Artificial intelligence or AI, was, for the most part, a technology of the future for me — something you wouldn’t lose sleep over unless you were Elon Musk, who thinks AI is “our biggest existential threat.” But for an ordinary mortal like me, blissfully oblivious to the ground-breaking research underway at Google, the status quo changed a couple of months ago when the tech giant debuted its experimental new AI — Google Duplex.
This mind-boggling version of Google Assistant makes salon appointments and restaurant reservations for you over the phone — by actually talking to people! If that isn’t remarkable enough, there’s more — the assistant accomplishes these tasks by not just sounding like any other human assistant (which, by itself, is an achievement), but by also understanding the intricate nuances of human conversations: At its annual developers' conference at Shoreline Amphitheatre where Google made the big reveal, the crowd cheered and laughed as the female version of the Assistant said, “Mm-hmm,” in response to an unsuspecting receptionist’s request to hold, while she checked for available time slots for a haircut appointment.
There were more cheers and applause, as the AI demonstrated its uncanny ability to comprehend seemingly unstructured flows of conversation. Google’s CEO, Sundar Pichai, explained that Google Duplex brought together years of research and investment by the Silicon Valley giant on natural language understanding, deep learning, and text-to-speech conversion.
As news broke about the revolutionary AI, social media users posted about the rise of Skynet, the self-aware AI that goes rogue and leads humanity to Armageddon in the Terminator, and on a lighter note, about how Duplex will be used for “legendary” prank calls; articles spoke about the ethical concerns around the tech development such as dehumanization of the service industry; and others about a “deceitful” AI that tricked people into thinking it was a human at the other end of the call. The big question on everyone’s mind though was one that had far-reaching implications for tech — Did Google’s new AI just pass the Turing test?
The big question on everyone’s mind though was one that had far-reaching implications for tech — Did Google’s new AI just pass the Turing test?
The Turing test has long been the gold standard for AI systems. Proposed by mathematician and computer scientist Alan Turing in the 1950s, the test is the outcome of an intriguing philosophical question posed by the polymath — "Are there imaginable digital computers which would do well in the imitation game?” In other words, Turing wanted to test if machines could think. But as “thinking” was difficult to define scientifically, he chose his words carefully. A machine that could think would be a quantum leap for tech which has long searched for self-perpetuating intelligence in its artificial creations. But its development may not be all gloom and doom. And Google’s AI, for starters, is still in its experimental phase — It is limited to what Google’s engineers call “closed domains”. While the Assistant may be able to handle simple appointments, more complex open-ended or abstract conversations are still out of the picture. As for building its own fleet of robots to take over the world, I’ll leave that vision of the future to sci-fi aficionados.
To be clear, Google Duplex did not pass the Turing test, as this article in Medium points out. Even if it did, as the chatbot Eugene Goostman proved was possible, it may not be anything that would help us with every-day tasks. But Google’s new AI is as close to usable artificial intelligence as we’ve ever been — one that could possibly be on our smartphones, conveniently booking a table for me at my favorite Italian restaurant. So, what a time it is to be alive!