FaceApp Apologizes For Racist Filter That Gave Users Lighter Skin: AI To Blame?
FaceApp, an app for iOS and Android devices that uses artificial intelligence to edit selfies in a photorealistic way, has issued an apology for a seemingly racist filter.
The apology came from FaceApp founder and CEO Yaroslav Goncharov, a former executive of Russian internet company Yandex, who claimed that the app's neural network was the one to blame for the issue.
What Is FaceApp?
Photo-editing app FaceApp utilizes artificial intelligence in the form of deep neural networks to make changes to selfies of users.
The capabilities of FaceApp are nothing new, as it is able to make changes such as making users look old and subtle edits such as a wider smile. However, FaceApp does things differently compared to other apps in the market.
While other apps rely on base-level picture editing, FaceApp uses a neural network to make the changes look more realistic and natural. However, the results of the app vary between images, the ideal FaceApp selfie that will possibly generate the best results, has small face, little to no beard, and eyes looking straight into the camera.
FaceApp's Racist Filter
It has been discovered that FaceApp will create better images for people with lighter skin, raising racism accusations against the app. The "hotness" filter of FaceApp, meant to apply a beautification process on a selfie by making it look sharper, also makes the skin tone lighter.
"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," apologized Goncharov in a statement.
The artificial intelligence technology used by FaceApp is based on code acquired from some open source libraries, including Google's TensorFlow. However, the data set used by the app to train itself for the "hotness" filter is its own, which means that the blame falls solely on FaceApp.
The racial undertones that FaceApp learned from its neural network simply means that the development team did not provide a data set that is diverse enough for the app to understand that being handsome or beautiful does not require lighter skin tones.
Goncharov added that the "hotness" filter has been temporarily renamed to the "spark" filter to exclude the connotation of a relationship between a lighter skin tone and beauty. However, perhaps it would have been better if the filter was entirely removed in the meantime while the app's creators work on a fix.
Another Controversial Photo Editing App
FaceApp is the second controversial photo editing app in recent memory, after the Chinese app Meitu. Meitu is capable of turning people into anime characters, but it also came with shady data collection practices. The company behind the app denied the allegations that it was selling user data.
Popular messaging service Snapchat also came under fire in August 2016 for a racist "yellowface" filter that superimposes exaggerated Asian features on top of selfies, including small and slanted eyes. Snapchat later apologized for the filter, claiming that it was inspired by anime.