A new feature of Alexa, Amazon’s virtual assistant, will allow to hear again the voice of those who have passed away. This was announced by Rohit Prasad, an Amazon engineer who at the recent re:MARS event showed a video of a child asking an Amazon Echo speaker for his grandmother to finish reading him the story “The Wizard of Oz”.
With less than a minute of recording Alexa could already play back a person’s voice
The voice that then sounded through the smart speaker happened to be that of an older woman, turning out to be a reconstruction by digital synthesis of the voice of the child’s grandmother, now deceased. The re:MARS event is focused on machine learning, automation, robotics and space and has served as a framework to showcase this revolutionary technology. It could be described as a deep fake vocal.
To achieve a satisfactory result at Amazon they opted not for a voice generation but for a conversion. In this way, instead of needing hundreds of hours of recording a voice with the highest quality (preferably recorded in a studio), a few seconds, barely a minute, could be enough.
This would open the door to recovering the voice of deceased people from old recordings of their voice, but obviously it could also serve, as in the case of the videos deep faketo falsify the statements of celebrities, politicians… Or, not to be too dramatic, to generate funny jokes among family and friends by making them say words they have never uttered.
From Amazon it is stated that while artificial intelligence seeks to emulate human attributes such as affection and empathy in order to generate trust among human users, the application of these models to a feature such as imitating the voice of the dead generating new speeches or oral interactions would not serve to eliminate the pain caused by the loss but at least it can make the memories of a loved one who has passed away last.