Amazon is working on a way for users to talk to their family members with Alexa even after they die.
At the annual Re: Mars conference in Las Vegas, Rohit Prasad, Senior Vice President of Amazon and Chief Scientist at Alexa, announced a range of new and upcoming features for the company’s smart assistant.
The most important potential new feature was the ability to group short audio clips into longer speech. Prasad separated the feature that allows the voice assistant to imitate a specific human voice.
In the scenario presented in the event, the voice of a deceased family member of Prasad (the grandmother) is used to read a bedtime story to a grandson.
Using the new technology, Prasad notes, the company is able to achieve impressive audio output using just one minute of talk.
“This required inventions as we had to learn to produce high-quality sound with less than a minute of recording versus hours of recording in the studio,” he said. The way we’ve achieved this is by framing the problem as a sound transduction task rather than a speech generation pathway. We live in the golden age of artificial intelligence, where our dreams and science fiction have become reality.
Prasad explained that the feature is currently under development. Amazon also did not say when it might be released to the public. While they could ostensibly be used to imitate any sound, Prasad suggested they could be used to help memorialize a deceased family member.
Amazon plans to make Alexa imitate anyone’s voice
Making the AI conversation become one of the primary focus areas, especially given that so many of us have lost someone during the coronavirus pandemic, Prasad said.
“While AI cannot eliminate this pain of loss, it can make memories last,” he added.
And the e-commerce giant wants to make talking with Alexa more natural in general. It has introduced a series of features that enable its voice assistant to repeat more human-like dialogue, even to the point of asking questions to the user.
This work is an area of technology that has been closely scrutinized for potential benefits and abuses.
Microsoft recently restricted companies that could use its software to imitate sounds. The goal is to help people with speech disabilities or other problems. But some fear that it can also be used to spread deep political fakery.