A cop doesn’t have to tell you they’re a cop, but should a robot have to tell you they’re a robot? If you aren’t worried about it yet, maybe you should be.
At its annual developer conference, Google showed off a jaw-dropping demo of “Google Duplex,” a feature where the AI-powered Google Assistant will call actual real humans at real businesses to do things like make appointments and schedule dinner reservations for you. The examples are assuredly cherry-picked examples of the AI’s absolute peak performance, but they’re impressive nonetheless.
This is obviously a dream come true for those of us who detest picking up the phone, and while it seems like it would be polite for Google Assistant to announce its true nature as a robot, this particular flavour of deception is pretty small fry. But of course that isn’t where the possibilities end.
There’s already a prevalent robocall scam designed to trick its victims into saying ‘yes’ with the goal of using that clip to authorize fraudulent charges. It’s not too hard to imagine how a more sophisticated and fluid AI caller might be able to extract audio clips or personal information in a more nuanced way, all with the victim being none the wiser.
On the other hand, this sort of tech could also be pointed back at human scammers, to keep them bogged down in calls designed only to waste their time. All this, combined with advances in audio editing that let software synthesise speech in virtually any voice, open a whole Pandora’s box of possibilities.
For now, the technology is still fairly primitive. Chances are that, in practice, Google’s AI callers are much less smooth than they are in these curated on-stage demos. But the fact that these conversations are happening at all speaks to how rapidly we’re speeding into the unknown—an unknown we won’t recognize until we’re already there.
You can read more about how Google Duplex works and hear more examples of it in action on The Google Blog.
First published on Popular Mechanics USA