Voice assistant from Amazon: Alexa should speak with grandma’s voice

Status: 06/24/2022 2:30 p.m

An audio sample of less than a minute should be enough for a new software for Amazon’s language assistant to imitate the human voice. But experts warn: The potential for abuse is high.

The US technology group Amazon has developed software that can imitate a human voice with less than a minute of speech as a template. This was announced by Amazon manager Rohit Prasad yesterday at a company conference. The voice assistant Alexa can thus imitate any voice.

The aim is to create lasting memories after so many people have lost loved ones during the pandemic, Prasad said. Artificial intelligence can help. The online company demonstrated how a networked loudspeaker with the language assistant Alexa reads a boy the book “The Wizard of Oz” with the voice of his grandmother.

In an interview with technology blog Techcrunch, Prasad stressed that the grandmother from the demonstration was alive. It wasn’t about a “dead grandma,” he emphasized.

Sale to consumers still unclear

From a technical point of view, the new software represents a leap. Until now, it was necessary to speak several hours of text for usable results from electronic voice imitation. However, it is still uncertain whether the technology will actually reach consumers. Amazon did not provide any information on this.

Experts pointed out possible dangers. Siwei Lyu, a professor of computer science at the University of Buffalo, expressed concern about the software’s potential for abuse. For example, criminals could pose as family members or try to influence the stock markets with bogus statements from top managers, he told the financial news agency Bloomberg.

The software would also make it possible to imitate the voices of politicians or public figures. This could facilitate the spread of false information and message manipulation.

Microsoft only wants to deliver software to certain customers

The US group Microsoft is working on a comparable software. Natasha Crampton, head of the company’s AI ethics department, recently commented on this in a blog post. Microsoft wants to set strict guidelines for using the software, the manager announced. It should be restricted which customers are allowed to use the service.

According to Crampton, the technology has great potential. But it is also easy to imagine how it could be used to deceive listeners.

source site