AUM logo CMYK-02-01
By Adrian • June 1, 2018

Google’s Duplex: Exciting or Worrisome? Part II

In part one of our two-part article, we provided an overview of what we know so far about Duplex – Google’s Artificial Intelligence device. In this second and final part, we do two things: discuss how Duplex is beneficial to corporations and the public in general; and consider some of the criticisms Duplex has met with since it was unveiled on the stage of Google I/O 2018.


How users will benefit from Duplex in the near future

google duplex part 2 image 1As Google fine-tunes the current iteration of Duplex, businesses opting to make use of the device will undoubtedly benefit from quite a few of its functions. For instance, businesses which lean heavily on appointment-making but haven’t yet taken their operations online can make/receive bookings through Duplex. This service includes options such as reminders for appointments, cancellations, rescheduling, and various types of inquiries.

Let’s look at this last function a while longer. Customers often call businesses to inquire about information that is not available online, such as hours of operation during a holiday. Now, however, a prospective customer need only interact with Duplex, which will itself call the business to inquire about service hours or other matters, and then duly make the information available online with Google. This will reduce the number of such calls businesses receive, while at the same time making the information accessible to nearly everyone.

In addition to calling businesses on their behalf, Duplex can help users overcome issues such as limited connectivity, language barriers, and hearing-impairment. One article in TNW (The Next Web) even applauded Duplex for its potential to help people suffering from social anxiety. In the author’s own words: “I was experiencing a certain kind of hope for the first time in my life. […] It felt like maybe, in my lifetime, I might finally stop feeling pressured to use the telephone. And that was a surprisingly powerful feeling. […] this could improve quality of life for people like me (even if it’s not a fit for everyone).” 

Google, of course, is one of the big tech players pouring resources into bridging the gap between humans and A.I. by continuously enhancing the latter’s verisimilitude, and increasing the number of scenarios and options available. Google maintains that, with innovations such as Duplex, it is improving the quality of people’s day-to-day interaction with computers. Yet Duplex has its detractors.


Controversy and criticism

Google Duplex part 2 image 2As with just about anything groundbreaking, Google’s public unveiling of Duplex resulted in an audience split right down the middle. Countering the wowed onlookers was an equally audible lot who, eyebrows raised, brought up quite a few ethical concerns. What was initially only verbal quickly congealed into more fleshed-out analysis in quite a few tech review articles.  

One of the first criticisms raised happens to relate to the “naturalness” of Duplex. Among other characteristics, the use of interjections such as “um” and “uh” does indeed make this chatbot more lifelike. Consequently, human speakers on the other end of the line are often duped into believing they are talking with a fellow human.

This could pose a problem when a call is recorded for legal reasons. Should the human speaker say something incriminating but fail to realize at the time that he is not talking to another human, he might later claim to have been subjected to a form of entrapment. And an A.I. device or program cannot be held legally accountable. With this in mind, one might get spooked by the whole phenomenon. Given the scandal of Facebook and Cambridge Analytica sharing people’s personal information, we’re also left to wonder how much Google wants simply to provide a service – without using potentially deceptive technology to gather private information on the sly.

The fact that Duplex can mimic just about any voice raises the related concern of misrepresentation. During the conference, Google showcased Duplex’s ability to carry out a conversation in an uncannily close approximation of John Legend’s voice, based on snippets of his songs with which it was provided.

For its part, Google addressed this matter in a post-conference statement: "We understand and value the discussion around Google Duplex – as we've said from the beginning, transparency in the technology is important. […] 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." 

The move to have Duplex disclose its A.I. nature prior to engaging in a conversation is indeed a laudable one –  or it would have been had this significant shift not come in response to misgivings evinced by critics. While it’s fair enough to encourage companies to adjust their strategies and approaches in accordance with feedback, it still leaves unanswered the question of how much megacorporations such as Google take ethical issues to heart in the design and development phases of their technology.

In other words, yes, we’ve reached a significant milestone in the progress and integration of robotics and A.I. But with the ominous messages imparted by dystopian literary classics such as Aldous Huxley’s Brave New World and George Orwell’s 1984 still fresh in our minds, and with Duplex’s shortcomings the subject of increased scrutiny, some of us understandably remain wary.

Indeed, the critical reaction to Duplex continues unabated, with this writer having a few observations of his own. As it happens, two of the phone interactions Google showcased at the public unveiling left me wondering if Google was overplaying its hand somewhat when claiming how natural and lifelike Duplex is, even in limited conversational contexts.

(Video time mark:  0:57 -1:14) Consider this example, in which Duplex is speaking with someone I presume is a non-native speaker of English trying to get a reservation at a restaurant. During the conversation, the A.I. device issues a few decidedly incongruous statements, the result of failing to properly process the human speaker’s responses.

(Video time mark: 3:48 - 4:39) Here, in another example, the human speaker at the end of the line is clearly deceived into believing that he’s speaking to another person. Duplex’s inability to modify its speech patterns and cadences once the conversation is underway increases pressure on the speaker to change the way he speaks. Notice how he goes from a casual tone to a more formal one, taking his cue from Duplex.

It might be a little premature to judge Duplex’s lack of flexibility on this score, given that we know the device is still in its infancy. However, considering that in the examples presented above, Duplex sounds natural enough to pass for human but nowhere empathetic enough towards human interlocutors, I cannot help having qualms about whether Google intends to make good on its promise that Duplex will facilitate all manner of communication services for users.

One more sticking point when it comes to Duplex is as much socioeconomic as it is ethical – given the potential fallout for people employed in the service sector. In a poignant article that appeared in The Verge , the author asks an eye-opener of a question: “When did human service workers become Google’s to experiment on?”

The author mentions two real-life instances that illustrate her point. In the first case, she and her colleague spend a week testing out Facebook M by using it to order breakfast every day from the same restaurant. After a few days, the restaurant owner asks them to place their order directly with him, as Facebook M is using a service called Eat24, which takes a cut of the profits.

What we see here is that, precisely because of its effectiveness on the technical level, the A.I. voice phenomenon exemplified by Duplex renders unnecessary any direct communication between people demanding services and those providing them. In addition to the resulting literal objectification of many a service provider (previously human but now a computerized device), this may lead to a reduction in wages for those men and women lucky enough to keep their jobs in the sector in question. After all, if a company’s profits decrease because an intermediary is taking a cut (as with the aforementioned Facebook M/Eat24 case), management may decide not only to lay off many of its employees, but to slash the salaries of those it retains.


While we wait to hear more about Google Duplex...

Let’s retrace, for a moment, the trajectory of our two-part examination of Google Duplex. First, we provided an overview of how this new strand of A.I. works. Once we laid this foundation, we looked at what Duplex, having gone public, promises businesses in the way of communication features. Finally, we dove into points of contention that have arisen in reaction to this chatbot’s unveiling.

As students at AUM, it’s important that you learn to approach innovations in the field of A.I., however exciting they are, from as many different angles as possible. This not only makes you better students within the walls of academe, but equips you to grapple with life in the “real world.” So even as you remain receptive to A.I.-related advances and trends, make sure to examine them with a critical eye, one that allows you to put matters in perspective and weigh their ramifications for both society and the individual.

Contact AUM