Remember meForgot password?
    Log in with Twitter

article imageNew Google Duplex AI sounds exactly like a human voice

By Ken Hanly     May 13, 2018 in Technology
Mountain View - After widespread criticisms about the issues raised by Google's new Duplex system that uses AI to imitate the human voice to make appointments, the company notes that the system will have disclosure built into it.
This appears to mean that whatever final shape the system takes it will give some type of verbal announcement that it is a robot or at least AI software. Given how human the voice sounds the person on the other end may be skeptical! Two examples of the AI voice ordering a hair appointment and making reservations at a restaurant can be found here.
However, the two can also be heard in the appended video.
The demonstration of Duplex
Duplex is not yet a working project. Sundar Pichai Google CEO introduced it onstage last Tuesday at a Google developer conference only as a pre-recorded phone call. The demonstration showed how Google Assistant was able to sound much more lifelike and human with the use of DeepMind's new Wavente audi-generation technique, and other advances in the processing of natural languages.
The technique creates a robotic voice that sounds exactly like a human speaking. Duplex can even produce the sound of human breathing when speaking as well as the inevitable "uh's and "ums" that are part of most people's speech. No doubt the system can mimic Canadian speech with "eh"s.
The backlash
The crowd were impressed by the demonstration as you can see in the appended video. However, it is quite clear that in each case the person answering the phone has no clue they are not talking to a human being.
This may be a huge technological achievement but brings up ethical issues as a Verge article asks:
For example, does Google have an obligation to tell people they’re talking to a machine? Does technology that mimics humans erode our trust in what we see and hear? And is this another example of tech privilege, where those in the know can offload boring conversations they don’t want to have to a machine, while those receiving the calls (most likely low-paid service workers) have to deal with some idiot robot? The AI is not an idiot in the area of making appointments. It will know quite a bit about that. Just don't ask it what the weather is like outside or who Donald Trump is.
Zeynep Tufecki, a technology critic, tweeted a trenchant response: "@zeynep 9 May Google Assistant making calls pretending to be human not only without disclosing that it's a bot, but adding "ummm" and "aaah" to deceive the human on the other end with the room cheering it... horrifying. Silicon Valley is ethically lost, rudderless and has not learned a thing." The cheering I expect was just with respect to the ability of the AI to sound so convincingly like the human voice. No doubt it did not at first dawn on the crowd that there was an ethical issue involved.
Google reacts to critics
The company did not explicitly state in their presentation that disclosure a robot was speaking would be mandatory. Nevertheless a blog post by two engineers Yaniv Leviathan and Yossi Matias said it was important that users and businesses had a good experience with the service and transparency was a key part of that. The two added that the intent of the call should be clear so that businesses understood the context. They said that Google would be experimenting with how to do this before the system enters testing expected to be this summer.
Google representatives told Verge that the company felt a responsibility to inform individuals who do come into contact with Duplex that they are talking to a piece of software. They also said that the team was looking for ways to safeguard the product from uses such as spam calling. The outcry appears to have had an immediate salutary effect on the further development of the Duplex system.
More about Google, Duplex system, Ai