuser before Chat GPT. I've almost eliminated them from my writing because I don't want people to think I'm using an LLM. I'm actually kind of bitter about this. I genuinely like the EM dash and now I can't use it anymore. I've changed to the semicolon and these passages. So besides the fact that there's no EM dashes here, it doesn't have that. You know, this is really common in AI writing. It's not X, it's Y. That sort of framing.

Jesse (13:54)

Yeah. I just, the, the, this, the idea that there. These supposed tells will give it away. Like, I feel like that's a little bit 2024. Like you can very easily prompt it not to use EM dashes. Not to use. Or just like you can prompt.

Katie (14:07)

That doesn't. That actually doesn't work that well. I have asked ChatGPT before to not use em dashes as an experiment and it will reply with em dashes. It's like it can. It cannot seem to get them out of its system.

Jesse (14:19)

Okay.

Katie (14:20)

Of course I don't pay for the. I don't pay for the premium subscription.

Jesse (14:23)

I'm just a little bit worried that people think there's like super obvious tells when it's pretty easy to work around them. If you're a half decent would be AI plagiarist. So I just.

Katie (14:32)

We're going to get to one later that I think is a much more obvious tell. I don't think this one. I don't think anything in that paragraph. To me, nothing screamed AI to me.

Jesse (14:39)

Not me neither.

Katie (14:40)

And some people, some people said that this is, you know, is this like. People in the comments on Reddit were debating whether Shy Girl was AI or was this just an example of bad writing? I actually didn't think it was bad writing.

Jesse (14:50)

I don't think it was particularly bad writing. Maybe a little over. Yeah. But it seems like it's going to be that. That kind of book.

Katie (14:55)

Yeah. It was vivid.

Jesse (14:56)

Every scratch keeps me in place.

Katie (14:58)

It's a book about a human pet. Yeah. Yes. So I also, I fed it into that paragraph into ChatGPT and asked if it sounded human or AI, and Chad said that it sounded human. Except for one thing. Too many metaphors.

Jesse (15:15)

So you asked AI if allegedly, AI text was AI.

Katie (15:20)

Yes.

Jesse (15:20)

Okay.

Katie (15:21)

And AI said no, it wasn't.

Jesse (15:23)

Well, case close.

Katie (15:25)

So a week later, after this Reddit post started to spread, a popular YouTuber, she goes under the name Frankie Shelf, she released a very long video. It's 2 hours and 40 minutes. And it's called I'm pretty sure this book is AI Slop. So Frankie goes deep into this book, but. And one of the things that she notices is that there's a lot of repetition in the book. And Frankie compares this to a book called Rejection. So this is a novel by an author named Tony Tulithamute. I'm sure I pronounced that correctly. Don't look it up. It's a strange comparison, but she just. She, like, wants people to know how frequently Ballard uses these phrases and constructions. I'm again, I'm not sure this is a helpful comparison, but I want to play a clip. All right. So the phrase to this, to that, or to X to Y that I've read to you many examples of appears 25 times. It is not used once in rejection. Hum or humming as A description appears 26 times in Shy Girl. Again, 0 times in rejection. Something carved or being carved happens 21 times, typically as a description of something abstract. 0 times in rejection. The word gaze appears 41 times in Shy Girl, 6 times in rejection. Something is described as unrelenting or relentless, 28 times here, three in rejection. Rhythm or rhythmic appears 42 times in shy Girl and twice in rejection. The word low is used as a descriptor 51 times, 31 of those, referring to voices. In Rejection, it appears twice, once to refer to a voice. Gia's breath catches nine times something presses against her ribs, and 19 times her chest tightens 21 times. None of these things ever happen in rejection. Ballard was also a big fan of the word sharp, which occurred 159 times in the book, or a little less than once per page. And Frankie says it's often used in a phrase with. With another adjective. So sharp and jagged, sharp and terrifying, sharp and loud, sharp and quiet, sharp and grounding, sharp and unforgiving, sharp and endless, sharp and fast, sharp and electric, sharp and piercing. I could go on. And to Frankie, all this, plus the fact that the novel to her just feels very empty and very flat, is evidence of AI. But she does point out that it's really hard to tell. And she says she would not have thought that large chunks of this were AI generated if other people hadn't picked it up first. Here's a quote from Frankie. It's really hard to substantiate a claim of AI fiction writing because humans can be really bad writers.

Jesse (17:43)

Yeah. A lot of the tendencies people attribute to AI this is a broader problem. Like, there's just a lot of time. It's just stuff people do, too. And that includes bad or repetitive writing.

Katie (17:53)

Right. AI is trained on human writing. AI is going to pick up the idiosyncrasies of human writing.

Jesse (18:01)

Do you think AI is mad that that's what it was trained on? Just like, this is such fucking garbage. This is my. This is my corpus. How dare you.

Katie (18:11)

So Frankie's video starts to spread as well. Reviewers on Goodreads started to call it out for using AI Goodreads actually froze the book's reviews so no one could post a new review or an edit an existing one. Then the found of Pangram. So that's an AI detection company. He ran the book through, through their program and declared that the book was 78% AI and to be clear, none of those AI detection programs are perfect. And there are often false positives. Have you ever experimented with these, Jesse?

Jesse (18:36)

No, but I keep seeing people like, they'll. What was it? Was it Moby Dick? It was some classical.

Katie (18:44)

Oh, super.

Jesse (18:45)

Was some classic novel.

Katie (18:46)

So many. It's not Moby, it's Dick.

Jesse (18:49)

Right. It was a classic novel. It might not have been Moby. Right. But it was like it was. And it said it was almost certainly AI So I don't think. I think very few of these have been actually validated.

Katie (18:59)

I mean, there is probably. I was trained on Moby Dick specifically.

Jesse (19:03)

Yeah.

Katie (19:04)

Anyway, there are other red flags too. Like, people dug up an old interview with Mia from last summer, and it's. These are. The answers to this interview do sound very much AI So this one seems much more obvious to me than the actual paragraph of the book that we read. So, Jesse, let's read this passage. I'll be the interviewer. You be MIA So this is from the website Bookster, which is like a content farm kind of site. Uh, the interviewer asks. Throughout history, body horror and psychological horror stories have been used as allegories for dark experiences in life that could be AI Yes. An abusive relationship plays a prominent role in Shy Girl. How did using monstrosity in Gia's animal transformation as a metaphor for abuse and dehumanization help you explore the difficult topic?

Jesse (19:44)

I've always believed that horror is one of the most honest genres because it doesn't look Away in shy girl. Gia's transformation is a direct result of her submission. It's not just physical. It's psychological.

Katie (19:53)

Em Dash.

Jesse (19:54)

And it's not. It's not just. It's duh.

Katie (19:56)

Yeah.

Jesse (19:56)

Her body is changing. Her body changing is the literalization of how abuse makes you feel in human othered animal. The fur, the crawling, the barking. They're all extensions of what happens when someone convinces you that you're no longer a person. Horror. Let me explore that degradation without softening it. Let me show how loss of autonomy can manifest as monstrosity, but also how monstrosity atrocity can be reclaimed.

Katie (20:16)

Check out the. So this, this interview also included photos of Mia. Check out the photos.

Jesse (20:21)

Holy boobies.

Katie (20:24)

I mean, yes, holy.

Jesse (20:25)

She looks pretty. It's just like. It's extreme. She looks fake. The level of filtering, it looks like AI.

Katie (20:31)

It looks like actually bad AI. I don't think it actually is AI. I think it's just like intense filtering. But this is like clearly not fully human.

Jesse (20:40)

I feel bad. Like, no male author would have to like, show off his physique to be. It's just. I don't blame her for this, but it's like a very boobsy photo.

Katie (20:48)

I mean, they probably asked her for a photo. Not a photo of her titties.

Jesse (20:52)

No, I'm saying she. I bet. Well, whatever. We don't, we don't need to get. I, I, I, I bet she thinks that this will help her, like. No, Whatever.

Katie (21:00)

Yeah. Okay. So everything.

Jesse (21:02)

It's fine.

Katie (21:02)

On March 19, the New York Times published an article AI is Writing Fiction publishers are unprepared, it says largely about shy Girl. And when the author, so her name is Alexander Alter, she reached out to his shot the publisher, they said the publication had been canceled, Quote, the decision to cancel the publication came after a lengthy and thorough analysis, Hachette spokeswoman said, noting that the company values human creativity and requires authors to attest that their work is original. Hachette also asked its authors to disclose whether they are using AI to the company, end quote. And so that was the first known cancellation from a major publisher over AI generated text.

Jesse (21:36)

So she made history for her. What did you, what did she say? She denied it or cop to it or what?

Katie (21:42)

So when Frankie's shelf first put up their video, Ballard commented on the YouTube page. That has now been deleted. But we have a screenshot. So Jesse, read this.

Jesse (21:53)

I can't because it's too small.

Katie (21:54)

Just zoom in.

Jesse (21:55)

Let me zoom in. Fine. Hello, Frankie. This is genuinely heartbreaking. To write because I've been a fan of your channel for a long time. But I need to clarify a few things. I did not edit or format Shy Girl myself at the time. I didn't have the money for a professional editor or formatter. What's a professional formatter like? Oh, she mean like layout Layout? Yeah. Those services are expensive and I was working on a broken laptop doing the best I could with what I had. Someone in my writing group offered to help and assured me she'd do a thorough job. So I trusted her in the process. She also changed a lot of the wording and encouraged me to lean more poetic because that's my background and I listen. I should have done one final careful pass before publishing, and that's on me. But the formatting and editing issues being pointed out are not present in the traditional published edition. Most important.

Katie (22:40)

So that was one thing that Frankie was talking about in her video was these formatting issues.

Jesse (22:46)

Most importantly, I did not use AI to write this book. What I can say is that the version you're referencing was edited by someone else. And I only later realized she may have run parts of it through an AI tool during her editing process. When my first book, Sugar, was edited by someone different, I never received these AI allegations. The accusations started after I used the same person for my novella, which I've since pulled, and for early versions of Shy Girl. I'm not able to watch the rest of the video for my mental health. But I needed you to hear this directly. Being a black woman in the publishing space has been exhausting. I get attacked constantly when all I'm trying to do is write Mia. That was the other thing I was going to ask you is what race she is. Because if she's not white, I was. I was going to attack her. I don't like that.

Katie (23:26)

You do have the pictures of her big titties, so I figured you could tell.

Jesse (23:29)

But no, I'm race blind. I've told you that.

Katie (23:31)

I think that last sentence was actually plagiarism from me. Drum o luo.

Jesse (23:37)

Or at least it was generated by AI trained on Ijoma Allua.

Katie (23:41)

So the New York Times also published a follow up piece that included this quote from Ballard. Quote, this controversy has changed my life in many ways and my mental health is at an all time low and my name is ruined for something I didn't even personally do. She wrote, noting that she could not elaborate on how the book had been edited with AI because she was pursuing legal action. What do you think, Jesse? Do you Buy it?

Jesse (24:00)

You gave it to someone to edit. You accepted their edits without, like, inquiring? No, I mean, that's not. No, it is something I didn't even personally do. You. Did you publish these words under your name and. Yeah, I don't. I mean, first of all, I don't really believe her, like this idea that a friend did it, but no, that's like totally your responsibility to make sure. It's like if, if you gave it to someone to edit and they inserted plagiarized material from a human author, that's on you. I mean, it's on them, but it's on you. It's your name. Or if you, if you hire a ghostwriter. I mean, it's like, I don't know, what do you think?

Katie (24:32)

I mean, yes, ultimately the person whose byline it is is responsible if she got these. Like. Let's just say that that's true, that she worked with an editor who inserted a bunch of language in her work. If she didn't notice it, that's also on her. Or if she noticed it and didn't realize that it was AI. I do have more sympathy if that was the case. But the thing is, like, she says that this was not an issue when she self published her prior novel, Sugar. Well, Jesse, I looked up Sugar on Reddit. Check out this comment from the horror subreddit from a year ago.

Jesse (25:07)

It's so bad. I'm almost. This is, by the way, Living Gazelle. 2474. Good name. Living Gazelle. Not Dead Gazelle. Living Gazelle. It's so bad I'm almost positive it's AI written. I just returned mine. What a waste of cash.

Katie (25:19)

Can you return books after you read them?

Jesse (25:21)

Maybe.

Katie (25:22)

I thought it was like, like a tampon. Like once you buy it, it's definitely

Jesse (25:25)

Katie, how are you? That's not how you read books, Katie. Its descriptions are unnecessarily long and so repetitive. Using the word finality four times in one paragraph is enough. And it's so cliche. Also, like, the writing format is weird as hell and my husband pointed out that there isn't even a publisher for the book. Well, that's because it was self published, right? Mine said it was printed the date I ordered it.

Katie (25:45)

Lol.

Jesse (25:46)

Whoever's recommending it on here is a bot. I mean, printed the day you ordered it would be like printing on demand. Yeah, half of this is just like bad writing or self publishing. But anyway, so she had attracted some suspicions before.

Katie (26:00)

Yeah. There are other comments on here similar. I Just read the first chapter and a bit more and it's absolutely AI written. Complete garbage. It's upsetting. She got a major deal for her second book. Someone else says the COVID of the damn book is AI generated and I wouldn't be surprised if it was written with it too. So. So when she says that there were no issues with the first book, that just doesn't seem to be the case. And Mia does have some defenders. So there's a content creator named Audrey Henson who goes by Dr. Dossier and she wrote a substack post about this and did a video and she points out that Pangram. So that's the company that does AI detection. The CEO published this report on Twitter and Hansen says that the copy of the text he scanned was from a pirating website.

Jesse (26:40)

I mean, that's crazy that Pentagram wouldn't just get the actual book. It also, it's very unlikely that matters unless there's a pirated copy that differ. I mean, this is a very easy thing to check. I don't know why Pentagram would do it. I hope they would explain. But this is not. This is actually a minor point, I think.

Katie (26:58)

I think so too. And I tested Pangram again. Like there's lots of false positives with all of these detectors, but I tested Pangram myself. I gave it AI text versus something I wrote. And it was. I did this four different times. And it was accurate, 100% accurate every time. Which I was actually surprised about that it was that accurate. Yeah. My work is just so human.

Jesse (27:18)

I think you just have so many typos. It's like, that can't be AI. Yeah.

Katie (27:21)

Yeah. Henson also thought that Ballard's race and sex were relevant here. So quoting from her substack, I'm not making a claim about how Mia Ballard writes. I'm saying that the tools themselves have demonstrated peer reviewed patterns of racial and linguistic bias baked into their architecture. When tools with that track record are used to inform a career ending decision and that person on the receiving end is a black woman, it's always capital black. Who was already navigating an industry with its own well documented equity problems, the compounding of those failures deserves considerably more scrutiny than it has received.

Jesse (27:52)

This is dumb. That's dumb.

Katie (27:55)

Best possible ending to this. It turns out that Mia Ballard is not actually black. That she's a white man using AI images to pass herself off as black. Which is like there's like a 50% chance that's true.

Jesse (28:07)

And in real life that white man is a pretendian too. That would be the best outcome. Yes.

Katie (28:11)

He's a professor at Brown.

Jesse (28:13)

There's like a long, well, long recent history of folks like Timnit Gabriel and Emily Bender talking about bias in AI and like there's obviously some truth to it, but it's like a little bit bullshit because it really depends on the subject you're prompting it about. If you prompt it on a subject where like I'm just thinking of some of the stuff I've reported on like microaggressions research, which is garbage. I'm sure if you prompted on that it will tell you that yeah, there's evidence that microaggressions do xyz so there's not. The bias is like really complicated and really context dependent. The idea that that has anything to do with how this story went down is just such a stretch to me.

Katie (28:51)

I'm going to do just a quick experiment. I'm going to ask AI to respond to me.

Jesse (28:57)

People love this. Commenters are overwhelmingly in favor of us doing this. But anyway, go ahead.

Katie (29:01)

I'm going to. I'm going to ask Chad to always respond to me in aave African American Vernacular English. Just asked it to. I get what you're going for, but I'm not going to lock into always responding in a V E. What I can do is match your tone, keep it casual, direct, maybe a little sharper, less stiff. I think that's racist. And then when I say Irish question mark it said yeah, I can lean into Irish tone and tone, no problem. So it will do. It will like parody an Irish voice but not a black voice. So I think it's anti racist.

Jesse (29:32)

Well, Katie, I did my own experiment and do you, would you read the results for me?

Katie (29:36)

Well, it depends. Are you using grok? If you're using grok, then no.

Jesse (29:38)

I said to chatgpt, can you translate this into Jamaican patois? I went to the store to buy some coconuts and then went down to the beach to hang out with friends. I would like you to read this, Katie.

Katie (29:48)

Okay. I'm surprised that it would do this for you when it won't do a for me.

Jesse (29:53)

And don't do this offensively or racistly

Katie (29:55)

me go at the shop, if I buy some coconut demi, go down to the beach, go chill with my friend down there's.

Jesse (30:02)

I think maybe the reason it does it is because like this is an actual like recognized dialect. Is an A V. No, it's recognized.

Katie (30:10)

It's a recognized dialect.

Jesse (30:11)

We probably spent too much time on this.

Katie (30:13)

Okay, so that was probably the most consequential of these recent AI generated author controversies. Just because someone's book got canceled and pulled from the shelves, that's a big deal. But it was not the only AI generated text that made it past editors, and at least that one was fiction.

Jesse (30:27)

So.

Katie (30:27)

Also, in the past Couple Weeks, a 2025 Modern Love column was revealed to at least be partially GPT. And this was first, I don't know, discovered. This was first posted about by a Twitter user named Becky Tuke. Tuke. Becky took. She posted a passage on X from a Modern Love column called I was deemed unfit to be a mother after losing custody of my son. I fought to prove my love to him. That was a mistake, and it's a sad column for many reasons, including this passage. Jesse. So please read this.

Jesse (30:55)

I thought the fight was how I proved I loved him. But fighting didn't bring him back. It only pushed him further away. I mean, I've been primed, but that strikes me as little chatgptish. And then one day, when he was 12, he said it. I don't love you. I never want to see you again. Not hate, not anger. Just the flat finality of a heart too tired to keep trying. That's when I stopped fighting. I didn't give up. I shifted. Okay, this is.

Katie (31:17)

That's Chat GPT.

Jesse (31:18)

I mean, I don't want to pretend that I'm a good judge, but having been primed to think it was Chat GPT. Yes, it comes across as chatgpt GPT.

Katie (31:26)

It's textbook. Chad. I know Chad very well. This is Chad's voice, and it was published in November 2025. So I'm surprised it made it through because that voice had already been like, it had been established by that point that this was the AI voice. And the Atlantic published an article on this by Vahuni Vara. She plugged the column into various AI detectors. They really varied in their assessment. So Pangram said it was 60% AI. Two others said it was 30% AI. One found no AI.

Jesse (31:53)

I mean, they're not reliable. But what did the. What did the author say?

Katie (31:56)

Yeah. So the woman who wrote the column, her name is Kate Gilgan. So let me read you from. From Vera's piece. She said Gilgan, quote, told me she hadn't copied and pasted language from an AM model in her work. However, I did use AI as a tool, she added, seeking inspiration. This is a quote, inspiration and guidance and correction. She said she'd prompted various products including Chat GPT, Claude Co Pilot, Gemini, and Perplexity to help her stay on topic in a paragraph, for example, or stick to a theme. Quote, I used AI as a collaborative editor and not a content generator.

Jesse (32:27)

This was the one that generated a great response from someone. Apologies if you're going to get to this. But someone like quote retweeted that was just like, you know, you don't have to be a writer or language to that effect. And that idea that you need AI to help you stay on topic in a paragraph. Not to be a dick.

Katie (32:44)

Maybe she's ADHD not to be a dick.

Jesse (32:47)

And I'm like, I'm currently editing some stuff I've written that's way, way over long and structurally bad. But if you can't stand topic within a paragraph. Thank you, Katie. You're maybe just not a professional writer. Like, that's like, you should not have to lean that heavily on these tools. That's the problem.

Katie (33:02)

Yeah.

Jesse (33:02)

And that assuming she's telling the truth.

Katie (33:04)

Yeah. And that passage, like, I just, I frankly like don't believe her when she said she hadn't copied, she hadn't copied and pasted language from the AI model. Because that reads what you read. It reads exactly like an AI model right there. And it's always possible that like she's picking up language from AI, which in itself is, it's was like its own problem.

Jesse (33:23)

Yeah.

Katie (33:23)

And in response to her questions, a spokesperson for the Time said, journalism at the Times is inherently a human endeavor that will not change as technology evolves. We're consistently assessing best practices for our newsroom. There's no sort of editor's note on that column, so they seem to be trying to hope that this particular one just goes away. And I, I do, I imagine that after this the editors will be looking for signs of AI everywhere, particularly because just this week or last week, as most people will hear this, another New York Times contributor, this is a UK based freelancer named Alec Preston. He was caught using AI in a book review that was published by the New York Times that was actually directly plagiarized from a Guardian column. And you know, that's again, that's not surprising because AI trains on existing work. And when questioned, Preston immediately admitted he used AI. But he said it was the first time, which is either terrible luck or bullshit. And at first I thought, yeah, that's probably bullshit. And then I read this piece by Sam Leith on his subsec. He knows Alex Preston and he reached out to him and asked him what happened and, and Preston said this. Jesse, will you read this oh God,

Jesse (34:24)

it's awful and I'm so ashamed. Such a total car crash. The short version is that I had written a draft review of the book, but it was under length and I was rushing badly and drowning slightly. God, I wish I could ever, just once, have that problem of being under length. I made the stupid decision to use an AI tool to help expand and smooth it. That's such a bad idea. With instructions about US spelling and house style at the NYT, which I always got wrong. Mr. So and so, etc. They do like to use those honorifics. I looked at how it had tidied up the end of the review, but didn't realize it. It had also dropped in language from Christabel Kent's Guardian review. I was rushed and stupid and I'm so sorry. That is the heart of it. I'm not trying to minimize it. It's a serious lapse and deeply embarrassing. But nor do I think the pure AI ghostwriting version of the story is quite right. The review was mine in the vast majority of its substance and the error was in using the tool at all and then failing to catch what it had done for some reason. Maybe because it's a white man. I find this a little bit more.

Katie (35:17)

But he's British.

Jesse (35:18)

Oh, Brits are not white according to my. My calibers. I find this a little bit more realistic. I mean, I don't know.

Katie (35:30)

Well, he admits to it, whereas Mia Ballard was like, blame someone else.

Jesse (35:35)

He admits it. First of all, right, my friend did it. Yeah. Yeah. Maybe that's what makes it a little bit more convincing. That would be incredibly dumb to just let AI expand your shit for you and just accept what it produces. It's not. It's not your work.

Katie (35:50)

Like, it's not as though he just asked the AI to copy edit his work. And like I. There have been times when I have asked an AI to copy edit something and it will insert new language, which absolutely enrages me.

Jesse (36:04)

No, you. But you shouldn't have. The output shouldn't be the clean text. You should say point by point, identify any copy editing errors in this.

Katie (36:13)

Well, yeah, but.

Jesse (36:13)

Sure.

Katie (36:13)

But also, it's just like, if you want it. I mean, if you also say copy, edit this. Do not change anything. It should perform that task as directed. Don't change anything.

Jesse (36:23)

It's weird.

Katie (36:24)

Change the typos. Fix the typos. Yeah, but that doesn't sound what he was doing. He says he used AI to expand and smooth it, which is writing. And he has certainly paid the price for this collapse in judgment. The New York Times has cut ties with him. My guess is that this won't be the only outlet that cuts ties with him. And I suspect there's also a lot of writers right now who are thinking, fuck, this could have been me. I think AI is probably much more common than just the cases that we know about.

Jesse (36:51)

Well, it's also like a genuine slippery slope because I use it regularly, including as an extra set of fake robot eyes. I don't ever copy and paste its output. But I could see a slippery slope where or maybe some people just somehow haven't yet caught onto the fact that you shouldn't copy its output. I'm not trying to be holier than that. It seems obvious to me I should never. But yeah, because it is useful if you know what you're doing for a lot of stuff. And there's like a fuzzy boundary area there.

Katie (37:21)

Right. So in those two cases, those are clear cut ethical transgressions because publishers like Hachette and publications like the New York Times require that authors disclose any use of AI pre publication. And so failing to do so is a contractual breach. And it seems obvious that using AI to actually compose your text itself is a form of plagiarism. That seems very clear cut. But what's less clear cut is using it in research. So the final AI scandal we're going to discuss involves that. Not plagiarism, but research. So Jesse, have you heard of Matt

Jesse (37:55)

Goodwin one he's like a far right commentator type, right?

Katie (37:58)

Yeah, he's based in the uk, presenter on GB News. He's a failed reform UK political candidate. That's the anti immigration party there. And he just self published this book called Suicide of a Nation, Immigration Islam Identity. It was published on March 16 and then a few days after it came out a UK writer named Andy12 published this on X. So Jesse, read this exclusive.

Jesse (38:18)

Goodwin MJ's new book Suicide of a Nation, Immigration Islam Identities out now and I'm only five chapters in and have found a huge amount of what I appears to be false quotes and basic misinterpretations of data that appear to be AI hallucinations. That's awesome. Matthew, can you explain the claims you made in the book that I've outlined in the below thread?

Katie (38:34)

Okay, so he then he outlines the factual errors. So for example, here's one of Goodwin's claims that that twelves found suspicious. In one year one classroom in Bradford, only four of 28 pupils spoke English as their first language teachers reported Spending large amounts of time simply mediating between dozens of languages, making normal teaching almost impossible and slowing down the rate of progress for everybody. There's no source there, no citation. It isn't reported anywhere else that 12s could find. And this is a recurring theme. So some of his claims, just like on their face, seem pretty outlandish. Like he claims that in, quote, virtually all of London, most primary school pupils, main language is no longer English.

Jesse (39:12)

Again, that's offensive. Immigrants cannot afford to live in London. Come on.

Katie (39:16)

He again offers no evidence to support this, which could, you know, it could just be straight up fabrication. But 12 thinks it's more likely an AI hallucination. And then other people start looking into this as well. Someone finds that in his foot. He has very few footnotes in this book. There's like 12 or 14 footnotes. And two of them, most of them are references to his own substack. And then in two of them, they. You can see in the URL that he got it from ChatGPT.

Jesse (39:41)

That doesn't mean anything, Right, Right.

Katie (39:44)

I think, like, it's fine to use ChatGPT as a resource. Like, I don't find it that much different than using Google for references.

Jesse (39:51)

Google. Google. If you. Yeah, if you do a Google search and then click, it'll sometimes leave a little Google thing.

Katie (39:55)

Oh, yeah, of course.

Jesse (39:57)

I mean, the process, it just means. It just means you arrived there from ChatGPT.

Katie (40:00)

I think the process is a little bit different. Like it's easier to find sources that affirm your position using AI than Google. You know what I mean? Like, if you're just maybe searching through. Through Google results, I think you're more likely to stumble upon countervailing evidence.

Jesse (40:16)

I know, but if you're Matthew Godwin, you're not gonna. It doesn't matter anyway. I think there's a version of truly hackish research where you prompt Claude or chatgpt. Find me 12 articles showing that immigrants are bad for England. Sure, but that's not. You can't prove that just from the URL tag, right?

Katie (40:32)

No, no, no, you can't. But it also appears as though he included fabricated quotes in his book, which seems much more likely AI hallucinations.

Jesse (40:41)

Or maybe it's the Matthew Godwin hallucination. Maybe he'll heal hallucinated the quote. How did he. How did the gentleman respond to this?

Katie (40:49)

Okay, so he posted this on X. He said the left are having a meltdown about suicide of the nation, cherry picking, misrepresenting and hate bombing the Amazon review.

Jesse (40:56)

This is Just this is literally just a flip side of it's because I'm a black woman, isn't it?

Katie (41:01)

Why? Because they think they own and control the public debate. He also said the left don't want you to read. They don't want you to know what is in this book because they do not want you to know what is happening around you both. Okay, this, the book is full of fabrications. And it's not just the left. Of course she says that in capital letters. It's not just the left that's been critical. Both the Critic and the Spectator, these are both conservative publications. The Critic is very populist, probably in agreement with Goodwin, especially on immigration. But both publications wrote scathing reviews of the book. So he's trying to position himself as the victim of an unfair cancellation campaign when really this is just lazy work. I was struck by this line, this one line. His Twitter post was very long, but I was struck by this one particular line. They are constantly trying to gaslight you. It is really that simple. I heard another right wing influencer use the term gaslight on on the media the other day. They have really picked up on woke vernacular in grievance politics. They have been training on wokeness so much that it has inflected their their vocabulary.

Jesse (41:59)

It really has.

Katie (41:59)

The other thing Matt Goodwin keeps doing is he keeps saying AI didn't write my book. The Spectator actually published an article by him called AI didn't write my book. But that's not the accusation, right? The act. The accusation is that he used AI to do his research and the AI hallucinated. So that brings us to the next topic of the day. So last week, friend of the pod and Washington Post columnist Megan McArdle posted this on X. Jesse, why don't you read this?

Jesse (42:27)

I use AI to do research. I find things to read, explain parts of academic papers I find ambiguous or confusing, transcribe interviews, generate pushback on my column, thesis suggestion trims when I'm over my word count, sharpen podcast interview questions and perform a final fact check on columns and editorials. But mostly it's compressing the ancillary tasks to the main job, reading, thinking and writing.

Katie (42:46)

So she expanded on her thread and she explained what AI is and isn't useful for. And she sums it up with this post, TLDR for a journalist. In my humble opinion, think of your chatbot as a combination of an intern, a first pass editor, and a fact checker. Its job is to do grunt work and help you turn in cleaner copy. Not to quote inspire you. Before we get to our own thoughts on this, let's first go to the reactions you would have thought from the reaction that Megan had publicly admitted to plagiarizing her work. So just here's just a few of the responses. Will you read these?

Jesse (43:17)

Jesse Ben Burgess no Ben In a healthier media culture, an admission like this would at the very least get her fired. Je suis Megan no it wouldn't. Because transcribe it one more time. Let's be clear. Find things to read, explain parts of accuracy. Academic papers I find ambiguous or confusing. That could be in the wrong hands. That could be problematic. But Megan is like knows she's not going to just take it at its word. Transcribe interviews, generate pushback on my column thesis suggest trims when I'm over my word count sharpen question Yeah, I. I don't see anything wrong with any of that.

Katie (43:49)

Yep, keep going.

Jesse (43:50)

J. Dale Shoem Shoemaker I think it should be deeply embarrassing for any journalist employed by a news organization to admit to this level of AI use. I don't. I don't understand what's wrong. She's not saying I prompted to write a column for me or that I take at face value anything it tells me. I don't. I don't get this. But yes, I think Megan should be fired for sure.

Katie (44:09)

There was also a substacker named Beggar Rothfeld who said Megan is violating Washington Post's own internal AI policies. So Jesse read from her post.

Jesse (44:17)

The policy states we are transparent about how and when we use AI, but McArdle has not appended notes to her columns explaining how she has used it in each, although she's pretty transparent, she just tweeted it out, although she's apparently quite heavily reliant on it. McArdle admits here that she often asks AI to generate ideas for stories for her, yet she has not attributed anything to it in any of the resultant columns, at least. This is such a stretch.

Katie (44:40)

Yeah, let's. Let's stop there. So Megan does not actually say that she uses it to generate ideas for stories. Here's what she says. Lower enough Red. It's also useful for formatting my stream of consciousness list of podcast questions into a usable script, suggesting podcast guests, and generating lists of potentially interesting stories for me to check out for columns or editorials. Things I don't do with it. Go into a chat with an AI without my thesis already strongly developed. Have it outline or write. Accept anything the AI says. Please provide link. Sources should be in every prompt. She seems Pretty clear that she is not using AI to think for her. I mean, am I misreading this? Am I biased because I like Megan?

Jesse (45:20)

I don't think so. You also missed the most annoying part of this Becca person's substack, which is what can we do about this? I think we have an opportunity to demonstrate what is and is not acceptable to the reading public. That's capital R, capital P. By writing to the Post Standards desk demanding that they enforce their policies by asking McArdle, asking McArdle adhere to them. That's not AI because it's a typo. Here is the email of the Post Senior Standards desk editor. Please. Like, this is so obnoxious.

Katie (45:46)

She's tattling as though Megan were, like, trying to hide this. I'm pretty sure her bosses follow her on Twitter.

Jesse (45:53)

If the policy just says we are transparent about how and when we use AI, that should be phrased a little bit more clearly because I do not think you need to admit to using AI for, like, basic research assistance anymore than if you have, like, a paid research helper. You don't necessarily need to attribute everything to the helper if they just tell you what papers and stuff like that. I think it could be the policy should be phrased more clearly. But, yeah, I don't. I don't really get this.

Katie (46:22)

Yeah. And the thing that most critics seem particularly mad about is that Megan said she uses AI as a fact checker. Megan is very clear in the thread. This is a first pass fact checker. This is what she does before she sends this to her editors. It's the first step, not the last step. She has editors and Washington Post does employ actual fact checkers at the paper. And as far as I'm aware, no one has pointed to a single correction or needed correction on any of Megan's post AI work, post LLM work. So people were acting like obviously AI had introduced errors into her columns without giving a single example of it. And obviously there is a risk to using AI in research. Matt Goodwin found that out. There's also a risk to using any source in research. I would say that there's a greater risk in AI because chatbots are. Are overly confident and they make research really easy. But I don't think Megan McCardell is skipping primary sources or just blindly inserting hallucinations into her columns. She's not Matt Goodwin.

Jesse (47:18)

It's basically, you basically just need to adhere to the same rules you do with Wikipedia. You would never copy and paste something from Wikipedia. You would never not check the sources. There's obviously some extra wrinkles with AI because it feels more like you're conversing with an intelligent being. But it's like if you basically stick

Katie (47:35)

to Wikipedia, it does not.

Jesse (47:36)

No, if you stick to Wikipedia rules, like I think 90% of the trouble you could get into, you won't get into.

Katie (47:43)

What do you think accounts for the pretty unhinged response to her post? I mean, this like really took over Twitter for this was like supplanting Lindy

Jesse (47:50)

for a little while.

Katie (47:50)

It was terrible.

Jesse (47:51)

It's like two storm systems combining. People hate Megan because she's like a libertarian. They are mad about stuff going on at the Post. I've also read about this in my newsletter, which Freddy DeBoer has not. Not enjoyed my takes on this and has let me know. Sorry, I'm going to change the episode since we mentioned Freddy DeBoer. Counterbatch0 there's this weird strain of like deep moral disgust with AI mostly on the left that look, I think it's going to be really disruptive and could take a lot of jobs. And that of course partly accounts for this. But some of the, the takes you see, in my view just aren't really thought through. And to me, in the category of not thought through, take I would say a Megan McCardell using AI basically the same way you'd use like an entry level research assistant. I don't think there's. There is something wrong with that in the fact, in the sense that it affects the labor market. But like every time you use Uber or Lyft or Seamless, the self checker

Katie (48:56)

at the grocery store.

Jesse (48:56)

Yeah, like every. Everything affects the labor market. Market. So I just, I, I find a lot of the lefty takes on this. I don't think they like sort of take the technology seriously enough. Because if you do, you should realize it. Of course this is going to be useful to someone like Megan McArdle.

Katie (49:10)

Yeah. You know, and I think people are rightfully scared about their future. There's this precarity in the industry and you know, Washington Post has done mass layoffs and Megan McArdle is one of the few existing columnist who kind of benefits from their like, restraint. I don't know about restructuring, but at least from their realignment of their values because she was already in line with Jeff Bezos's mandate for the editorial board to focus on free minds and free markets. She's a libertarian.

Jesse (49:38)

Yeah.

Katie (49:38)

And that mandate was definitely not good for someone who spent a career arguing for American Maoism or Whatever. But that's not Megan. So first off, like you said, a lot of people don't like her just for her politics. And they're resentful that this woman with a job that barely even exists anymore is. Is using what they see as a cheat. And look, I'm scared about this too. I spend a lot of time thinking about what I'll be able to do when this podcast inevitably burns.

Jesse (50:02)

It's going to be spectacular. It's going to be like 12, 9, 11.

Katie (50:07)

So far I've settled on janitor at the police department in the hopes that maybe occasionally they'll let me help solve crimes. I looked it up recently and you can't just be a detective. You have to start. You have to go to the police academy and you have to start at the bottom, which I don't want to do.

Jesse (50:21)

What I think is going to happen is one of the detective conference rooms. They're going to leave a complicated case on the blackboard. You'll just be bopping up at 2am There should be a movie about that. It was the butler in the conservatory.

Katie (50:34)

Yes. That's my plan B.

Jesse (50:36)

That's my understanding of police work.

Katie (50:38)

Yeah. So I completely understand why people.

Jesse (50:41)

Can I make one other point about this? This is really important. I was talking to someone and they said that in San Francisco. And let me double check this on the fly age limit. Police.

Katie (50:51)

Hopefully it's 80.

Jesse (50:53)

There's no. I think I could become a San Francisco cop. And I was thinking, should I do that?

Katie (50:58)

You.

Jesse (50:58)

Should I be a cop?

Katie (50:59)

No, I'm gonna be. You can't steal my job. I'm gonna be a janitor, slash cop.

Jesse (51:03)

No, you're gonna be a cop's janitor. No, I want to be. I think being a cop would like. Jesse, you know, all those gaps I have and like not knowing anything. Think about the world. That would fill those in really quickly.

Katie (51:13)

I have a really great. I have great news for you.

Jesse (51:15)

Go for it.

Katie (51:16)

They just increased, I believe, the age at which you can enroll in the military to 42. Now would be a really good time to join the military.

Jesse (51:25)

That one I'm gonna. That one I'm gonna skip.

Katie (51:28)

I think the other thing, like you mentioned this sort of spiritual contamination of AI and I think that's also a huge part of it. And I saw.

Jesse (51:36)

I don't think I did. I did. I mentioned it.

Katie (51:37)

Did you not? I saw a, a great example of this later in the week. Jesse read this tweet right here trying

Jesse (51:47)

to find a hex code For a specific color, dropping it into ChatGPT, Claude, Gemini. They each give me a different hex code. I run it through again multiple times on each model. Now each of the models are giving me different hex codes. I love this system. And that's by Taylor Lorenz.

Katie (52:01)

Okay, so read. I posted some of the. I pasted some of the responses to her. Read these.

Jesse (52:07)

These are gonna be. These are gonna be hinged, right?

Katie (52:10)

Absolutely hinged.

Jesse (52:11)

I don't know.

Katie (52:11)

If things get called hinged, we only get unhinged.

Jesse (52:13)

It is extremely confusing for you to do so much journalistic work regarding air quality and how it systematically impacts marginalized communities, then go on to use a technology that poisons black communities with air pollution and by your own admittance, doesn't even work. Can I just. Can I say something? I've. I can't remember if I brought this up on one of my posts or an ama. This is so dumb. Because you're. You're. Well, because you're missing. That's what I wanted to say. This is so dumb. No, the first of all, I think there's a lot of bullshit statistics going on around, like energy and water use. I'm sure AI uses a lot of energy, but if you're going to yell at Taylor about the, about this, you need to know what you're comparing it to. Are you comparing it to her over the course of 20 minutes doing Google searches for each of these codes, each of these colors, paging through many websites, all the extra work she would have to do? Are you sure that would use less energy?

Katie (53:09)

What I'm curious about is, is you do so much journalistic work regarding air quality. Is that about COVID Because Taylor does not do, like, environmental reporting. She complains about people not wearing masks.

Jesse (53:21)

People just. I think this is what I mean by, like, sort of annoying. Lefty takes. They just like apply wildly disparate standards to anything having to do with AI. Like, if Taylor didn't use AI, I bet she would have used more energy. Or we should at least acknowledge that possibility.

Katie (53:36)

Go ahead and read some more of these responses.

Jesse (53:39)

For someone who talks a lot about masking and the importance of clean air, you'd think using a tool that runs on data centers that poisons the air of the local communities would be a big no, no. At the risk of being called dumb again for not understanding Taylor's AI use. Why are we running futile experiments on planet broiling nightmare machines when the search bar and letters hex code finder are right there? I can't. I Can't anymore with these. These people are so insufferable.

Katie (54:03)

Okay, so I, I think this is the problem. This is sort of a self imposed problem that, that Taylor has made for herself. Because if you make your brand ally of marginalized people, which Taylor has attempted to do, I don't think very convincingly, but I think she's attempted to do this by scolding other people, you are held to an insanely high standard. She's never going to be able to take that Covid mask off. But Jesse, do you have any moral qualms about using AI from an environmental perspective?

Jesse (54:29)

I don't. As like, I just got to say man, I look, maybe I should look more to the energy usage stuff, the extent to which we're all destroying the planet every day. And look, first of all, I'm a vegetarian, so I can cash in that moral high ground. I'll be one who buys milk and eggs. But I suspect that there's a subset of ChatGPT or cloud prompts that save energy over the alternate googling you'd have to do to answer the same question. I don't know that for sure. That's just an intuitive guess.

Katie (54:58)

We do know that these, these data centers use incredible amounts of power. That's. There's no question about that. How bad that is for the environment really depends on the source of your energy. Right? So if your energy is, if, if all of the energy powering a data center is coming from nuclear power plants, the, the impact on smog, on air quality and on the climate is minimal. If he's using fossil fuels, that's not the same. So I'll link some to some reporting in the, in the show, notes on this that specifically looked at at Memphis. Somebody responded to Taylor, said something about how, how AI is poisoning black communities. And this does sound on its face like woke nonsense. But there is, if you look at individual cases, there is some truth to that. Although I think it probably tracks more cleanly with socioeconomic status than race. But like if you look at this, at this colossus, this is a data center in Memphis. It's run by xai. When it was built, it was the world's largest supercomputer. That data center has its own power plant with 35 methane gas turbines. They're only permitted to have 15. They have 35. And Xai has said that their power plants are going to be super clean. But activists there in Memphis claim that smog is up by 30 to 60% in the area since the data center went up and residents claim that you can taste the air pollution. So a reporter for the Atlantic went there and he said he could feel it in his throat. I think it's probably too soon to definitively connect that data center, any data center with increases in disease or cardiovascular issues. But no one wants to live near 35 gas turbines.

Jesse (56:28)

I'm sure, like, in long run, AI is and will continue to use a lot of energy. But I just think that every fucking app I download now or update has integrated AI stuff I didn't ask for. And that is often completely goddamn worthless.

Katie (56:41)

So, like, it's all AI.

Jesse (56:43)

Like, Taylor Lorenz is a journalist. This makes her job a lot easier. It often makes my job a lot easier if you use it right. Although she's right that, like, it'll randomly shit the bed.

Katie (56:51)

Well, I mean, you know, you mentioned the fact that you're vegetarian. That has. Right now eating meat has a much higher impact on the climate and on water than AI. This could change as there's more AI. The.

Jesse (57:05)

This is all guided by people's moral discussions just du jour. They're mad about AI right now, so they try to backtrack into some, like, reason to be mad about AI. And there's legitimate reasons to be mad about AI. I just don't think the energy of individual user searches is one of them.

Katie (57:17)

If, I mean, if you're eating Hamburger Helper twice a week, cutting that out would make a much larger impact than not. Than not using chat GPT.

Jesse (57:25)

Given that, given the sorts of people who get mad on the Internet, half of these people are literally dripping hamburger juice on their keyboard as they yell at Taylor Lorenz about using AI.

Katie (57:35)

But it is, I think there's this idea that there's this, this, that using AI is inherently unethical, not just because of the job stuff, but also because of the environmental impact. And so this is a way of, of doing this sort of moral posturing. If you insist that you don't use AI, here was Taylor's own own justification. She said, people, people asking why I, a technology journalist, would ever search something via AI. I am a technology journalist. I cover tech from the user, user side. I love technology and love trying new technology. And it's crucial to understand the capabilities and limitations of consumer products. I think that's fine. But I also think that Taylor, because she pretends to be this sort of moral arbiter, of course her audience, the audience that she has cultivated, which is dumb.

Jesse (58:20)

We're the. When we're the moral arbiters and we've made that clear for years. There can only be two.

Katie (58:24)

She is a woke skull who has cultivated an audience of people who get mad about the smart smallest things. Of course she has pissed off a bunch of her own followers by admitting that she uses AI for dumb like looking up a hex code. Yeah, Jesse, I'm curious, what's your sort of like, what's your AI stack? What's your. I'm not using the word stack correctly, but what's your standard like use for if you're working on an article? How will you use AI

Jesse (58:48)

sometimes to like get me started on research sources, depending on the subject. Sometimes if I like think I remember having seen a quote somewhere, I'll be like, did anyone say this?

Katie (58:57)

You'll give it access to your. To your Gmail?

Jesse (58:59)

It did briefly. That was a mistake. Although since I did that now in Google Mail it just. There's a thing there that has access like you can prompt it to search your inbox. So a Google AI agent thing agent has access to all. Well, it's. Is it. An agent really usually means something that can like go out in the online world and do stuff, but whatever. I basically use it the same way you'd use like entry level human assistants. Now I literally just hired a new assistant editor type, so it doesn't replace human stuff. But yeah, I'll use it for copy editing. It's really good, I find. I use Claude. I'm a Claude boy Boi. If I paste in a draft of something I'm working on and I'm just like critique this. Let me know if this makes sense or if there's weak points in the argumentation, I find it to be pretty good at that. Of course the whole reason I find these tools useful and I think they make me more productive is because I entered this having about two decades of experience in journalism. If I was 23, I'd have to be so much more careful because I wouldn't have the expertise needed to know which of its output is useful. How do you, how do you use it?

Katie (60:08)

Basically the same. I use it to poke holes in my own arguments and you know, basically things that if I were still working,

Jesse (60:14)

hey chatgpt, call me a piece of shit.

Katie (60:16)

Does it in like that. Like the same things that I when I was at an institution that I used to use an editor for copy editing for one. Although I have to be very specific like do not change any fucking language. Just figure out, fix the typos. And then I find it really useful for, for fact checking and you have to be very specific. Like you. If you're. If AI is fact checking. If Claude is fact checking something for me, I need it to give me the exact source. And it's oftentimes it's wrong. It will give me a link to. It will give me a link, and then I'll search the link.

Jesse (60:48)

And so why do you use it for fact checking if it's often wrong?

Katie (60:51)

Because it's often right too. And I also don't like, I don't work with an editor. So this is. It's a proxy for an editor. It's a starting place.

Jesse (60:59)

If you're working on a print story at one of the diminishing number of outlets that has fact checkers or print or online, or you're like writing a book and hire a professional fact checker, you can basically take what they find at more or less face value.

Katie (61:16)

Well, actually, they're probably using AI too.

Jesse (61:20)

Okay. If the Atlantic has a staff fact checker fact checking your story, then norm is you basically accept their conclusions unless there's a reason to push back. You would not triple check every single thing they say because that would take forever. And if there's an error, you then have a log showing that it was their fault. With an AI fact checker, it's very different because you basically have to really carefully. You can't just be like, well, the AI fact checked it. It's very different.

Katie (61:43)

Right, Right.

Jesse (61:45)

I was in like a group chat with other journalists and I joked about how I didn't want to say this publicly because it makes me sound like a holier than thou douchebag. I can't imagine using it to produce writing under my name. Like, I'm a fucking writer. I do this for a living. That's not my work. And I bet you feel similarly. It's just like, it is like this pretty visceral thing of like, ew, no,

Katie (62:05)

that's how I feel. It's like, it's the ick factor. I also enjoy writing, though, to me, that's the more interesting part of the job, is actually writing. I like writing and I like. Yes, that's the part that I like.

Jesse (62:15)

If you were like a stressed out 20 year old with a term paper who didn't like writing, the 10 temptation would be enormous.

Katie (62:20)

Right. And you know, there are. We talked on another. I'm not sure if this is a primo episode or a free episode, but we talked a couple weeks ago about the Cleveland Plain Dealer. An editor there wrote an editorial about how they're using AI. And he basically was like, look, we have our reporters go out and they take their notes, they dump it into an LLM, and then we have human editors and fact checkers, and the reporter gets final say on what goes out, but we're using them more as straight up reporters, as information gathering humans. And the LLM is doing the writing. And you pointed out in that episode that some journalists, especially kind of old school ones, are shit writers. And so that. And like, I've been an editor before. That's absolutely true.

Jesse (63:01)

Yes.

Katie (63:02)

That some, like some journalists, including some journalists that you've all heard of, are really fucking bad writers themselves. And so I can see the utility there, personally. It's just like I take way too much pride in my own choice of words to actually do that. But I think what will be a more interesting question is in the future, what happens, you know, if and when the tech is so good that it doesn't hallucinate? Or when Megan McArdle can say to an LLM, Write me a thousand words on the national debt in the style of Megan McArdle. And it does it perfectly.

Jesse (63:34)

Yeah.

Katie (63:34)

What happens when AI novels are truly indistinguishable from human novels? Or even better, you know, people keep saying that AI is empty, it's soulless, that this, you know, this writing is an art. It can't truly replicate human creation. I'm not so sure that's true. But at the same time, the experiments

Jesse (63:52)

we have suggest that the average person can often cannot tell the difference.

Katie (63:56)

Yeah, yeah. And at the same time, do I want to read AI novels or articles or listen to AI podcasts or music? No, I don't. But I'm not totally sure I can tell you why, other than it's icky, you know, but if I read a Megan McCardell story or a Jesse Single story and found out later that this was actually a Megan McCardell bot or a Jesse Single bot, you know, that would probably be the last Megan McCardell or Jesse single story I'd read.

Jesse (64:19)

Yeah. I'd be like, why am I bothering with my attention, giving my attention to this person?

Katie (64:23)

Right, Right. But of course, you know, pretty soon it might be very well impossible to tell the difference.

Jesse (64:27)

Can't wait. Katie, should we wrap it up?

Katie (64:30)

Yes.

Jesse (64:31)

This has been blocked.

Katie (64:32)

We once again forgot to do housekeeping. We've got last week and this week.

Jesse (64:36)

Well, should we do it? We're a podcast.

Katie (64:38)

Yeah. No, do it, please. Please subscribe.

Jesse (64:40)

We're a podcast. Please subscribe. Go to blockedreport.org Rate and review us on Apple podcasts. This has been blocked.

Katie (64:46)

$7 a month.

Jesse (64:47)

This has been blocked. Reported. No AI.

Katie (64:49)

Lots of extra content. No AI. No AI yet.

Jesse (64:52)

Well, not no AI, but no we.

Katie (64:54)

Do we. Absolutely.

Jesse (64:56)

I don't know that you're not a. I don't know that you're we.

Katie (64:58)

We. You know what? I will admit the amazing art that we use on our show Notes. That's AI.

Jesse (65:05)

Oh, really? That's AI.

Katie (65:06)

I know. People are shocked to hear this.

Jesse (65:07)

Was it all the humans with 24 fingers that gave it away? This has been blocked and reported. As always, we are produced with help from Jessica the 80s baby. Thank you for listening. Bye bye.