As the world grows more connected, it has also given all kinds of people a platform to broadcast and share their views and thoughts. However, some of these views can be considered to be hateful and racist, and also with the ease of sharing, it is often quite easy to accidentally share something that could be outright false or perhaps wrong.
With the number of users they have and with the reach they have, platforms like Facebook have a duty in which they need to curb such posts, at least as far as their platform is concerned, which is why Facebook has turned to the use of AI to help enhance their ability to detect hate speech and misinformation in an attempt to curb it.
How Facebook plans to do this is by creating an AI that has the ability to detect things like near-identical images, which to a human will look different, but to a computer might look the same, which is how some people skirt around Facebook’s current system by uploading images that have been ever-so-slightly altered.
To achieve this, Facebook still needs the help of human moderators who will first flag an image as containing false claims or hate speech. The system will then seek out near-identical images which moderators can then apply warning labels to. The idea is that the AI will apply a similar logic/approach as a human moderator, thus allowing Facebook to scale up the process and to make it faster.
To that end, Facebook has also announced what they’re calling the Hateful Meme Challenge, where if you can develop an AI system that can identify and label these memes, you could stand to win yourself as much as $100,000. More details about Facebook’s approach and the challenge can be found on its blog.