We all are on social media for the memes. They¡¯re funny, inspiring and the perfect thing to calm our nerves while staying locked indoors due to the novel coronavirus. However, there is another side to these memes where they can be used to spread bullying and hateful information on social media.
While AI exists on Facebook to stop any explicit or sensitive image or video from being visible to everyone, the same technique doesn¡¯t really apply to memes as they¡¯re a lot more dynamic. However, Facebook wants to put an end to hateful memes. It is asking developers around the world to help them fight hateful memes by developing an AI algorithm to accurately identify the memes.
Facebook states that memes are hard for AI to decode as computers don¡¯t really get contextual information like humans do (yet), so a text which otherwise reads fine but could have gross hateful content could go undetected.?
To provide data to researchers, Facebook has created a meme database of sorts that includes meme text with the actual images replaced by similar content images from stock image platform Getty Images.
The idea behind this project is to create a tool that can take the different modalities -- image as well as test and analyse them together in the classification process. Basically, training AI to read text and image like a human would and finally decide if it¡¯s hateful or not.
Facebook has also stated its definition of hateful, stating, ¡°A direct or indirect attack on people based on characteristics, including ethnicity, race, nationality, immigration status, religion, caste, sex, gender identity, sexual orientation, and disability or disease. We define attack as violent or dehumanizing (comparing people to non-human things, e.g., animals) speech, statements of inferiority, and calls for exclusion or segregation. Mocking hate crime is also considered hate speech.¡±
Facebook has teamed up with DrivenData -- a platform that hosts such challenges for data scientists. The prize pool for the team who tries to crack this code will be awarded $100,000 which is approximately Rs 77 lakhs.
In case you or your team were interested, click here to apply.?