Facebook launches a competition for AI developers to fight “hateful memes” on its platforms

In recent months, the ongoing coronavirus pandemic has been showing exactly how social media platforms are subjective to misleading and harmful information. Facebook is also working on improving its social media platforms against misinformative content, as well as hate speech and harmful posts. Now, The Independent reports about a competition, started by the social media platform, that offers prizes to developers who develop an AI algorithm that will be capable of identifying “hateful memes”.

Facebook stated that memes are hard for artificial intelligence systems to decipher as computers do not understand contextual information the way humans do, so an image with text that’s supposed to be read together is not an easy task for AI. In order to use an AI system for this, the meaning of the image has to be combined with the text and interpreted as a whole.

To provide more data for researchers, Facebook created memes, based on popular ones being shared on the social media platforms, but using royalty-free images. The data set provided by the company amounts to more than 10,000 examples of memes. The point is to identify “hateful memes” with AI so that they can be eventually removed from the platform.



The aforementioned competition for developers is hosted by DrivenData and has a $100,000 prize pool. Additionally, in the post, announcing the news, the company states that they are continuing to make progress in improving AI in order to detect hate speech and harmful content that violates Facebook's Community Guidelines. The company adds that the new project will enable Facebook to do more in an effort to keep people safe on its social media platforms.