Last week, google showed off a really cool demo phone call. Google’s AI called a hair salon to setup an appointment. It went something like this:

AI: Hello, I'd like to make a hair appointment for my client for, umm, the Wednesday the 7th for 12pm.

Salon (Actual person): Ok. Give me oneee second.

AI: mmhhmm

Salon: We don't have anything for 12, we do have a 1:15.

AI: Umm, I'm looking for something between 10-12 actually.

Salon: Depends what she is looking for. What service does she want?

AI: Just a women's haircut for now

Salon: Ok, I think I can get her in at 10am. Thanks.

AI: Ok, thanks! bye.

The transcript of the conversation doesn’t really stand out. It’s not too far off from the Google Assistant.

The biggest surprise was how natural the AI sounded.

Ordinarily, it’s pretty easy to tell when you’re speaking to a robot versus a human. But, this robot was convincing. Both for the Salon and me as an outside listener.

Having listened a few more times, I think there’s two things that set it apart from previous iterations.

First, the AI speaks with imperfection and variation. It doesn’t pronounce words identically each time.

Second, it adds filler words, like umm and ahh. It even uses mmhhmm instead of a more defined affirmative response like yes, yep, or sure.

These may seem like trivial things to us. But it seems to make a big difference in our interpretation of what is natural speech and what is not.

Regarding the intelligence of the robot, I don’t think there’s a great leap here. Making an appointment is a very well-defined task. There’s only so many possibilities for where the conversation can go.

This takes me to the bigger picture. Clearly, robots will start sounding more and more human-like. In 5 years, most of our appointments may be made by robots. Heck, x.ai has already been setting up meetings through email for a few years.

But, I still don’t see anything here that signals progress towards general intelligence. Like ours. And I think that’s a good thing. I’m with Elon Musk, Ray Kurzwell and the other doomsdayers here. Once a machine can reach our level of intelligence, there’s nothing stopping it from going beyond it. It will be the end for us.

(If you’re interested in going deeper down the AI doomsday rabit-hole, Wait But Why has a great post about it.)

So, it seems that all we have here is a robot that can handle calls that were pretty robotic in the first place. I mean think about it. If the hair salon was online, I’d just click a few buttons to set up an appointment. Or if they had an app, I could probably ask google assistant to do it for me already.

If this means robots can handle more of these robotic tasks with less work for me, I’m all for it.

Back to posts