Facebook Is Now Using Artificial Intelligence To Censor Hate Speech On Memes
This year social media sites have been cracking down on offensive material, and the party has just begun. Facebook has just deployed a new A.I. on its platform that is capable of analyzing memes, images, and video frames to detect language that it believes violate its terms of service.
The system is called Facebook Rosetta, and it’s an impressively intelligent A.I. that uses a multi-layer model (seen below) to read text on image and video content, analyze its contents, and then interpret if it should be filtered. Each identified word receives a recognition score that determines how offensive the word may be. Which words Facebook is censoring is unclear.
More than one billion media items are uploaded on Facebook each day, and Rosetta will supposedly analyze all of them. Facebook hopes that this will reduce the workload of the moderation team, which has been working a sweat censoring fake news, misleading material, violence, and more.
Also Read: Does Social Media Make People Meaner?
Censorship has been a huge topic in the tech industry lately, with Facebook, Apple, Twitter, and several other major platforms banning Infowars radio host Alex Jones from using their services.