Deepfakes harness artificial intelligence to produce videos that aren’t real. They’ve been used to entertain, or to embarrass public figures, and it’s usually easy to tell they’ve been fabricated. A group of engineers in Canada hopes to master the most challenging part of creating a believable deepfake: generating a synthetic voice that sounds convincing. They’re racing the giants of their industry to create what they hope will be the perfect deepfake, integrating high-quality synthetic audio with doctored video. Meanwhile, researchers and legal experts are watching these developments with dread. They warn that the risks of unleashing this kind of technology are devastating. The well-timed release of a convincing deepfake could spark violence, or swing an election — further eroding public confidence in governments and institutions, and our fundamental ability to believe in what we see. (Episode 14 of 30)