At its re:MARS conference in Las Vegas this week, Amazon showcased a new Alexa capability that will make the digital assistant capable of imitating any voice (via Reuters). The feature was demoed during a keynote segment about empathy, where the company wanted to emphasize how AI assistants can enable “lasting personal relationships” with loved ones.
Rohit Prasad, SVP and Head Scientist working on Alexa AI introduced the feature in a video showing a kid asking Alexa if his grandmother could finish reading him the Wizard of Oz. The digital assistant complies and starts reading the story in the voice of the kid’s grandmother.
The exec highlighted the “companionship relationship” users have with Alexa, and how human attributes of empathy and affect are “key for building trust.” He also explained that in the context of a worldwide pandemic that claimed the lives of so many loved ones, AI could “definitely make their memories last.”
According to Prasad, Amazon developed the capability to reproduce a human voice with just “less than a minute of recording versus hours of recording in a studio.” Depending on the way you look at it, it can be quite impressive or really concerning, especially if this technology can be used for harmful purposes, including deepfakes.
Amazon didn’t say when this new feature would roll out to Alexa. However, the company may want to think twice before putting this technology in the hands of millions of Alexa users.