Quite controversial from an ethical point of view, Amazon demonstrated yesterday at the Re: Mars conference in Las Vegas. The company is developing a speaker that can speak to the voices of the user’s loved ones – even if they’ve already left this world.
Rohit Prasad, the company’s senior vice president and chief scientist on the Alexa team, spoke about a feature that allows the voice assistant to “clone” a specific human voice.
In the demo video, the child asked his grandmother to finish reading him a fairy tale – the plot is quite worthy of the next Black Mirror series. Alexa acknowledged the request with a normal robotic voice, then switched to a softer, more “human” version that imitated the voice of a child’s family member.
According to Parsad, the Alexa development team created a model that allows an assistant to create a high-quality speech matrix that takes less than a minute of audio to teach.
The feature is under development and it is not yet known when it will be released to the general public. While it claims the technology should be able to create “duplicates” of each voice, Prasad specifically stressed that it helps commemorate departed relatives, as “so many of us lost ones during the COVID-19 pandemic.” have that we love”. According to him, while AI cannot remove the bitterness of loss, it can definitely salvage memories.
Overall, the tech giant intends to make conversations with Alexa more natural. A number of new features are being added to the Assistant, allowing for more “humanoid” dialogue, including asking the user questions of their own initiative.