Amazon is determining the best way to make its Alexa voice assistant deepfake the voice of anybody, lifeless or alive, with only a brief recording. The corporate demoed the characteristic at its re:Mars convention in Las Vegas on Wednesday, utilizing the emotional trauma of the continued pandemic and grief to promote curiosity.
Amazon’s re:Mars focuses on artificial intelligence, machine learning, robotics, and different rising applied sciences, with technical consultants and trade leaders taking the stage. Through the second-day keynote, Rohit Prasad, senior vice chairman and head scientist of Alexa AI at Amazon, confirmed off a characteristic being developed for Alexa.
Within the demo, a baby asks Alexa, “Can grandma end studying me Wizard of Oz?” Alexa responds, “Okay,” in her typical effeminate, robotic voice. However subsequent, the voice of the kid’s grandma comes out of the speaker to learn L. Frank Baum’s story.
You may watch the demo under:
Prasad solely mentioned Amazon is “engaged on” the Alexa functionality and did not specify what work stays and when/if it will be out there.
He did present minute technical particulars, nevertheless.
“This required invention the place we needed to be taught to provide a high-quality voice with lower than a minute of recording versus hours of recording in a studio,” he mentioned. “The way in which we made it occur is by framing the issue as a voice-conversion process and never a speech-generation process.”
After all, deepfaking has earned a controversial popularity. Nonetheless, there was some effort to make use of the tech as a instrument fairly than a method for creepiness.
Audio deepfakes particularly, as famous by The Verge, have been leveraged within the media to assist make up for when, say, a podcaster messes up a line or when the star of a venture passes away all of a sudden, as occurred with the Anthony Bourdain documentary Roadrunner.
There are even situations of individuals utilizing AI to create chatbots that work to speak as if they’re a misplaced cherished one, the publication famous.
Alexa would not even be the primary client product to make use of deepfake audio to fill in for a member of the family who cannot be there in individual. The Takara Tomy sensible speaker, as identified by Gizmodo, makes use of AI to read kids bedtime tales with a mum or dad’s voice. Mother and father reportedly add their voices, so to talk, by studying a script for about quarter-hour. Though, this notably differs from Amazon’s demo, in that the proprietor of the product decides to offer their vocals, fairly than the product utilizing the voice of somebody possible unable to present their permission.
In addition to worries of deepfakes getting used for scams, rip-offs, and different nefarious activity, there are already some troubling issues about how Amazon is framing the characteristic, which does not actually have a launch date but.
Earlier than exhibiting the demo, Prasad talked about Alexa giving customers a “companionship relationship.”
“On this companionship position, human attributes of empathy and have an effect on are key for constructing belief,” the exec mentioned. “These attributes have turn out to be much more essential in these instances of the continued pandemic, when so many people have misplaced somebody we love. Whereas AI cannot remove that ache of loss, it may positively make their recollections final.”
Prasad added that the characteristic “allows lasting private relationships.”
It is true that numerous persons are in critical search of human “empathy and have an effect on” in response to emotional misery initiated by the COVID-19 pandemic. Nevertheless, Amazon’s AI voice assistant is not the place to fulfill these human wants. Alexa can also’t allow “lasting private relationships” with people who find themselves now not with us.
It is not laborious to imagine that there are good intentions behind this growing characteristic and that listening to the voice of somebody you miss generally is a nice consolation. We might even see ourselves having enjoyable with a characteristic like this, theoretically. Getting Alexa to make a good friend sound like they mentioned one thing foolish is innocent. And as we have mentioned above, there are different corporations leveraging deepfake tech in methods which are just like what Amazon demoed.
However framing a growing Alexa functionality as a option to revive a connection to late members of the family is a huge, unrealistic, problematic leap. In the meantime, tugging on the heartstrings by bringing in pandemic-related grief and loneliness feels gratuitous. There are some locations Amazon does not belong, and grief counseling is one among them.