It’s not Google Duplex, it’s you

Tito Sarrionandia
4 min readMay 13, 2018
He would use Duplex

In Asimov’s Robot series, several societies uphold the custom that robots must introduce themselves with the prefix “R”. For example, Daneel Olivaw, the humaniform robot so convincingly lifelike that it can deceive most actual humans, introduces [it/him]self as R. Daneel Olivaw.

If you’re not particularly interested in tech ethics, you’ve probably seen the video of Google’s AI assistant making a convincing phone call on behalf of its human owner to book a haircut, a technology they are calling Duplex. If it really works as well as the demonstration, it’s pretty incredible.

If you are interested in tech ethics, you’ve probably also seen a take like this one from CNET’s Bridget Carey.

Bridget (and everyone else on YouTube) identifies the primary ethical issue as being the same one identified by Asimov: that human beings have a right to know when they are and aren’t interacting with other human beings.

I think this is far fetched. If you’ve ever used Twitter you probably know that ship has sailed, and if anything this technology makes me think that voice calling as a format will die a slow death as the number of humans making calls shrinks compared to the number of robots making sales, research, and political calls, and restaurants get a few bad experiences with empty tables caused by unclear conversations with robot assistants of varying quality. Especially since it sounds like the production version will identify itself somehow during the call.

That’s a shame, because there is an ethical issue here, it’s just not the kind of broad, scary, political ethics that we’re used to AI invoking, instead it’s just about good manners.

Unlike in Asmimov’s world, we aren’t used to talking to robots as if they are people. The robots also don’t really have any agency in our world — at least not the ones we interact with, yet.

Ignore AI planetary takeovers, my problem is this: If you use Duplex to interact with another human, at least for the foreseeable future, you are an arsehole. Every time I see you I will be thinking about Michael Scott or Alan Partridge.

First, you’d better be pretty damn sure this technology is perfect before you unleash it on somebody else. If your robot phones me up to make an appointment and I have to repeat myself, or it is unclear whether or not you have booked, you are no longer optimising your own time, you are just stealing mine.

Even if things do go smoothly, if I know I’m talking to a robot over the phone — either because I come to recognise the voice, or it introduces itself to me — it is going to be incredible embarrassing for me to have that conversation. I have to choose whether or not to talk to it like a human or like I talk to my own voice assistant.

I can either keep up the facade and waste even more of my own time asking it about the weather and wishing it well, like an idiot, or I can start barking commands down the phone, which is something I am not willing to do if other people, even strangers, are in earshot. I don’t even use the voice assistant on my own phone if there is a single other person in the room because it’s kind of outside the social norm, and yet I have to suffer the same embarrassment to save somebody else’s time? That’s rude.

Google, of course, has done this before. Glass was kind of a cool project, I really liked the idea of maps overlays into your vision and taking photographs without taking out my phone. It would make sense in a society where people had normalised the new interactions it introduced, but in this society it was just too much.

It’s weird that you can’t tell if somebody is taking a photo of you. You also force the issue of using screens in company. Using phones over dinner seems just about socially acceptable now, if you look around any café it’s hard to avoid that conclusion, but there’s still a little bit of social theatre required to pick the right moments. If someone is opening up about something to you, you had better put your phone in your pocket. Having the screen affixed to your face didn’t allow for that subtlety. Just wearing Google Glass was impolite. Maybe in 10 years people will care less.

I just about accept that some people are too busy to make their own hair appointments. If your time is incredibly valuable then maybe that 120 second call is really worth converting to a 15 second instruction to a PA. However, you must accept that by doing so you are sending out a very clear message about how much you value your own time.

Namely, it is worth paying somebody a salary to make a 105 second saving for trivial admin tasks. My time isn’t worth anywhere near that, and neither is yours. I know this because I wrote this post, and you are reading it.

Clearly the cost of a robot assistant is much lower than a human one to the point that maybe, in 10 years, using assistants won’t carry this connotation, but for now it absolutely does. You still think it’s worth making that 105 second saving every 5 weeks, you still think that’s worth the extra time and embarrassment for the recipient of the call. You just are unable or unwilling to pay for it.

I get that we might move fairly quickly to the point where enough of these conditions have changed. Assistants are reliable, we don’t feel stupid asking robots how their day has been, it’s not considered self important for me to outsource my small admin tasks to something else while spending my Sunday watching YouTube clips — but until that happens:

  1. Don’t get your robot to phone me
  2. Maybe just sign up to OpenTable?



Tito Sarrionandia

Head of frontend engineering at Babylon Health. Formerly ThoughtWorks, Made Tech, school teacher.