On Wednesday, FaceApp released a string of ethnicity filters, such as Asian, Caucasian, Indian, Black and so on, allowing users to change their appearance. However, after massive backlash from users, the feature was discontinued on the same day, but the damage is already done.
Neural network wasn’t at fault this time
Earlier this year, users forgave FaceApp once after the “hot” filter was released, which could be used to lighten the skin tone. At the time, the company was quick to take down the filter, with CEO Yaroslav Goncharov apologizing for it, saying that it was an “underlying neural network.” However, the ethnicity filters this time cannot be even distantly related to an artificial intelligence mistake.
This time again, Goncharov was at the front to defend the decision, stating that there are no positive or negative connotations linked to the filters. The CEO told TechCrunch that these filters are designed in a way that they are equal in all aspects.
However, following massive amounts of criticism, the CEO later said that the “new controversial features will be removed in the next few hours.” The CEO told CNNMoney that the feature would be removed from the company’s servers entirely so that users will not have to update their apps to remove the new option.
Racial parodying has proven costly for many other apps too. Last year, Snapchat was also subjected to backlash for creating filters which could change a face into a black or yellow face. But for those thinking that Snapchat learned its lesson, it did not. In August of the same year, Snapchat released another controversial feature known as the “anime” option, which turns the faces of people yellower and makes their eyes narrower and their teeth bigger. As expected, the company was again reprimanded for the filter, notes Slate.
FaceApp not learning from mistakes
Earlier this month, the company added paid selfie styling effects which allow users to beautify the appearance of their selfies. Paying for beautifying apps does not sound very convincing, given the fact that there are hundreds of such apps that offer a series of effects for free.
However, it is worth noting that FaceApp introduced some of the most radical filters, such as gender swap, which was quite popular. Goncharov has not disclosed any new usage metrics for FaceApp, but last week, the company said since its initial release in January, it has had more than 45 million downloads.
At the time, when asked if they have fully fixed the race controversy issue, Goncharov said that removing algorithmic bias needs continuous monitoring.
“When dealing with such complex algorithms and neural networks, it would be naive and even irresponsible to declare such an issue completely fixed. It is an ongoing process,” Goncharov told TechCrunch.
Further, the CEO said that FaceApp takes these issues seriously and keeps this in mind at all phases of the product cycle, i.e., creating data sets and training neural networks, quality assurance, and processing customer feedback. He further stated that the style filters are meant to preserve ethnicity origin, and tests conducted shows that they are doing quite well.
However, the latest ethnicity filters do suggest that Goncharov and his company have not learned from their past mistakes.