One of the methods Amazon demonstrates to improve Alexa involves making the voice assistant seem more human. While Alexa has certainly become a better conversationalist, learning when to whisper and offering a follow-up mode, it still has a distinct and slightly-robotic tone. Yes, even for Alexa celebrity voices.
But soon Alexa could be able to speak through a filter of a more human-like voice, specifically the voice of your departed loved ones. Rohit Prasad, senior vice president and head scientist of Alexa AI, demonstrated this experimental feature during Amazon's annual MARS conference (opens in new tab) this week (via The Verge (opens in new tab),) claiming its AI can mimic a person's voice using a single minute of recorded audio.
While it could've demoed the feature using anyone alive and perhaps even a person with a recognizable voice, Amazon instead opted to have Alexa narrate a bedtime story to a child using the voice of his deceased grandmother.
Though this choice is questionable, it seems to illustrate the advances in humanizing the Alexa experience. I'm not saying I support disillusionment, but connecting a user to someone they've lost could deliver an otherwise priceless experience. As long as it's within reason, of course. Even to hear the voice of someone in a different time zone, or your favorite show character, could make Alexa communications feel more authentic.
Sure, copying someone's voice with a mere minute of learning or intake marks a creepy progression for AI. But is it a surprising progression? Not really. Deepfakes, which were coined back in 2017, are responsible for conjuring up convincing-yet-fabricated visuals.
And maybe you saw the Metaphysic founders who auditioned for America's Got Talent with an unbelievably persuasive Simon Cowell AI (opens in new tab). I felt equally amused and unsettled by the act, not because it it felt like a magic trick but because I know the current state of AI technology more advanced than many give it credit for.
In other words, Alexa gaining mimicking abilities was inevitable. At the MARS conference, the feature was not promised for the best Alexa speakers or the Alexa assistant consumers currently use on their smartphones, headphones or other connected Alexa devices. If this capability does eventually come to Alexa, it probably won't be for some time.
I'm curious to see if, how many and in what ways users leverage such abilities. I personally don't need to hear the voice of someone who's passed, but I wouldn't diminish the value that might hold to another person.