People can usually tell when a video of a famous politician speaking is actually an AI-generated deepfake, but when listening to the audio or reading the purported text transcript, it becomes harder to distinguish fake from real.
“In my opinion, audio deepfakes are a bit risky at the moment because visual deepfakes are still hard to create,” he said. Aruna Sankaranarayanan At Massachusetts Institute of Technology.
Sankaranarayanan and his colleagues collected text recordings, audio and video of political speeches…
Source: www.newscientist.com