Social networking authority Facebook on Thursday made it official that a new artificial intelligence, or AI-based systems that will serve as main proponents of detecting hate speeches and misinformation, or more known as "fake news" that continues to spread all throughout the site's platform.
The enhanced system is projected to show improvements in the coming days as users of the Facebook app may get to experience this newly developed feature by the AI system, although it would take a while to fully maximize this. Slow and gradual improvements are expected by Facebook Chief Technical Officer (CTO) Mike Schroepfer, who mentioned a small portion or impression of the social media's overall automated systems which may be not affiliated with the mentioned issues surfacing Facebook.
The Components of Facebook's New AI System
Technology systems such as the ObjectDNA are heavily integrated to the continuing development of the AI. This technology revolves more on targeting the specific key objects primarily seen on an image, and denies the clutter or background around it, allowing the system to be more accurate as it focuses on less subjects than being involved on possible multiple target points.
ObjectDNA's participation on the system was revealed on the study that was published this year, named the 'An Analysis of Object Embeddings for Image Retrieval'. This would open the possibility for Facebook to locate and analyze copies of claimed objects and visuals that may resemble an image that they already banned or pointed out, despite it may be different from the ones flagged by the system.
Another technology that is now a part of Facebook's AI system is the 'LASER' - a system that helps review semantic similarities of sentences and statements, which utilizes its cross-language sentence-level embedding capabilities. Despite being a text-based content reliant system, images could also be catered by LASER.
A new tool developed by Facebook itself that is purposely made to manage and oversee scanning of over a billion posts a on the social network per day as estimated by the company. They call it the "Linformer" (derived from the combination of two words "linear" and "transformer"), which instead of exact calculations by the system, it will approximately rely on the central attention mechanism of transformer-based language models.
Hate Speeches and Misinformation is not just on Facebook, its Everywhere
All of these new additions by Facebook could really help the social media company to further provide a more monitored surveillance across the platform, detecting hate speeches, unnecessary words and texts, and wrong information that is seen widely not just on Facebook but all over the different social media platforms and sites available on the internet today.
Last week, fellow social media site Twitter named world-renown hacker Peiter Zatko, with the moniker "Mudge" as their new head of security. This move was done by Twitter as ongoing charges are being presented to both Twitter and Facebook concerning issues of account hacking, hate speeches, fake news, and 'biased' opinion-based points of with the two social media sites used as a tool to execute it.