At the recent Amazon re: MARS conference, Rohit Prasad, senior vice president and chief scientist of Alexa demonstrated a new technology they have been working on that allows their voice assistant to mimic any voice. All they need is a sound clip of the voice, about one minute should be enough, and then Alexa will be able to speak with a similar voice.
The on-stage demonstration pointed out that it can be used to imitate the voice of a deceased relative to remember them, which drew many reactions.
The demonstration shows a child asking Alexa if “Grandma” can read him a bedtime story. Alexa listens, and a voice is heard that sounds at least like that of an elderly lady. As you can see in this example, it’s not Alexa’s voice reading the book, but the voice of the child’s grandmother, said Rohit Prasad, Amazon’s chief researcher for Alexa AI.
Prasad also said that adding “human attributes” to artificial intelligence has only become more important during the pandemic because so many have lost someone close to them.
“Although artificial intelligence cannot alleviate the pain of a loss, it can definitely make memories last,” Prasad said. Advanced audio programs also have similar functions for cloning voices based on recordings. This feature can be used in a podcast recording. As long as the clip is short and simple, it will often not be possible to hear that it is “fake”.
