"Representative" "Representative" "A f**king representative!"

By far the most exciting announcement to come out of I/O this year was the announcement of automated calling from Google Assistant. For those that did not watch the keynote (or read my blog post), Google is experimenting with having the Assistant be able to make calls for you to do things such as make reservations or book appointments. In an expertly coordinated demonstration, the Assistant was able to book an appointment without significant issue while sounding extraordinarily human; the human on the other end of the live call seemed to have no knowledge of talking to a robot.

Now, this type of technology, which Google calls Duplex, offers tremendous potential for the future of artificial intelligence as well human/technology interaction; you can't tell me you have never wished to have J.A.R.V.I.S. Google's goal is to give time back to the people, and let machines do what they are more than capable of doing. While I completely love the potential of this venture, there are clearly some problems that could arise.

I will not even bother going in to the ethical questions that are quickly popping up about this technology such as should humans we notified about who they are speaking with. If that is what you are concerned about or interested in, I highly recommend reading this excellent article published by The Verge reporter James Vincent.

No, my concerns lie more with the technological restrictions because my moral compass must point in a different direction. As the demonstration showed, Duplex has the potential to be an incredible breakthrough in terms of formulaic tasks. For instance, it will know my calendar, commute time, and price range due to the data I give to it so it can easily navigate a conversation for scheduling most events and making reservations. However, it will not be to make judgement calls on the fly without information from me.

For instance, I golf every Wednesday and sometimes I will make my weekend tee time while I am there. However, in those very common times when I forget, I always call to make my tee time the next day for the weekend. On a normal time, it would be easy and convenient for the Assistant to call and confirm a time for me as well as put it on the calendar. However, what if they ask to group us with another group? My decision would depend on the people in that group so I could not make it a simple Boolean decision. In addition, what if I call for a specific time and they ask to schedule me a close time that is not full?

I also think about the frustration that could arise on the recipient end of the call as it is not always a joy to speak with robots. Yes, the naturalistic ability of the Assistant during the demo call was amazing, and the ability of artificial intelligence to calculate reasonable responses these days is incredible. But, sometimes even the most simple phone calls can turn in to lost nightmares. For instance, maybe I just want to call and schedule my car to have an oil change. Simple right? I worked for Firestone, and trust me, that call is far from simple. If someone called me and asked about that, I am asking 300 other questions to see if there is anything else I can do for them.

As an employee, would I get pissed if I was asking questions only to discover that I was speaking a robot that probably doesn't know? You're damn right. In these scenarios, the Assistant would not have enough information input from me to know how to adequately address the questions, and then you may have a pissed off employee screaming "REPRESENTATIVE!" at the phone just to find themselves making a circle on an automated loop.

Is this technology far more advanced than those automated machines? Absolutely; it is significantly better. And, these issues could be easily solved I am sure. I could give the Assistant a range of tee times that I would take, and as a fallback for anything, I could have the Assistant put me on the call if it hits a roadblock in answering questions. The technology is still new and very much experimental, and I am sure that it will only get better as testing moves forward. However, I think it is important to look at the potential pitfalls early to make sure they can be addressed so that Duplex can be a benefit to us all.