
"Following backlash over an increase in "AI slop" taking over users' feeds and making it less useful, Pinterest on Thursday added new tools that let users limit how much AI-generated content they see in their feed. Users can now personalize their feeds to restrict GenAI imagery in some categories, and the company said it will make its existing GenAI content labels more noticeable in the days ahead."
"The site, widely used to browse and bookmark inspirational content and potential purchases, has come under fire from users who complained about the massive uptick in GenAI content. The media also documented the problem, openly wondering if AI has already ruined Pinterest or if it could still be fixed. The problem, if left unresolved, could destroy Pinterest's reputation, and ultimately, its bottom line. It's also potentially a tough nut to crack as more of the web fills up with AI-generated content."
"Citing academic literature, Pinterest notes that GenAI content now makes up 57% of all online material. Beginning to address the issue, Pinterest earlier this year introduced "AI modified" labels that would appear on images whose metadata indicated AI generation, or if Pinterest's own systems detected that the content was AI-generated. At the time, the company said it would introduce consumer-facing tools that would let users choose to see less AI content."
Pinterest added controls that let users limit how much AI-generated imagery appears in their feeds and personalize reductions for categories such as beauty, art, fashion and home décor. The platform plans to make GenAI content labels more noticeable. Users will find the controls under Settings > Refine your recommendations, with the tools launching first on the website and Android and rolling out to iOS in the coming weeks. Earlier labels flagged images by metadata or Pinterest's detection systems. Pinterest cited academic literature estimating GenAI now comprises 57% of online material and plans to add more categories based on user feedback.
Read at TechCrunch
Unable to calculate read time
Collection
[
|
...
]