Stop me if you’ve heard this one before: An app that uses augmented reality to make face filters gets in trouble because of a short-sighted feature.
You’d think apps would learn their lesson after Snapchat’s yellowface filter controversy, or its backfired Marie Curie lens or the digital blackface filter for Bob Marley. FaceApp is the latest filter flop after its “hot” filter made people look whiter.
FaceApp is riding a wave of popularity in recent weeks as people started posting memes of the photo-editing tool, giving smiles to grimacing faces like the Undertaker’s.
The app is intended to be goofy and light-hearted, by letting people give themselves a younger or older look, or swapping their genders. It uses artificial intelligence to scope out parts of your face in photos and transforms them based on the filters you select.
But its “hot” filter is landing the app in hot water on social media for literally whitewashing people’s photos in an attempt to make them look more attractive.
The app’s creator and CEO, Yaroslav Goncharov, apologized for the feature’s flaw, blaming the issue on a neural network — an aspect of AI design meant to help computers function more like human brains.
“We are deeply sorry for this unquestionably serious issue. It is an unfortunate side effect of the underlying neural network caused by the training set bias, not intended behavior,” Goncharov said in an email to CNET. He added that the company is working on a fix, starting with a change in the name from “hot” filter to “spark” filter.
Updated at 8:23 am PT: To include comments from FaceApp CEO Yaroslav Goncharov.