Your Alexa speaker could soon imitate your deceased loved ones’ voice by leveraging AI (artificial intelligence). Speaking at the company’s annual global AI event, re:MARS, Amazon Senior VP and Alexa Head Scientist Rohit Prasad demoed the feature with the voice of a deceased grandmother reading a bedtime story to her grandson.
Alexa will use one-minute voice clips to create high-quality and longer voice clips aided by AI
The feature requires at least one minute of audio from the source, which Prasad claims can offer decent audio output. Alexa would effectively synthesize the audio clips provided by the user and turn them into lengthier speech, TechCrunch reports (via Gizmodo).
Prasad said the company “required inventions” to develop a high-quality audio clip with only a minute of voice data. “The way we made it happen is by framing the problem as a voice conversion task and not a speech generation path. We are unquestionably living in the golden era of AI, where our dreams and science fictions are becoming a reality,” he added.
Amazon didn’t say when the feature would officially roll out. But it’s safe to say that the company is only testing the waters with its announcement. The response so far appears to be mostly negative. There are also some unanswered questions.
We’re unsure if this feature could be used negatively, maybe in the form of a cloned voice to imitate another person. Synthesizing the voice of a dead loved one is not something that everybody wants. But as is the case with technology, there are no limits on innovation. This can often lead to features or products that are downright creepy.
The feature will almost certainly be opt-in, considering the level of data sharing involved
Since there’s no release timeline, it’s safe to assume that the feature will not be available anytime soon. Moreover, considering the kind of data sharing involved, it would require interested users to opt in. So customers should be able to ignore this if they don’t want to bring up memories of deceased people.
Amazon holds the re:MARS (machine learning, automation, robots, and space) conference each year to showcase its products in the field of ambient computing. The event takes place in Las Vegas, Nevada, beginning on June 21, 2022, and lasting until June 24, 2022.