Amazon’s Alexa now has the ability to imitate multiple voices, which the company said can be used to have it speak in the voices of dead relatives.
The feature was showcased at Amazon’s annual MARS conference last month in Las Vegas, according to The Verge.
It’s unclear if the feature will ever be made public.
A video at the MARS conference showed a child asking Alexa to read a story in the voice of his dead grandmother.
“As you saw in this experience, instead of Alexa’s voice reading the book, it’s the kid’s grandma’s voice,” said Rohit Prasad, Amazon’s head scientist for Alexa AI.
Amazon is working on a feature that lets Alexa mimic the voice of dead family members using short voice clips. Yeah, it can read you stories in Grandma’s voice. Would you wanna try it or too creepy?
Here’s a glimpse at how the feature works (source: Amazon re:MARS keynote) pic.twitter.com/zKjeIdcZkA
— Beebom (@beebomco) June 24, 2022
Prasad talked up the concept, saying that giving “human attributes” to AI was important “in these times of the ongoing pandemic, when so many of us have lost someone we love.”
“While AI can’t eliminate that pain of loss, it can definitely make their memories last,” Prasad said.
Would you want this capability in your house?
Some on social media were not sold. One user tweeted: “Creepy.” “Morbid.” “Monstrosity.” Amazon Alexa Creeps Out Internet With Voices of the Dead.”
This is very, very creepy and disturbing.#Alexa #AmazonAlexahttps://t.co/5Fcuesm4Dy
— Mahmoud “Moody” Shabeeb?? (@ItisMoody) June 25, 2022
“Creepy.” “Morbid.” “Monstrosity.” Amazon Alexa Creeps Out Internet With Voices of the Dead https://t.co/0IFdavFM3S
— Rsnews (@Rsnews19Rsnews) June 24, 2022
No, Amazon, I don’t want Alexa to bring my relatives back from the dead https://t.co/B8ItfbjhnP
— Side effect (@Agenorchannel) June 25, 2022
Just because we can do something doesn’t mean we should. This Alexa feature is too creepy for me https://t.co/Snkra9Ioc2
— T3.com (@T3dotcom) June 25, 2022
Microsoft has set limits on how voices can be created and used, according to NBC.
“This technology has exciting potential in education, accessibility, and entertainment, and yet it is also easy to imagine how it could be used to inappropriately impersonate speakers and deceive listeners,” Natasha Crampton, who heads Microsoft’s AI ethics division, wrote in a blog post.
Amazon said one minute of recorded audio is enough to do the trick. That’s a massive change from current technology in which Amazon has had celebrities spend hours recording audio, according to Variety.
Prasad said engineers were able to cut down the time “by framing the problem as a voice-conversion task, and not a speech-generation task.”