Facebook is now using an artificial intelligence to identify malicious memes.
Under pressure from both politicians and users to better police content shared on its platforms, Facebook has developed a new, machine-based learning program called “Rosetta.” In a blog post last week, the company outlined how the AI helps Facebook identify content that does not conform to the company’s “community values.”
Text laid over pictures (“memes”) and videos are ubiquitous on social media. According to Facebook, the problem of recognizing the relationship between text and a corresponding image “is quite different” from the comparably simple task of sorting through text-only posts.
Rosetta tracks rectangular areas that may contain text and then scans them through a recognition system, making the text readable by machines. It can detect a plethora of languages other than English, including languages that read from right to left.
Rosetta puts that text into a “recognition model” trained to understand the relationship between text and related images. The machine then deciphers the context of the text and image simultaneously. In theory, Rosetta can determine the intentions behind a meme, whether that be humor or hate speech.
The software will automatically flag hate speech in posts that violate Facebook’s community guidelines. Facebook says it will use this system not only to ensure that memes do not promote discriminatory language but also to make the site more user-friendly for the visually impaired and to determine which posts should surface higher on personalized news feeds.
According to Facebook, Rosetta, which is already live, examines text from more than a billion public Facebook and Instagram posts on a daily basis.
Facebooks says it plans to expand the number of languages used by Rosetta and make it more effective in monitoring videos. Currently, Rosetta’s application to video is limited, as the process of monitoring every video on Facebook and Instagram would take an enormous amount of computing power.
The blog post, noting how social media evolves constantly and rapidly, described the company’s efforts at policing user content as an exciting but “ongoing challenge.”
Image from Pexels.com