GOOGLE HAS USED all of its machine learning and artificial intelligence (AI) tech to enable Google Assistant to make phones call for you.
Called Google Duplex, the new feature taps into Google's efforts in natural language recognition to enable Google Assistant to call places such as salons and restaurants and make booking or queries on the user's behalf.
At Google's I/O 2018 event, where the firm also detailed its upcoming Android P release, chief exec Sundar Pichai showed how Google Assistant can phone up and book a salon appointment based on a user's data by seemingly talking naturally with the human on the other end of the line.
Duplex will also automatically call businesses during public holidays to figure out when shops will be open, a move to stop people from inundating stores with calls and allow it to update its search listings for the shops in respect to the latest information.
It all sounded a bit too good to be true, with the Google Assistant sounding a lot more human than it does in our Android phones or Google Home, but Pichai claimed the calls 100 per cent legit and took place in the real-world, not test environments.
During the demo (above), Google Duplex hit a brick wall when someone seemingly didn't understand Assistant's rather clear request to book a table at a restaurant. In this case, the smart tech realised it had hit a brick wall and politely but swiftly got off the call - we suspect that there are more gremlins and hiccups behind the scenes that need to be ironed out which didn't crop up in the slick stage demo.
While Duplex looks like some interesting use of AI tech and makes it easy for people who hate making calls to people they don't know, it's also creepy as heck.
The tech is effectively pretending to be a human and hoodwinking those on the other end of the line, even inserting 'uh-huhs' and pauses to sound more convincing. Depending on your moral compass and prejudice towards robots, you'll either be shrugging or declaring that nobody got time for that.
Furthermore, Duplex will work in the background making calls and booking based upon a users data and calendar appointments, which treads the line between useful and intrusive, if such a line exists in the first place.
There's also the potential for the tech to be used by crafty developers or savvy hackers to automate AI-powered prank calls if it's not sufficiently secured.
Pichai noted Google is still working on the tech so there's no official launch window for the feature.
"We're still developing this technology, and we want to work hard to get this technology and the expectations right," he said.
We doubt Steve Wozniak will be too chuffed with the tech, but then again he popped up on The Big Bang Theory so who gives a f**k about his opinion.
While gaining the ability to trick people into thinking it's a real man/woman isn't likely anytime soon, the Google Assistant is getting six new American accent voices, including that of crooner John Legend.
Using its WaveNet machine learning tech, Google took choice words and phrases from the signer and used it to make up a full vocabulary in order to add his voice to the Assistant.
The same tech has been used to make the Google Assistant sound more natural and conversational, rather than just responding to queries prefixed by 'Hey Google', at least that's according to Pichai, who noted the Assistant will be better at handling multiple questions in one.
For example, Pichai asked it: "Who was the Governor of California when Kevin Durant was drafted and what team drafted him?" And the Assistant answered: "Arnold Schwarzenegger was the governor in 2007. Kevin Durant was drafted by the Seattle SuperSonics."
And to top it all off, the Google Assistant will get a setting called Pretty Please that will have it respond to "please Google" when kids want to ask it a question in order to encourage people's sprogs to be more polite.
Welcome to the tech dystopia, please wipe your feet before coming in. µ
What could possibly go wrong...
Committee clams firm failed to implement 'adequate security'
Meme Ban means Meme Ban
It's anonymous data at first but the NYT figured out how to make it personal