At Google I/O 2018, it showed us what it could do. It showed the AI calling a hairdresser and setting up an appointment. The catch: The person taking the appointment didn’t know it was talking to a computer.
This is both amazing and creepy. According to the company, it is a tool that uses Artificial Intelligence (AI) to accomplish ‘real world-tasks’ over the telephone. But thinking that the tech giant is pouring millions of dollars into something that makes your appointments or reserves that table, wouldn’t that be too naive? The tech is bound to expand into other areas.
So how does it exactly work? It is just like any other task you will order the assistant, but this has a twist. Per AndroidCentral: ‘Duplex is using what’s called a recurrent neural network. It’s built using Google’s TensorFlow Extended technology. Google trained the network on all those anonymised voicemails and Google Voice conversations you agreed to let it listen to if you opted in with a mix of speech recognition software and the ability to consider the history of the conversation and details like the time of day and location of both parties.’
Of course, duplicating the human speech one hundred per cent is still a far-off goal. There are far too many variables to consider. Of course, if impersonating a human becomes a reality, the future is really tense.
But one question we need to ask: do we trust Google enough? You see, one of the main plus (or negative) points of Internet is the anonymity. With technologies like Duplex, if there is no full disclosure, will we know whether we are talking to a human or machine?
After all, AI has its negatives, too. The recent investigation following Trump’s election has opened up a whole Pandora’s box. From paid users spreading misinformation on social media (and thus trying to influence opinions) to deepfakes, which are AI-generated videos of entirely new facial expressions of a target person created by stitching together two faces in an eerily convincing way. And good old photoshop is not far behind.
Amidst such a minefield, it is but guesswork to understand the implications of what may happen if such a tech falls prey to hackers. Alongside veg/non-veg classification will we have to see human/machine classification, a la Blade Runner?
According to the company, it is a tool that uses AI to accomplish ‘real world-tasks’ over the telephone.