The ethics of Google Duplex
Google Duplex is an AI assistant from google, which hit the news in summer of this year. Duplex is able to call up businesses and book appointments for you. For example, it can make your dinner reservations or haircut appointments over the phone while you carry on with other daily tasks. It is a real enhancement of the smartphone’s ability to make your life easy, convenient and fast. While it is expected to be released relatively soon, no official release date has been announced.
The AI has been fitted with incredibly human sounding speech, including discontinuities like “um” and “er”. It also has a selection of different voices, ranging across age, gender and nationality. Its understanding of the nuances of conversation and problems which may arise in bookings are better than ever before. This has been achieved through years of research and the implementation of deep learning technology. Watch Duplex in action here.
As you can see, there is definitely the potential to confuse the voice on the phone for a human rather than a robot, unless you already know you’re talking to Duplex. For these reasons, Google have announced that the AI will identify itself as a “google assistant” at the beginning of any phone calls.
This decision has come after some uncertainty as to the ethics of Google Duplex in the months following its press coverage. There has been, predictably, a fear of robots and AI becoming more human than ever before, and some people have described this as “disconcerting” and even “creepy”. The fear of an AI being indistinguishable from a human has long been around, exacerbated and blown to the extreme by Hollywood. There is a clear anxiety about the power and agency that this could grant already powerful AIs.
However, we are a very, very long way away from the point at which AI could become out of our control (if this ever happens at all). There is also a huge benefit to the AI sounding traditionally human. If a robotic voice is heard over the phone the automatic response usually is to hang up – we are all used to automated sales or spam calls. A more human sounding voice invokes empathy and so means that the receiver is more likely to stay on the phone, and comply with requests. It is also important to remember that Duplex can only converse within a certain context, and cannot make conventional conversation, so its human believability has an end point.
And what about the convenience of Duplex for those with speech impediments, those who don't speak the local language, or those with anxiety disorders? A common anxiety symptom is a nervousness or even fear of making or picking up phone calls, and obviously this brings up a lot of hurdles in everyday life. An AI like Google Duplex means these issues can be dealt with effectively.
Many have argued that we are beginning to rely too heavily on technology in order to live our lives, and Google Duplex is a prime example of where this heavy reliance is reflected. Are we becoming too dependent? Perhaps. But it is important to remember that Duplex can only go so far. It cannot be used as a complete replacement for things like calls with family members or phone-interviews, for example, because of the fact it must identify itself as an AI at the beginning of the call, and its inability to stray from certain contexts.
The benefits which this kind of technology could bring are all too often forgotten in the midst of fears and apprehension. While we are a long way off from a world in which AI literally walks among us, this is an amazing step forward for widespread commercially-used Artificial Intelligence.