At the beginning of the year, Meta began tagging AI-generated photos. The goal was clear: to protect users on Facebook, Instagram and Threads from such content and to always inform which photos are real and which are not.

Problem? Categorizing AI-generated images isn’t that easy, even for Meta. As reported by TechCrunch, in recent months we’ve been seeing actual photo examples (rather viral) where Meta isn’t very clear on what AI is and what it isn’t.

A perfect example is the photo of Kolkata Knight Riders winning the Indian Premier League Cricket tournament. The photo looks like it was “generated by AI” when it is nothing more than a real photo taken during the celebration.

Former White House photographer Pete Souza also complained about the label online. The black and white photo, taken during a basketball game, had the label “Created with Artificial Intelligence”. Do you have doubts? I had cropped the image using Adobe tools, and this small change could have caused the algorithm to classify it as altered (something that has little practical meaning since all professional photos are edited).

“Our goal has always been to help people know when they see AI-generated content. We’re taking the latest feedback into account and continuing to evaluate our approach to ensure our labels reflect the amount of AI used in an image.”

Speaking to Techcrunch, Meta said it was aware of the recent complaints and was “evaluating its approach” to ensure the labels reflect the amount of artificial intelligence used in the images.

Image | Feast Feyissa

in Xataka | The European elections are approaching and they wanted to help us by voting on Facebook and Instagram. AEPD banned this