Facebook’s $1 million ‘Deepfake Detection Challenge’ winner registers 65.18% accuracy

Facebook announced the winners of its $1 million “Deepfake Detection Challenge”, which is exactly what it sounds like— a contest for AI models to detect doctored deepfake videos that “convincingly” imitate real people. However, as the winning model scores 65.18 per cent accuracy rate, it looks like there is much work that needs to be done before we rely on such algorithms to spot deepfake content.

With over 2,000 participants, the competition saw more than 35,000 AI models attempting to detect deepfakes from a black box dataset of 100,000 videos. Facebook reportedly hired more than 3,000 actors to create these video clips where they were recorded holding conversations in natural environments.