I hear dead people – Alexa should imitate the voice of the dead

Amazon Echo Spot

An echo device like the spot is in many German households – the voices will soon be able to be personalized.

© Andrea Warnecke / Picture Alliance

A current project for Amazon’s Alexa is causing a worldwide sensation: the language assistant’s voice will soon be able to imitate living and even deceased people.

The news agency “Reuters” reports what Amazon currently working. At a conference in Las Vegas, the company showed that it will soon be possible to freely change the voice with which Alexa answers. And that with your own voice, that of other living people and also that of dead people.

Alexa is designed to preserve memories

The intention is “to preserve memories” after “so many of us have lost loved ones in the pandemic,” Amazon said. As an example, the company showed a film in which a child asks Alexa to read a story with the voice of the deceased grandmother.

“Reuters” attests to the short video that the grandmother’s voice actually gave the impression that it had been spoken by the real person.

The technology magazine “Techcrunch” explains how Amazon wants to do it. Using artificial intelligence, it was possible to extract enough material from an audio recording that lasted less than a minute to then form whole sentences.

Fear of very good fakes

Amazon did not respond to possible criticism of such projects. Most recently, Microsoft significantly restricted access to similar technology in order to prevent the misuse of foreign voices for so-called deepfakes, i.e. fakes that appear authentic.

Except for the presentation Las Vegas However, there is currently no information on how far the development has progressed and when the voice imitations can be expected to be available on home devices.

For Amazon, the novelty seems to be just part of the strategy the group is pursuing with Alexa. According to Rohit Prasad, Alexa boss and leading Amazon scientist, the goal is for Alexa to be more adaptable to user environments and to learn new concepts with very little input. This is “not to be confused with an omniscient, all-talented, general artificial intelligence” that companies like Alphabet or OpenAI want to create.

source: Reuters, Techcrunch