By: Joseph Marcinowski
The way the public consumes news online needs to change. With the rise of social media and proliferation of deep fakes, it’s not difficult to convince an audience of thousands of people to believe a lie. The fake news problem may be graver than you might think; statistically, 86% of internet users have fallen for one or more of these tactics, and the consumption of fake news is now threatening American democracy.
Manual verification isn't capable of moderating the vast number of articles and posts created by bots, and fact-checking is tedious and cumbersome. Additionally, posts which have evolved to appear as legitimate as possible often fly under the detection radar for long periods of time. It doesn’t take long for these false stories to cause irreparable damage because reposts can cause a cascading effect. These articles can only be stopped when a moderator catches wind of them, and sometimes the damage has already been done.
An emerging technique involves using artificial intelligence to determine the truthfulness of a post. This is done by analyzing word patterns using automated cognitive intelligence. The AI method is essentially similar to the way that parents and teachers preach about being wary of articles that contain typos and poor grammar; AI technology has the capability to pick up on very subtle hints that improve the classification. Using the AI tool, fact-checkers can be notified of articles that might otherwise go unnoticed, so that online misinformation can be caught before making a lasting impression.
One of the most frightening forms of fake news– deep fakes– can also be detected using AI. In a 2019 research paper, computer scientists developed a deep-fake classifier that detected pixel manipulation. The state-of-the-art classifier was not only able to pick up a deep fake with 95% accuracy (with humans being 50% accurate), but the model was able to identify the section of the video that was manipulated, as well as the tactic used.
Citations:
“AI: The Modern-Day Defender against Misinformation.” Sutherland Global, 22
June 2020,
https://www.sutherlandglobal.com/our-thinking/ai-the-modern-day
defender-against-misinformation.
Villasenor, John. “How to Deal with Ai-Enabled Disinformation.” Brookings,
Brookings, 9 Mar. 2022,
https://www.brookings.edu/research/how-to-deal-with-ai-enabled
disinformation/.
FaceForensics++: Learning to Detect Manipulated Facial Images.
Comments