Gradient Photo Editing App Criticized Over 'Racist' AI Face Feature
Photo editing smartphone app Gradient is under fire on social media for a new advertised feature that lets people view an image of themselves in various ethnicities. While the Gradient app hasn't exactly gone viral yet, the initial feedback from the online community appears to either be simply having fun with images, or deeply concerned about the potential racial bias at play.
The Gradient app was launched in the Google Play and App Stores in the fall 0f 2019 by Ticket to the Moon, Inc. Ran by two artificial intelligence and machine learning enthusiasts, it initially began as a facial editing app utilizing AI for more effective manipulation of selfies and other social media images. The developers soon added more trivial features for fun, such as the ability to turn one's image into a classical portrait. On the next update released in March, the Gradient app could scan a person's face and predict their DNA ancestry followed by a feature that predicts what animal a person most closely resembles. However, team Gradient soon began to venture into more risqué territory, offering people on the app the ability to scan themselves and receive a 'beauty score.' This past week, the app's latest update uses AI and facial recognition to morph an image and demonstrate how that person would look in 'Europe,' 'Asia,' 'India,' and 'Brazil.'
According to Gradient's website, this new feature, called AI Face, allows a person to, 'find out how you would look if you were born on a different continent.' The website and screenshots of the app on Twitter reveal that several of the locations offered are actually countries, not continents, including Brazil and India. While India and Asia are two separate choices available on Gradient to manipulate one's skin tone and face shape, it's important to note that India is in fact a part of Asia. This sort of broad association of various cultures and ethnicities has already started to garner negative publicity for the app, especially after haphazardly using B-List reality stars to promote it on social media.
So far, many people using the Gradient app seem to try it once for fun to see how they look as a person of other ethnicities, then move on. However, others do not see it as innocently and are more concerned about bias the AI itself encodes. Reality personalities Scott Disick and Brody Jenner both Tweeted their own images using Gradient's AI Face feature and were both very clear in their Tweets that it was a paid ad. However, Twitter was quick to attack, questioning the app's intentions while expressing outrage toward the images. Both Disick and Jenner disabled comments after being bombarded with accusations of racism and promotion of 'blackface.' Rather than simply jump on someone for promoting an app for money, some more informed individuals in the AI community called for constructive action.
Professor Anima Anandkumar, Director of AI research at Nvidia, Tweeted that while the Gradient app clearly uses racial bias which is unwelcome, it could serve as an effective opportunity for Twitter to perform a controlled study of its own AI cropping technology. Prof. Anandkumar used the app herself in several Tweets to exploit the AI racial bias Twitter performed in which the photo with the darker skin was never picked as the candidate for featured crop.
While Anandkumar still used her platform to point out the clear racial flaws of the Gradient app, Anandkumar has also implored Twitter to use the opportunity to correct its own racial bias convincingly present within its AI. Should the Gradient app gain more momentum and move over to a red hot app like TikTok would certainly lead to more controversy. For now, however, the Gradient app (while clearly flawed) is still flying under society's radar for the most part.
Source: Gradient, Prof. Anima Anandkumar/Twitter