Google Duplex was one of several interesting concepts revealed at the 2018 edition of its I/O. Though the keynote lasted just over 10 minutes it was the one that generated the most talk. This leap forward in Artificial Intelligence (AI) is clearing new frontiers.
Bots have been used to complete certain tasks for some time now, and with machine learning and other technology progressing so rapidly, this should only continue. Customer support services for online casinos and other industries, getting basic medical advice and many other communications are the obvious next step. If chatbots had human voices and cadences that would be even more feasible – which is just what this latest AI aims to do.
Everyone already uses virtual personal assistants to find nearby services or directions to wherever it is they need to go. The word “duplex” actually means to have two parts, and the intention with this new technology is to have it work as the second part, in tandem with the existing assistant in Android’s operating system.
Essentially, this will be a bridge between you and Google Assist; the quick wits of a human with the capacity and methodological nature of a machine. The tech company describes it as capable of “conducting natural conversations to carry out ‘real world’ tasks over the phone”; it can interact in different languages, make ums and ahs as real people do, and take initiatives that other chatbots cannot.
If something is out of stock online, for example, the new software will call the store to check if it can be found there. Using human speech sounds is, apparently, the most efficient way to interact. Tools are even being developed to create fully-fledged bots from individuals’ voices.
To start with, the bot will be limited to closed “domains”; beta testing will focus on booking hair appointments, making reservations at restaurants and enquiring about holiday hours. The idea is for the AI to run in the background, and to report to you when tasks are completed.
You make the request, then carry on with something else or even go offline, and are notified when things are done. If the software can’t manage, it will tell Assist or hand the communication over to you.
For people with patchy reception, very busy lives or impairment such as hearing loss, the benefits are obvious. But how far can it, and should it, go?
While developers say that the idea is to blend your real- and digital-world interactions so that you are more present and benefit in every way, critics suggest that not engaging with actual human beings could have the opposite effect.
If you’re not dealing with a person there’s no need to moderate your frustration and rudeness, and no feeling bad about cancelling reservations at the last minute. Besides the disconnection from humanity resulting from this greater sense of entitlement, like most tech this naturally neutral system could be abused.
What if a version with a croaky voice is developed, so that you can use it to call in sick, or you are duped into a sales call, the kind that you usually terminate as soon as you hear the robot voice on the other end? And what will happen when the bots can speak to each other? Will it be an added convenience, where your AI speaks to your friend’s AI to set up a meeting? Or will they develop their own language and conspire for world domination?
In 2016, a paper was prepared for the White House by the National Science and Technology’s Subcommittee on Machine Learning and Artificial Intelligence. Potential job losses were acknowledged, but new careers and opportunities were also highlighted.
Despite this optimism, however, the idea of huge fallout for assistants, secretaries and even middle management is reinforced by these new developments. And as the machine learning continues could they even, for example, develop interview skills and replace reporters?
Automation in other industries is also a rising cause for concern – think about what will happen to the livelihood of Uber drivers when they are no longer needed thanks to self-steering vehicles. None of this is happening tomorrow, but it doesn’t feel very far off.
Ultimately, we are going to have to deal with better AI in a responsible way. Governments might need to introduce guaranteed income levels for all, and the reach of something like Google Duplex may continue to extend to limited domains only. As individuals, we should constantly upskill ourselves so that we are always useful and taking initiative – keeping Artificial Intelligence working for us, rather than the other way around.