Turing Test Failure

Turing Test Failure

Yesterday, I had a weird Turing moment.  As background to this post, I should tell you that I was listening to a podcast on robots and artificial intelligence where Sam Harris was chatting with Kate Darling. Had I not been listening to this, I might have forgotten about my Turing moment. Here's what happened. Yesterday, I received a call from a telephone solicitor, trying to sell me some emergency service for my furnace, insinuating as they spoke, but they were some safety issues that needed to be addressed. At first I was confused by what he was saying, despite being perfect English which is probably why I even allowed the conversation to continue. As I spoke to this person, I started wondering whether I was speaking to a chatbot. So I asked the person on the phone, point blank, whether they were human being, or a chat bot.

 

The person responded that yes, they were a human being, but only after a long pause. After another pause, he repeated his pitch regarding the service that my furnace presumably needed. He added that he worked with Enbridge and some other organization. For the record, I have a regular service contract with the company that installed my furnace, so I have no need of such services. Nevertheless, by this point, I was fascinated by the idea that I might actually be talking to a chatbot of some kind, so I asked again whether he was merely reading from a script, and whether it was possible for them to go off-script. There was a long pause again and the person, or chatbot, seemed confused, then repeated his sales pitch regarding my emergency service in a different way.

This went on for a couple more minutes with him trying to sell me and me trying to get him to talk about something else. Eventually, I grew tired of the exercise and said that if they could not go off-script and answer questions that weren't related to furnaces, that there was no point in continuing the conversation. I said, "Good bye," and hung up.

But it did make me think. If I was speaking to a human being, they were actually worse at convincing me of their humanity than some chatbots I've encountered. If they were a chatbot, I suppose it was pretty good in the sense that it tried to convince me that they were a human being. I guess. However, if they were human, then they didn't fully convince me of it and, as such, failed the Turing test. I find that funny, but the idea that the person might be worse in than an artificial intelligence chatbot, gave me pause.

Comments