More than a million LinkedIn users have hit the “Seems like AI slop” button

The social network LinkedIn has shared the first results from its “Seems like AI slop” button, launched on July 30. As chief product officer Hari Srinivasan announced on Thursday, it has already been clicked by “more than a million people.” The button is available from the three-dot menu next to every post and lets users flag entries that look machine-generated.

The feature was prompted by an unpleasant discovery. A few weeks before the button launched, the AI-detection service Pangram found that 41% of long-form posts on LinkedIn were flagged as entirely generated by artificial intelligence — as reported by the outlet 404 Media. In response, the platform not only added the reporting button but also rolled out “new and improved” classifiers for detecting AI content, and removed a feature that offered to “enhance your post” with the help of a neural network.

“Fighting AI slop is a priority for all of us,” Srinivasan said back at launch. By his account, the measures are working: according to the company, users now see 40% fewer posts classified as “slop” than they did a few weeks earlier.

Feedback, with a caveat

LinkedIn is going further and adding a notification for post authors in case their entry has raised suspicions. The message will read: “Some members told us this post seems like AI.” At the same time, the company stresses that it isn't out to brand anyone. “We assume good intent; I myself increasingly think about how not to sound machine-like, and the goal here is to provide useful feedback,” Srinivasan explained. The wording is telling: it turns out that even the platform's top executives now watch to make sure their own texts aren't mistaken for the work of a neural network.

This isn't LinkedIn's first attempt to set things straight. Earlier this year, the social network promised to crack down on comments created en masse “with minimal human involvement or none at all.”

It's still unclear how accurate the new classifiers are and whether real authors — those whose style strikes the algorithm as too polished — will suffer for it. But the very fact that a million people are readily hitting the report button suggests that fatigue with artificial text on the professional platform has reached a notable level.

Source: The Verge