The filter should have made the faces more attractive, but in the end just made them whiter.
The former Yandex employee and creator of the FaceApp application Jaroslav Goncharov apologized for the filter that made users whiter. It is called “Hot” and, according to the idea of the creators, should turn faces into more attractive ones.
According to Goncharov, the problem with the application arose due to the “side effects of the neural networks.”
The FaceApp application appeared in January
The application for processing photos using FaceApp neural networks began to rapidly gain popularity in Russia in early 2017. The FaceApp service allows you to add a plausible smile to a gloomy person in a photo, to turn a man into a girl, and anyone who wants to dream up how he will look in old age.
, but only in the spring began to gain viral popularity. First of all – thanks to the filters that turn the user into an old man, a child and a woman. The results of processing their own and others’ pictures in social networks published hundreds of people.
In addition, thanks to the built-in neural networks, the service puts a smile on the images, which makes the photo look weird or weird. April 15, FaceApp came in first place among the most popular applications in the Russian App Store, and on April 20 – on Google Play.
We deeply apologize for this serious problem. This is an annoying and side error installed in the neural network service.
Initially, this was not supposed, the conflicting data in the files for training the neural network are to blame. We are working to completely fix the problem.Jaroslav Goncharov, the creator of FaceApp
The creator of FaceApp added that the filter was renamed and is now called “Spark” (“Spark”) in order to “avoid all positive connotations”.
On April 25, on Twitter, we noticed that FaceApp whitens the faces of black and Asian users.