“Hello? Hello, uh, I’m calling to reserve a spot. This is an automated booking service from Google, so the call is being recorded”. You immediately hang up. Again and again. The fourteenth today. Google announced its futuristic Duplex function on the I/O conference in 2018, which took place in may. Duplex — this is essentially a Google-helper, which calls for, for example, in a Barber shop or restaurant to sign up from your face.
The tech world is excited. People portend a new era of artificial intelligence, able to deceive an unsuspecting person on the other end. However, pop and serious ethical concerns. Does the person who answers the phone says that with the AI? Does the AI position of the person calling?
Two months later, Google has some answers. The company showed a new video, which the bot uses inserts like “um” or “uh”. Even the author called Recode.
At first glance, Google got rid of the problem may demo. Assistant Google now calls himself an “automated confirmation service”. And Yes, it worked.
Google also said that Duplex is deployed on a limited list of the companies with which Google has partnerships, and on the “trusted list of subjects”, however, is not particularly going into details.
There is another question that Google needs to answer before the Duplex will become commonplace: what problem does it actually solve? Surely people can’t spend two minutes of your life to book a table? Is the robot who calls up restaurants, so necessary in the era of online backup? What is going on?
And even if the Duplex will find a compelling reason for its existence, whether he actually had a chance? There are many examples of technologies that have existed for a long time before they are finally accepted by the public. Take, for example, the device Echo from Amazon — they were introduced in November 2014, but they became popular only in 2017, when it sold 22 million units. Some technologies, like smart glasses, no success and not acquired. Google Glass: not found the way to people’s hearts, and Snapchat desperately trying to convince people to use their points.
And yet Google is in full swing developing a project Duplex. The company is struggling to make her AI sounded indistinguishable from an ordinary person. When the voice sounds too “like a robot”, as noted by Scott Huffman, Vice President, technology Google Assistant, “it doesn’t work. People do not go on contact if a voice sounds unnatural”.
Solution? Add “deficiencies in speech.” A short pause filled with coughing and interjections, have significantly improved the results on Google, making the AI more friendly sounding and just.
What are the objectives of Google? The company will have to solve contradictory tasks: to earn the trust of those who do not want to be deceived, talking to the AI, and make the voice of the robot is indistinguishable from human.
Despite the furor, made Duplex, Google may face difficulty convincing people to use this technology (if it will ever make it public). For most people the conversation with the robot on the phone, regardless of how vital, it is an unnatural experience that requires you from unnatural approach — at least, according to modern technological standards.
There is nothing surprising in the fact that someone working in a Barber shop or restaurant, just hang up, shrug off the robot. About the same we vyprovazhivaya Jehovah’s witnesses and other Herbalife.
Even if Google is able to accurately reproduce the human voice to bring us out of uncanny valley, people can not immediately accept this technology. Even the Internet, which has existed since the end of 1980-ies, has entered the mainstream only in the mid-90s. E-mail has been around for nearly half a century, but in the active use of entered recently.
The main thing is not work with Google Glass. But in this case at least the phone can be put.
Caller robots from Google is cool. But why do they need?
Ilya Hel