Facebook was the first social media platform to reach the widest market share with users all over the world. For years it was considered a giant until it became oversaturated, eventually, the hype died down. But Facebook as a company is still huge with billions of users in WhatsApp, Instagram, and Facebook. A huge problem with the oversaturation on Facebook is the spread of misinformation. It’s very easy to spread fake news and ill-informed innocent users might fall for it with the number of likes and views that it gets.
They have now decided to leverage AI to help fight misinformation and hate speech. They have also decided to ban ads and commercial listings for medical equipment to prevent exploitation. Many scammers and conmen try to gain an easy profit from the panicking population of users during the pandemic.
Originally Facebook relied on human intervention via fact-checkers and collaborators in 3rd party companies around the world to filter fake news and malignant articles. They have now supplemented this with AI technology and widely expanded their reach. Around 95% of users were successfully prevented from clicking on malicious links and misleading content.
Facebook’s AI systems have been trained via Machine Learning to detect images that appeared similar to a human eye. An example of this is to differentiate actual images from screenshots as these applications can identify the differences in minute pixels. This is called Similarity Detector on Facebook’s platform and helps reduce misinformation connected to clickbait titles from actual images.
Another addition is the content analysis tools that can interpret the content to block ads targeting Covid-19 prevention protection products. These ads target the panic surrounding the pandemic to sell low quality and even fake products to vulnerable users. To improve the detection rate of these fake products the algorithm checks for specific objects that are likely to be fake and adds them to the ban list. They also use a list of negative images to prevent inaccurate detection from similar products.
It will take numerous trials and even further errors to perfect the system enough that all sources of misleading information and malicious products can be stopped. The sheer volume of users and the data that they generate makes it a monumental task. If and when they do figure it out, it will be a great achievement for the entire field of Data Science.
