Editor’s Note: Eliza Strickland is a former SSRC Abe Fellow.
In September, Facebook sent out a strange casting call: We need all types of people to look into a webcam or phone camera and say very mundane things. The actors stood in bedrooms, hallways, and backyards, and they talked about topics such as the perils of junk food and the importance of arts education. It was a quick and easy gig—with an odd caveat. Facebook researchers would be altering the videos, extracting each person’s face and fusing it onto another person’s head. In other words, the participants had to agree to become deepfake characters.
Facebook’s artificial intelligence (AI) division put out this casting call so it could ethically produce deepfakes—a term that originally referred to videos that had been modified using a certain face-swapping technique but is now a catchall for manipulated video. The Facebook videos are part of a training data set that the company assembled for a global competition called the Deepfake Detection Challenge. In this competition—produced in cooperation with Amazon, Microsoft, the nonprofit Partnership on AI, and academics from eight universities—researchers around the world are vying to create automated tools that can spot fraudulent media.
[…]
Source: Facebook AI Launches Its Deepfake Detection Challenge IEEE Spectrum – IEEE Spectrum