Google Duplex: Google Shows Off Realistic Voice Capabilities Of Assistant In I/O 2018 Demo

If you thought that predicting flight delays was a neat trick, just wait and see what Google’s AI has in store for you next. The AI-driven Assistant is getting more sophisticated with each passing year and will soon be able to carry out simple day-to-day tasks for you, such as placing a call to make a restaurant reservation or book a hair appointment when you’re too busy to do it yourself.

What’s even more exciting is that it will be able to do so in a completely natural-sounding voice as Google revealed in a demonstration at yesterday’s keynote for the 2018 I/O developer’s conference in Mountain View, California.

The highlight of this year’s I/O keynote was the unveiling of Google’s AI Assistant’s newest feature, namely the Google Duplex. To showcase its astounding capabilities, Google CEO Sundar Pichai went up on stage and played back a phone call recording placed by the AI Assistant to a hair salon, reports the Verge.

The strikingly natural-sounding voice of the AI dazzled the crowd, and that’s just the reaction Google was aiming for. According to a blog post detailing the new feature, Google Duplex is designed to enable the AI Assistant to conduct “natural conversations to carry out ‘real world’ tasks over the phone.”

This is bananas. Google Assistant just called to book a haircut for someone. #io18 pic.twitter.com/gr2V0NjfiH

— Daniel Bader (@journeydan) May 8, 2018

The Google blog post explains that the objective of this technology is to facilitate specific AI Assistant tasks, such as calling people to schedule certain types of appointments.

“For such tasks, the system makes the conversational experience as natural as possible, allowing people to speak normally, like they would to another person, without having to adapt to a machine.”

And, as shown by the I/O 2018 demo, that’s exactly what it achieved. The AI-driven Assistant spoke in a very natural, human-sounding voice, interjecting an “uh” and a “hmm” here and there before replying to questions, with a perfect intonation that could barely be told apart from that of the person at the other end of the call.

This is a far cry from the monotonous, robotic tones we normally see in bots.

“The amazing thing is that Assistant can actually understand the nuances of conversation,” said Pichai.

The demo also included a call to a Chinese restaurant, during which the bot managed to respond to complex sentences and react intelligently even when the conversation didn’t “go as expected” and deviated from the given objective.

And here's a call that didn't go so well ? pic.twitter.com/MQuNfZE9SJ

— Danny Sullivan (@dannysullivan) May 8, 2018

To pull this off, the Google Duplex system has to be “deeply trained” in certain domains to help it carry out a natural conversation. According to the Google blog post, the new feature is only trained in streamlined conversation topics and can’t carry out spontaneous and general conversation.

Nonetheless, everyone is excited about the new Google Duplex, with Engadget even saying that the Google AI Assistant may have beaten the Turing test.

Although the new feature is not yet ready for release, it’s still a pretty thrilling thing to look forward to.

“We’re still developing this technology, and we want to work hard to get this right,” Pichai pointed out at the I/O conference.