Facebook apologized after it mislabeled a video of Black men as “primates,” the latest in a series of racial gaffes by artificial intelligence systems that technology companies use to automate recommendations and other features.
Facebook users who recently watched a video from The Daily Mail featuring clips of Black men in altercations with white police officers and civilians received a prompt asking if they would like to “keep seeing videos about Primates,” the New York Times reported late Friday.
“This was clearly an unacceptable error and we disabled the entire topic recommendation feature as soon as we realized this was happening so we could investigate the cause and prevent this from happening again,” Facebook spokesperson Dani Lever said in a statement to USA TODAY.
“As we have said, while we have made improvements to our AI we know it’s not perfect and we have more progress to make,” she said. “We apologize to anyone who may have seen these offensive recommendations.”
Facebook often touts its artificial intelligence that it trains using images uploaded by users.
This is not the first time that a technology company has come under fire for racial bias in its automated systems.
In 2015, Google apologized after its Photos application mistakenly identified black people as “gorillas.”
Another incident a year later had Google image searches for “three Black teenagers” showing mugshots and “three white teenagers” showing smiling white teens. That same year, Microsoft’s AI chatbot Tay began spouting racial slurs and had to be yanked offline.
Other signs of racial bias creeping into technology products include Instagram filters that lighten skin or fetishize ethnic features and Snapchat filters that use blackface or caricatures of Asians.
___
(c) 2021 USA Today
Distributed by Tribune Content Agency, LLC.
