More Than Just An Assistant: Should We Be Concerned About Google Duplex?

[Note on the author: Daria Kravchenko is Group Digital Marketing Manager at Hays Recruitment. Daria is currently a student on the City, University of London Writing for Business short course which started in May 2018. She wrote this blog as part of a homework/in-class exercise on that course.]

Whether you personally are a loyal fan of virtual assistants, trusting Alexa to set your alarm and handle Amazon shopping, or a sceptic avoiding Artificial Intelligence (AI) at all costs, the fact remains that a substantial amount of global effort is going into developing technology that can interact with humans seamlessly, in the same way that we do with each other.

Despite all the technological advances, progress in this field has remained limited so far. Sure, quizzing Siri on the meaning of life can bring on some giggles, but the overall language processing capabilities of existing chatbots are far from advanced. We would never assume there’s a living and breathing person on the other end of the line, when it’s just a machine providing scripted answers in a stilted computerised voice. Right?

Well, that might have been true up until now, but it looks like things have begun to change.

Meet Google Duplex

Early last month at the Google I/O conference, the company’s CEO Sundar Pichai introduced Google Duplex, a next-generation AI virtual assistant powered by deep learning and natural language processing algorithms. In the demo, which immediately made headlines, the program made two phone calls: to make a reservation in a restaurant and to book a haircut appointment at a beauty salon. In both cases, the computer sounded so convincingly natural that the person on the other end of the phone did not realise they were talking to a machine.

How did Duplex manage to achieve this? Well, not only was it able to adapt and respond when the conversation didn’t go as expected, but its speech was also heavily sprinkled with interjections like ‘uhm’ and ‘mhm’, as well as pretty convincing upward inflections people use to turn a statement into a question. In other words, the bot used a fair number of language tricks to pretend to be a human.

Friend or foe?

The demo provoked mixed feelings. However impressive it was from an engineering point of view, a number of journalists raised the question of how ethical it is, in the era of fake news, to release to the market technology that has the ability to imitate human speech so well that it can trick people into thinking they’re talking to a real person.

Following the concerns expressed in the media, Google rushed to assure the general public that transparency is key and that the assistant will always disclose to people that they are dealing with a machine. “We are designing this feature with disclosure built in, and we’ll make sure the system is appropriately identified. What we showed at I/O was an early technology demo, and we look forward to incorporating feedback as we develop this into a product,” a Google spokeswoman said in a statement to CNET.

This, however, seems to contradict common sense: why waste all the effort, resources and development work needed to litter the chatbot’s speech with fake interjections, if the goal is not to trick the listener but to perform a simple and concrete task, such as booking a restaurant table? Considering the amount of personal data Google stores on each and every one of us (think geolocation, search history, phonebook details), do we really want an AI assistant out there that can easily call somebody pretending to be what (who) it’s not?

Let’s focus on what really matters

By all means, it is probably a bit too early to worry about a chatbot calling your bank to arrange an unsolicited transfer to Gibraltar. The calls used in Google demo must have been the most successful ones. We do not know how heavily edited they were or how many unsuccessful attempts preceded them. It is also unclear whether Duplex can maintain a meaningful conversation for longer than a minute.

Nevertheless, it seems just the right time to start thinking about what regulations should be introduced around releasing such technology to the market and making it accessible to businesses, political powers or anyone else.

However cool and exciting it might be to finally see a machine that looks, walks, and talks just like us, we should remain focused on what we, as a society, should be really striving to achieve: technology that makes our lives better and easier, whether it is handling our restaurant appointments to spare us an extra half an hour to spend with our loved ones, or dealing with more serious issues such as treating cancer or preventing global warming.