Facebook (FB) has been investing significantly in artificial intelligence and machine learning capabilities. The company will now use these capabilities to step up monitoring of hate speech and offensive content on its platform.
The latest machine-learning tool being rolled out for this purpose is an AI system called Rosetta, which can recognize inappropriate or toxic content in images, or to put it in simple terms, identify offensive memes. The tool’s capability is not restricted to memes and it can be used to recognize text in different kinds of photos like that of signboards or restaurant menus.
Fake news and misleading images have plagued Facebook in the past and the adoption of Rosetta is being seen as a meaningful step towards tackling this issue. Rosetta is currently being used on over 1 billion images and video frames on Facebook and Instagram with the aim of making photo searches more relevant and appropriate while also trying to reduce hateful content on the platform.
The tool supports several languages, including Arabic, which means it can read text both ways
Rosetta’s capability involves two steps – finding images with text and then identifying the content of the text. Apart from English, the tool supports several languages, including Arabic, which means it can read text both ways.
Although Facebook is on the right path with the implementation of Rosetta, the company still faces several challenges in refining the tool and its applications. AI tools can sometimes be ineffective in dealing with distorted or unclear images and text and can also fail to detect offensive or toxic content due to simple reasons like spelling mistakes.
Considering the fact that millions of images and videos get uploaded to its website on a daily basis, Facebook has a lot of work ahead of it with regards to Rosetta.