A book app used AI to ‘roast’ its users. It went anti-Uk instead

Fiction, a popular one Social media apps Which describes itself as a haven for “bookworms and binge watchers,” has created an AI-powered year-end summary feature that describes what books users read in 2024. It was playful and fun, but some of the briefs sounded strangely combative. For example, author Danny Groves’ summary, after labeling him a “diversity fanatic” asks if he’s “in the mood for straight, cis white people’s views.”

Book influencer Tiana Trammell’s synopsis, meanwhile, ends with the following advice: “Don’t forget to show the surface for white writers sometimes, okay?”

A reader summary shown on the 2024 statistics page from the Fable app.

Courtesy of Tiana Trammell

Trammell was shocked, and he soon realized he wasn’t alone after sharing his experience with Fable’s essence in threads. “I got multiple messages,” she says, from people whose abstracts made inappropriate comments about “disability and sexual orientation.”

Since its debut Spotify wrapAnnual recap features have become ubiquitous across the Internet, providing users with a list of how many books and news articles they’ve read, songs they’ve listened to, and workouts they’ve completed. Some companies are now using AI to completely produce or enhance the way these metrics are presented. Spotify, for example, now offers one AI-generated podcasts Where robots analyze your listening history and make inferences about your life based on your tastes. Fable started the trend by using OpenAI’s API to generate a summary of the past 12 months of reading habits for its users, but it didn’t expect the AI ​​model to turn up the commentary of an anti-Oak pundit. .

Fable later apologized on several social media channels, including Threads and Instagram, where it Posted a video “We are deeply sorry for the hurt caused by some of our readers this week,” the company wrote in the caption of an executive at issue. “We’ll do better.”

Kimberly Marsh Alley, Fable’s head of community, told WIRED that the company is working on a few changes to improve its AI summaries, including an opt-out option for people who don’t want them and clear disclosures that indicate they’re AI-generated. “For now, we’ve removed the part of the model that roasts the reader for fun, and instead the model simply summarizes the user’s taste in books,” he says.

For some users, adjusting the AI ​​doesn’t seem like an adequate response. Fantasy and romance writer AR Kaufer was shocked when he saw some screenshots of the synopsis on social media. “They need to say they’re ditching AI altogether. And they need to issue a statement, not only about AI, but also apologizing to the victims,” said Kaufer. “This ‘apologist’ in the thread comes across as absurd, mentioning the app is ‘funny’ as if it somehow excuses the racist/sexist/disabled quotes.” In response to the incident, Kaufer decided to delete his Fairytale account.

So did Trammell. “The appropriate course of action would be to disable the feature and conduct rigorous internal testing, incorporating newly implemented safeguards, to the best of their ability, so that no further platform users are exposed,” he said.