At its re:MARS conference on Wednesday, Amazon showed off a new feature that can imitate any voice with a one-minute recording. The firm would like to help users keep in touch with their deceased loved ones.
This week is the event re:MARSwith a series of conferences dedicated to the and space. In Wednesday’s keynote, Rohit Prasad, Senior Vice President of Amazon, outlined his vision for the future of the Alexa. One of the features Amazon is working on is the ability to imitate someone’s voice.
This is one of the controversial uses of Artificial Intelligence, because of the risk of abuse with. However, this feature can also be beneficial, such as who gave back his voice to Val Kilmer for the film Top Gun : Maverick. To illustrate how it works, Amazon showed a clip of a little boy asking have her grandmother read her a story. The voice assistant then modifies his voice to imitate that of his grandmother.
Watch Wednesday’s re:MARS keynote, featuring Rohit Prasad’s talk starting at 43:36. In English, enable automatic translation of subtitles. © Amazon
A system that works with a recording of less than a minute
The idea would be to be able to prolong the relationship with a deceased. The firm’s AI is able to reproduce a voice with a recording that lasts less than a minute, while current techniques require several hours of recording in a studio. Amazon has not indicated a date for a possible launch of this function, nor if it will be available on all devices equipped with the voice assistant Alexa.
Rohit Prasad also underlined the complexity of Alexa which integrates more than 30 decision systems based on machine learning. However, the firm would not aim, also called Generalized AI (AGI), a form of AI capable of rivaling human intelligence. Instead, Amazon prefers to talk about Generalist Artificial Intelligence (or “ generalizable), able to adapt to different environments and learn new tasks and concepts.