Facebook apologized after his AI slapped a blatant label on a video of black men. According to the New York Times, users who have recently looked at a video posted by daily mail with Black Men have seen a prompt to ask them if they wanted “[K] EEP see videos on the primates”. The social network apologized for the “unacceptable error” in a statement sent to the publication. It is also disabled the recommendation functionality that was responsible for the message it examines the cause of avoiding serious errors like this to occur again.

The spokesperson for Dani Lever said in a statement: “As we said, as we have made improvements to our AI, we know that it’s not perfect, and we have more progress to do. We apologize to anyone who may have seen these offensive recommendations. “

Gender and racial bias in artificial intelligence is barely a unique problem with the social network – facial recognition technologies are still far from perfect and tend to identify POC and women in general. Last year, false faces of facial recognition led to the unjustified arrest of two black men in Detroit. In 2015, Google Photos labeled the photos of the Blacks like “Gorillas” and Wired found a few years later that Tech Giant solution was to censor the word “gorilla” search and image labels.

The social network shared a dataset created with the AI ​​community in order to combat the issue a few months ago. It contained more than 40,000 videos with 3,000 paid actors who shared their age and sex with the company. Facebook even hired professionals to illuminate their shot and label their skin tones. AI systems can learn what people from different ethnic groups look like various lighting conditions. The dataset was clearly not enough to completely solve the AI ​​biases for Facebook, demonstrating as well as the AI ​​community still has a lot of work ahead.