AI technology independently deletes terror propaganda on Facebook
Facebook relies on artificial intelligence (AI) in the fight against terrorist propaganda. According to the company, 99 percent of posts linked to terrorist groups such as the Islamic State (IS) or Al-Qaeda are detected and deleted using automated technologies before they are even reported by users.
The new method should not only work for texts, but also for images and videos, as the Pressetext news agency writes. "Our efforts to combat extremist content on our website are now much more systematic than in the past. We work together with various partners and use different techniques internally", quotes BBC News from a Facebook statement. In principle, the company continues to rely on a mix of human reviewers and software. "However, automated systems are increasingly taking over the primary control," says Facebook.
According to the online network, these use special AI algorithms to detect unwanted propaganda material. "We use text-based machine learning, where software is taught over time to work out for itself whether certain postings could be a problem or not." It works in a similar way with image and video material. Each photo or video has a unique digital fingerprint that can be recognized immediately once it has been classified as problematic. "This allows the content in question to be deleted quickly," emphasizes Facebook.
AI not yet error-free
However, Facebook's AI system is not yet working flawlessly. "It will take many more years of development to perfect this automatic control," admits Monika Bickert, Facebook's Global Policy Chief. For example, the software currently only focuses on IS and Al-Qaeda. "We chose these groups because they pose the greatest threat globally. If we also want to include other terrorist groups, it's not that easy because there can be very big linguistic and stylistic differences in their propaganda," explains Bickert.
In the future, it should also be possible to use AI technology not only to monitor content on the site, but also to analyze behavior and classify what is behind individual accounts. "But we know that it's a cat-and-mouse game. Those who want to take advantage of the enormous reach of social media sites will do everything they can to circumvent automatic controls," concludes the Facebook expert. (pte/hae)
