Earlier this month, one writer’s career came under fire. Hachette, one of the major publishers often referred to as the “Big Five,” has announced that it will no longer publish horror novels in the United States. shy girl The book, written by Mia Ballard, was published in April this year and was due to be pulped from an existing copy that had already been published in the UK late last year. reason? artificial intelligence.
Rumbles gained traction in January (and earlier) on platforms like Reddit and Goodreads, where readers in the UK who picked up the book said they had the following to say: shy girl There was a characteristic of the way ChatGPT was written. A nearly three-hour video by book-focused YouTuber Frankie Shelf dissected the entire novel, pointing out things like the endless repetition of words. sharpexcessive “rule of three” structures, and various other assumptions about AI-generated sentences. These accusations eventually reached the New York Times, which verified the claims made by Max Spero, the founder of the AI detection program Pangram. shy girl It seems that the sentences generated by AI are showing a characteristic pattern. In response, Hachette announced that he would withdraw the publication of the novel, marking the first known example of a major publisher withholding a book due to accusations of AI use. (Ballard herself hasn’t said much, other than to give the Times a vague statement echoing now-deleted comments she left on a Frankie’s Shelf YouTube video, claiming that an acquaintance of hers edited the book and may have introduced the AI to do so.)
There has been a lot of uproar in recent weeks, much of it in anger, since this bombshell rocked the publishing world. shy girlthe threat of machines invading what we read, and also what AI-generated writing might look like. I just read it shy girl After obtaining a copy of it through covert means, you can confirm that it is not a very good novel, whether the person who wrote it is human or not. I wanted to see if I would be fooled as much as the person involved in the publication of this title by Hachette. But I also wanted to understand how this happened in the first place. In retrospect, how did a book that was clearly written, at least in part, by generative AI end up on the market?Many in the book industry continue to send out warning signals that something like this could someday happen. What made them so sure? I spoke to various people in the publishing industry to find out.
Author and critic Emily C. Hughes read: shy girl I thought it was okay a few months ago. It’s not life-changing, but it’s okay. When the AI accusations come to light, she can’t help but think about the implications of not realizing it on her first reading. Many have asked why Hachette’s editors didn’t notice that there was something wrong with this book. This probably has something to do with the fact that the book was originally self-published. “When a publisher takes a previously self-published book to reach a wider audience, the amount of editing between the self-published version and the traditionally published version is minimal,” Hughes says. “It’s probably copy editing. And the logic there is, Well, people already like this so we don’t need to tinker too much”
The biggest problem is that editors often don’t understand this issue. A.I. This could be because you simply don’t have the time to edit your book the way you want. To.
But it goes deeper than that, and the people I spoke to said they expect to see the use of AI permeate print books across the industry. Books whose selling points are based on ideas and plots, rather than the quality of the writing itself, will likely be most affected by AI, said one editor who worked at a major publishing house and asked not to be named. “Editors have made it clear that it’s not acceptable, but it’s hard to police it without potentially hurting people who write poorly.”
The problem isn’t just that large-scale language models are now widely available to people who want to use them to clean up their writing. And it’s not just that the technology is still so new that publishers are rushing to refine their stance on it. However, both of these are certainly true. The biggest problem is that editors are often unaware of the use of AI because they don’t have the time to edit the book as accurately as they would like.
“As the publishing industry has shrunk over the past 20 years or so, fewer employees are being asked to do more work and take on more roles. So the expectations for editors are higher, and editors can spend their workdays just like editors,” Hughes said. Industry insiders say most editors edit at night and on weekends, and must devote much of their professional lives to acquiring books, working on campaigns to sell them, and managing relationships with authors. “I hope editing becomes the core of the job again,” said an anonymous editor. “If you conducted a survey asking if editors have ever told you that editing takes too long, I think you would get a 100 percent response rate.”
Another editor said: “Quality control has collapsed.” John Baker, a literary agent at Bell Lomax Morton, said editors are seeing their work becoming more and more about project management, and while publishers are making a lot of money, “the last thing they want to spend that money on is hiring more staff. It’s the norm to have everyone working above capacity.”
Assuming unfortunately that the publisher is unlikely to suddenly change direction and increase staff to free up editors to focus on actual editorial work, how do we move forward from here? What’s to stop one more? shy girl After it happened? While AI detection software is improving, it has been shown to be fallible at best, often incorrectly labeling text produced by non-native English speakers as AI-generated. And the whole point of an LLM is that you’re constantly learning. What may seem like easy ways to identify AI language today (like em dashes, wink, wink) will soon be eliminated as the model learns that these em dashes indicate that.
What I read shy girlI discovered that I think as AI says: overuse of parallelism, endless repetition of the same vocabulary, and awkwardly rich similes. I could cite any number of individual passages that ring alarm bells, such as:
Breathing calms down, the contours of possibility take shape. It’s work. That idea will never go away, it can’t be solid. A ray of hope.
The call ended and I sat there for a while, the phone warm in my hand and the silence echoing loudly. That fleeting swell of hope that tries to break through the cracks. But that feeling doesn’t last. I snapped back when I heard the notification buzz, the vibrations echoing through the table like an electric shock.
Nathan’s name flashes on the screen.
Heart tightening and anticipation rising rapidly, I unlocked my phone with a clumsiness that betrayed me.
But many human writers also write that way, including non-native speakers, autistic writers, and people who choose writing styles simply because they want to write. After all, LLMs write in their own way because they learned from us. What finally convinced me that AI was involved in the text I was reading was the feeling that there was literally no human being behind the words. Of course, knowing the context, I would want to look for it, but while reading I felt it nagging shy girl. There is a consistent flatness to the novel, a single emotional note echoing throughout, and an empty note.
It was a feeling. It literally felt like there was no one behind me. words.
I spoke to one of the first AI detectives on Reddit about the potential uses of AI. shy girl. Let’s call this woman Dora. Until a few years ago, she worked as a freelance literary editor for 12 years, when she decided to take a break from the game due to the amount of material ChatGPT generated that required reading. “I think it’s no surprise, but I love books, I love art, I love people, and I was in the publishing industry, so I found that a little soul-destroying,” she said. she got a copy of shy girl She was a huge horror fan, and the concept of a woman kept as a pet by an abusive man who finds herself exhibiting increasingly animalistic qualities seemed compelling. From the first sentence, she sensed the all-too-familiar badness of the AI. “I continued reading, but there were no phrases, word choices, grammar, structure, syntax, or punctuation in my opinion that could not be achieved in ChatGPT.” But again, more than individual details, it was about the emotions she felt when reading the novel. “I think it all comes down to pattern recognition,” she told me. “I mean, humans are good at that.”
Dylan Garrity, a Brooklyn-based freelance editor who has edited more than 500 books, agrees that this is not an individual story. It’s more about the overall feeling you get when you read it. It sounds kind of vague and impossible to back up, but that’s pretty much what we have to go on.”
This instinctive feeling is something that AI companies will inevitably develop LLMs to avoid triggering. But for the rest of us, in the absence of reliable software or reliable means to confirm our suspicions when those feelings are triggered, we have to rely on trust. The Authors Guild, the largest professional organization for authors in the United States, has launched a program called Human Authored. The program allows authors to register their books and earn the right to use a certification mark on the cover to certify that no AI was used in the creation of their work beyond minimal spelling and grammar checking. This will be an honor-based system. Authors can say they don’t use AI, and we, as readers, are obligated to believe it.
This is what is so harmful about this shy girl A fiasco: It’s not that a single book that appeared to contain AI-generated content ended up in the unwitting hands of a small number of readers, but that such a high-profile scandal has eroded trust in the publishing process as a whole. “My big concern is that as an industry we are blind to how much public goodwill there is for the publishing industry, and that some major scandal could completely ruin it,” another anonymous editor told me.
Author and critic Hughes is also most concerned about the trust factor. Now, when readers are reading a book, she says, “there’s a desire on the part of the audience not to be drawn in, not to be fooled.” Reading is “an activity that has brought great joy to many people for a long time, but it’s really depressing to have such doubts introduced into the process. I say this as a reader, as a writer, and as a critic,” she lamented.
I personally hate AI. If I could dig a hole in the desert, dump all my servers there, and never talk about it again, I’d be right there with a shovel. And I’ll raise my hand too. I too have been fooled many times by AI-generated videos I’ve seen on Instagram, such as cats doing strange things. As a result, I now scroll through social media with a new, unpleasant skepticism that cuts into what little joy there is in casually browsing social media. Is this true or am I being fooled? I can’t help but think. I don’t want to do this while reading. It is notoriously difficult to regain trust once it has been broken. regain.
