One of the world’s biggest internet companies went back to the drawing board after highly advanced technology failed to tell the difference between African Americans and wild animals.
The Daily Mail shared a video of a verbal and physical conflict between multiple black men, a white male, and police on June 27, 2020. At the end of the video an automated Facebook message appeared, asking if Facebook users wanted to “keep seeing videos about primates.”
The message came despite the content having nothing to do with apes, monkeys, or lemurs. BL understands the social network’s artificial intelligence (AI) software had confused one large African American with a gorilla.
When mainstream media reported extensively on the error, Facebook went into damage control.
“This was clearly an unacceptable error,” Facebook spokeswoman Dani Lever said in a statement obtained by USA TODAY. “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.”
Lever blamed AI technology’s limitations for being “not perfect” and admitted her colleagues have “more progress to make.”
“We apologize to anyone who may have seen these offensive recommendations,” she said according to the paper.
This is not the first time AI technology incorrectly labeled content. In 2015, Google apologized after the company’s “Photos” application incorrectly identified black people as “gorillas.”
A Brooklyn-based programmer had complained about the program labeling photos of him and a black friend as powerfully built apes with large heads and short necks.
Google+ Engineer and Chief Architect Yontan Zunger promised the “Photos” team would quickly find a solution. In the end they chose to completely remove the “gorilla” tag.
Facebook previously lifted a ban on pro-life advertising for displaying a “graphic or violent” human fetus back in 2019.
“In this instance we made a mistake in applying a warning screen over the image used in The Iona Institute’s ad,” a representative said at the time. “We have removed the warning screen and apologise for any inconvenience caused.”
YouTube separately flagged a well-known Croatian chess player’s video as racist, and suspended his channel, for discussing white versus black pieces. After realizing the error, the video sharing website restored the channel one day after.
BL understands Big Tech employs AI to tag, recognize and warn users about different content. However, the technology still has many flaws.