Fake news is an issue that developers are working hard to rid. They’re doing this using advanced artificial intelligence (AI) tools that can help detect suspect stories and alert them to the relevant authorities.
One such algorithm that’s been designed specifically to help in this area is called AdVerif.ai. It’s run by a startup company with the same name and works by detecting fake stories, malware, nudity, and a range of other questionable content. November saw the company launch a beta version of the software that works with advertising networks and content platforms throughout Europe and the United States that want no part in being related with fake news. Both advertisers and content platforms have a lot to lose if they’re seen to be advertising bad content.
The way AdVerif.ai works is by scanning content to look for any obvious signs of being fake such as headlines not matching the rest of the ad, or too many capitals in the title. It will also cross-check content against its already existing database of fake and real news stories that’s updated on a weekly basis. The client gets a report that shows them each piece that the system considered and the likelihood of it being fake or containing nudity or malware, or anything else that the client requested.
However, it’s not completely foolproof and some fake stories do still creep through. One such example is that involving Michael Bennett in September where a photograph was doctored to show him burning the American flag when in fact he was simply doing a “now-traditional” post-victory dance. To try and combat these few slipping through the net, fake stories are uploaded weekly so the system can learn from them.
Other startups are also offering a similar service to that of AdVerif.ai. Cybersecurity firms have been swift at jumping on this particular bandwagon in providing AI-powered services for online companies. Facebook has also been working to fight fake news by altering its algorithms to detect and banish untruthful information from its site.
The Fake News Challenge is a competition run by volunteers that was set up to encourage others to develop AI tools that could detect fake news. It was won by Talos Intelligence – a cybersecurity arm of Cisco, who developed an algorithm than was more than 80% accurate in detecting fake news.