With audio recordings up to a minute long, Alexa will be able to reproduce human voices, to make conversations “more natural”.
On the occasion of the Re: Mars conference, dedicated to artificial intelligence and organized by Amazon this Wednesday, June 22, the company presented a new function of its voice assistant Alexa, reports the American channel CNBC. This novelty consists in making it possible to communicate with close people, even after their death.
During the demonstration, Rohit Prasad, vice president in charge of the scientific team of the Alexa program, presented a function to imitate a specific human voice. In the demonstration video, a child asks if her grandmother can read her a book.
Alexa then accepted the request using her “default” voice, a robotic voice. Then the voice assistant switched to a more human and soft voice, seeming to mimic that of the little girl’s grandmother.
One minute audio recording
This technology is possible with “less than a minute of audio recording” from the person, assures Rohit Prasad. This function aims to humanize the voice of Alexa, by imitating loved ones, but also to pay tribute to a deceased person by making the sound of their voice resonate.
This function is currently in development, and the company has not indicated a date for its implementation.
“So many of us have lost loved ones to the pandemic. While artificial intelligence cannot remove the pain of bereavement, it can, however, perpetuate the memories,” the vice president said.
Amazon wants to “humanize” its voice assistant, to make it a “companion” to relevant conversations. In 2020, the company rolled out one of its new features, which is to make him ask questions to its owners, to allow it to take part in more “natural” conversations.