Rohit Prasad, Amazon’s head scientist for Alexa revealed on the second day of the re: Mars event that soon Alexa will be able to speak anyone’s voice, dead or Alive. He said, his team is working on the new feature where Alexa can mimic anyone’s voice by listening to a one-minute recorded clip. Alexa will be able to speak an hour-long speech in this clip.
He showed a glimpse of the feature by representing a video clip where a child asked, “Alexa, can grandma finish reading me the Wizard of Oz?”
In return, Alexa confirmed the request in its default Robotic voice and started narrating the story in the child’s grandma’s voice.
“This required inventions where we had to learn to produce a high-quality voice with less than a minute of the recording versus hours of recording in the studio. The way we made it happen is by framing the problem as a voice-conversion task and not a speech generation path,” Prasad said.
Prasad highlighted the loss of many lives in COVID 19 and said this could be a way to last memories. He said, “While AI can’t eliminate that pain of loss it can make their memories last”.
There was no information revealed about the launch of this new feature. This technology is not the same as an all-knowing, all-capable AGI or artificial general intelligence, on which every tech-giant is working and trying to perfect it.
He also highlighted how the company is trying to make conversation with Alexa better and in a human-like way.
How do Alexa Conversations work?
To fill the gap between the wide array of possible conversations and what you can compose manually, Alexa Conversations leverages AI. AI extrapolates the range of phrasing variations and dialog paths for you based on a few sample dialogs you supply that indicate your desired dialog paths and templates for the APIs you’ll need to call. Additionally, the AI will be in charge of managing the dialog state and context, which includes carrying context from turn to turn, keeping lists, and assisting with corrections. With less development work required thanks to Alexa Conversations, you can devote more of your attention to developing high-quality experiences for your customers rather than flowcharts.
Alexa conversations work by generating training data and end-to-end conversational management architecture. A proprietary engine applies algorithms to the sample data that then generates thousands of dialogs between Alexa and the user. These datasets include many elements to provide a range of dialogs between the two. A novel model trains a deep-learning algorithm that can predict the conversation based on history, current state, and the API of the developer. With the help of API calls through the skill, it can drive the conversation ahead. It uses AI to manage the highly complicated elements of the conversations for the developers.
Since Alexa conversation works on the training data, the training dialog is given in the new dialog editor and then annotated with Dialog Acts, Utterance Sets, and Responses. They can specify to invoke APIs for the required arguments so the dialog manager can trigger the skill code. The debugging tools help to improve the conversation experience. Skills created with Alexa Conversations can be published using the current tools and procedures for certification and publication, and Alexa Conversations can be included in already-existing custom skills. Amazon provides a guided tour with fully functional templates to get you started with the dialog annotation process and Alexa skill-building.
User’s Reaction
While the intention of amazon behind the voice mimicry is to help people cope with the loss of a dear one, most Alexa users are not liking the idea. The reaction has been negative and people feel a mimicked voice will not make them feel better after their loss.
While some are tensed about the potential threat voice mimicry can have, others find it unethical to mimic a dead person’s voice without their consent.
FAQs
-
How does alexa work ?
Alexa is always ready to answer questions, tell jokes, play music, find the fastest route to work, adjust the thermostat with your compatible smart devices and much more. We continue to add new features to make Alexa even more helpful to you.
-
Can you get new voices for Alexa?
Tap on the device you want to change Alexa's voice on. Tap the Settings cog wheel button in the upper right. Scroll down to the option for Alexa's Voice and tap on it. Choose either Original (the feminine voice) or New (the masculine voice).
-
Can Alexa respond to two different voices?
Amazon's Alexa can now recognize different voices and give personalized responses.
-
Can i use alexa to record conversations ?
Alexa does not actively record and store all your record conversations, but it's always listening for "Alexa," the wake word.
-
How many voices can Alexa learn?
Alexa speakers can be trained to recognize up to 10 different voices, ensuring everyone gets a tuned experience.
