Those weird “loved ones come back from the dead to visit you” stunts pulled by celebrities like Kanye West may soon become a reality through your digital assistant.
At Amazon’s re:MARS conference, the company announced it’s working on a feature that can synthesize short audio clips of a person’s voice and then reprogram it as longer speech. Amazon’s Senior Vice President and Head Scientist for Alexa, Rohit Prasad, showed off a demonstration where, as TechCrunch described, “the voice of a deceased loved one (a grandmother, in this case), is used to read a grandson a bedtime story.”
Prasad noted that the company can do this sort of audio output with merely a minute of speech before continuing: “The way we made it happen is by framing the problem as a voice conversion task and not a speech generation path.”
There aren’t many more details beyond this initial demonstration. Reuters reports that Prahad mentions the goal of this technology is to “make memories last” after “so many of us have lost someone we love,” which makes it seem rather intense.
It does bring to mind the Takara Tomy smart speaker we reported on a few months back, which could imitate a parent’s voice at bedtime so they could read to their kids even when physically unavailable. It’s a neat idea for parents who often travel for work or are just sick of reading the same book. But it raises the question of if a deep fake copy of someone’s voice can offer the same comfort and security as a warm-blooded person in the room.
To that end, this kind of technology also brings up other questions, like what kind of emotions would hearing the voice of a loved one lost to COVID or some other incurable disease evoke in a person? Wouldn’t it just make them angrier their person is gone? And how does this interaction with the deceased work with someone actively working through grief? Does it interrupt the process or help it along? Further, what would the deceased think of it?
There’s, of course, also always the security risk of deep fake impersonation, although it’s unclear how the feature’s voice samples will be stored.
Amazon’s re:MARS conference is meant to highlight the company’s ventures in ambient computing, which includes Alexa’s functionality. The “MARS” stands for machine learning, automation, robots, and space. It’s expected that Amazon will be sharing more news bits at its Las Vegas conference over the next few days.