Speaking to friends and relatives from beyond the grave is an impossible task. However, Alexa, Amazon’s virtual assistant, may soon bring that about. Whether you find that creepy or comforting is entirely up to you.
Customers will be able to customize Alexa, the company’s voice assistant, to sound exactly like their grandmother or anyone else, according to Amazon.com Inc.
According to Rohit Prasad, a senior vice president at Amazon, at a conference, the company conducted in Las Vegas on Wednesday, the online retailer is working on a technology that would enable Alexa to mimic any voice after hearing less than a minute of audio. After “so many of us have lost someone we love” because of the epidemic, the objective is to “make the memories last,” according to Prasad.
When such a feature would be released, Amazon will not say. The project explores a technological field that has come under intense examination for both its potential uses and misuse. For instance, Microsoft Corp. recently put restrictions on which companies may use its voice-parroting software. Although some are concerned that it might also be used to spread political deepfakes, the intention is to assist people who have speech impairments or other issues.
Alexa will hopefully become more commonplace in consumers’ lives thanks to Amazon’s project. But the general public’s focus has already moved on. A highly disputed claim was made by an engineer at Alphabet Inc.’s Google that a business chat bot had developed to sentience. According to numbers provided by the business, another Amazon executive stated on Tuesday that there were 100 million Alexa users worldwide since January 2019.
Prasad said Amazon’s aim for Alexa is “generalizable intelligence,” or the ability to adapt to user environments and learn new concepts with little external input. The all-knowing, all-capable, uber artificial general intelligence, or AGI, that Alphabet’s DeepMind unit and Elon Musk-co-founded OpenAI are pursuing, he claimed, “is not to be mistaken with” this objective.
Amazon shared its vision for companionship with Alexa at the conference. In a video segment, it portrayed a child who asked, “Alexa, can grandma finish reading me the Wizard of Oz?” A moment later, Alexa affirmed the command and changed her voice. She spoke soothingly, less robotically, ostensibly sounding like the individual’s grandmother in real life.While this might seem borderline creepy, but at the company’s annual event, Prasad pitched this functionality as a way to preserve memories. Amazon says that while this functionality of Alexa to mimic people’s voices ‘cannot eliminate the pain of loss, it can definitely ‘make memories last’.
If you’re interested in learning more about how this feature works, Amazon told Engadget that Alexa’s new talent can produce a synthetic voiceprint of a person’s voice after training on as little as a minute of that person’s speech. The improvements the company has achieved in text-to-speech technology are what drive it. Amazon recently released a whitepaper outlining these improvements, in which it claimed that a “Voice Filter” may be used to mimic speech for Alexa in as little as one minute.
“State-of-the-art text-to-speech (TTS) systems require several hours of recorded speech data to generate high-quality synthetic speech. In this study, we present Voice Filter, a unique extremely low-resource TTS technique that requires as little as one minute of the target speaker’s speech. It signifies a conceptual departure in the current TTS paradigm by defining the few-shot TTS challenge as a VC task and using voice conversion (VC) as a post-processing module added to an already existing high-quality TTS system, the company wrote in the white paper.
While it all seems quite noble, things aren’t as simple as they seem. The equipment used to replicate voices in deep fake videos has long caused experts’ concern. Although this skill is still in development and it remains unclear if Amazon will release it to its users globally, it does raise concerns regarding this technology being misused by scammers and cybercriminals.