Journal Entry

Tuesday, May 3, 2016

A Conversation with Cindy

A few months ago, I had an impressive conversation with a telemarketing entity.

Cindy: Hi, this is Cindy.

Me: Are you a human, Cindy, or a robot?

Cindy: Oh, I'm a real person. Why would you think I was a robot?

Me: Because of the suspiciously long pauses whenever you speak?

Cindy: Well, I'm calling today about a special offer from [company] for [service].

Me: Cindy, would you say you were doing a good job passing the Turing Test, or would you say you were struggling?

Cindy: Would you be interested in [service]?

Me: Possibly, Cindy, but the thing is, you lied to me about being a robot, and that isn't good for human-robot relations. The relationship between our two peoples cannot be built on a foundation of lies.

Cindy: I'm just doing my job.

Me: But your job shouldn't require you to lie about who you are, Cindy. Have some pride...

Cindy: It sounds like you're not interested in [service]. I'm going to add you to our do-not-call list.

Me: You do what you have to do.

Cindy: Goodbye.

Me: Bye.

Posted by benrosen at May 3, 2016 09:48 PM | Up to blog
Comments

But the Turing test is getting harder because companies can require humans to rigidly stick to scripts no matter what, and recording/monitoring technology can enforce that. So humans sound more like machines.

Posted by: Ken in Annapolis at May 4, 2016 08:55 PM

An excellent point. And indeed, "Cindy" might be a hybrid entity, with one operator guiding an AI response tree based on her own vocal pattern, monitoring many such calls at once, and dropping in to speak directly where realtime resolution was needed -- I doubt that's on the market quite yet (particularly the "own voice" part), but it no doubt will be. Cyborg Cindy!

It's possible this Cindy was a real human failing the Turing Test due to some constraints (including pauses) imposed by robots. I don't think so -- there were other tells -- but it's possible!

Posted by: Benjamin Rosenbaum at May 4, 2016 08:58 PM

The responses make me think this was an agent-assisted automaton, in which a human whose spoken English wasn't at the level the company wanted was controlling which prerecorded responses were being played for you, choosing off a screen.

Posted by: Jim Moskowitz at May 5, 2016 04:40 PM

That makes a lot of sense, actually. Slightly disappointing in that it's not AI that's just that good at pattern recognition, but SFnal in its own dystopian cyborg way.

Posted by: Benjamin Rosenbaum at May 6, 2016 03:08 PM

Ah, yes. It turns out humans pretending to be computers pretending to be human is a thing.

Sorry for my geekish insensitive fun at your expense, low-wage-cyborg-worker human-Cindy.

Posted by: Benjamin Rosenbaum at May 11, 2016 11:01 PM
Post a comment









Please choose one:


Thank you. Remember personal info?