We’re fortunate that deepfake videos aren’t a big problem. The very best deepfake detector to emerge from a major Facebook-led effort to combat the modified videos would just capture about two-thirds of them.
In September, as speculation about the risk of deepfakes grew, Facebook challenged artificial intelligence wizards to develop techniques for finding deepfake videos. In January, the company also banned deepfakes used to spread misinformation.
Facebook’s Deepfake Detection Difficulty, in partnership with Microsoft, Amazon Web Services, and the Partnership on AI, was go through Kaggle, a platform for coding contests that is owned by Google. It offered a huge collection of face-swap videos: 100,000 deepfake clips, developed by Facebook using paid stars, on which entrants checked their detection algorithms. The job drew in more than 2,000 participants from industry and academia, and it produced more than 35,000 deepfake detection models.
The very best design to emerge from the contest found deepfakes from Facebook’s collection simply over 82 percent of the time. But when that algorithm was tested against a set of previously unseen deepfakes, its efficiency dropped to a little over 65 percent.
” It’s all great and great for helping h
