Facebook apologizes after artificial intelligence labeled a black man’s video “exceptional”. According to The New York Times, users who recently watched a video featuring black men published by the Daily Mail were asked if they “wanted to watch a primate video.” The social network apologized for the “unacceptable error” in a statement sent to the publication.
It also disables the recommendation function, which is responsible for messages, as it looks for causes to prevent serious errors like this from happening again.
Facebook Spokesperson Statement
Company spokesman Danny Lever said in a statement: “As I said, although we have improved our AI, we know it is not perfect and we need to make further progress. We apologize to everyone who saw this offensive recommendation. “
Gender and racial bias in artificial intelligence is rarely a unique problem in social networks – facial recognition technology is far from perfect and tends to misidentify POS and women in general. Last year, a fake facial recognition game led to the illegal arrests of two black people in Detroit.
In 2015, Google Photos tagged photos of black people as “gorillas,” and Wired discovered a few years later that the tech giant’s decision was to censor the word “gorilla” from search and image tags.
The social network shared a dataset it created a few months ago with the AI community to address the issue. It contains more than 40,000 videos with 3,000 paid actors sharing their age and gender with the company. It is very good sign that Facebook apologizes on mistakenly labelling black man as primates.
Facebook even employs professionals to light up their photos and highlight their skin tone so that its artificial intelligence system can learn what people of different races look like in different lighting conditions. That dataset is clearly not enough to completely resolve Facebook’s AI bias, which further shows that the AI community still has a lot to do.