The Turing test is a famous and reasonably well-respected test of a machine’s ability to exhibit intelligent behavior. I’m prompted to write this by news of Cleverbot, an AI web application that learns how to mimic human conversations by communicating with humans. It apparently does well in the Turing test or similar.
I wonder what proportion of humans would pass the Turing test? Judging by my experiences with call centres and support desks, I’m confident many wouldn’t.
Here is a recent example. I have a Facebook account, and I wanted to register a Username to give me a simpler Facebook URL. My attempts to use the on-line forms kept failing, so I emailed Facebook support. Here are some extracts:
Me: “When I tried to set up a username, (in my “Account Settings”), in a red box it told me I needed to verify my account…. I am trying to create the following URL, which doesn’t yet exist; [omitted here] is confirmed to be available as a username…. What am I missing?”
Facebook: “Facebook has several features in place to limit the potential for spam and abuse on the site. One such feature is the “words in the box” (i.e., captchas) that you see when poking, messaging, or sending friend requests. There are two ways to stop seeing these tests. First, you can verify your cell phone, which allows us to verify that a real person is in control of the account…. The second way to stop seeing captchas is to affiliate with a college network.”
Me: “Sorry, but this reply doesn’t appear to have anything to do with my problem. I wasn’t talking about captchas at all…. This is surely nothing whatsoever to do with spam and abuse…. I don’t have a mobile phone (cell phone). What has a mobile phone to do with establishing a username? …. I’m not at school or college. I’m 64 and retired, as seen by my profile.”
Facebook: “Unfortunately, if you cannot verify with a mobile phone or add a college network, you will continue to see captchas and be unable to establish a username / alias.”
First there was the disconnect between my problem, (failure to create a Username), and the response, (spam, abuse, capthas). It isn’t an excuse that these may be technically connected within Facebook. Second, there is the concept that only with a mobile phone can they tell that I am a real person.
Consider the above exchange from a “Turing test” point of view:
I see little or no evidence of a human being at the other end. Facebook support appears to fail the Turing test.
But surely Facebook support could see evidence of a real person at my end? Could a computer have written my emails? Would a mobile phone really be better evidence of a real person than my emails?
I’m sure that in practice there are real people at the other end, following scripts. And the likely reason they can’t accept my emails as evidence is because they have to obey the rules. I have seen people in call centres obeying computers, and know that initiative often has to be limited. But … these exchanges always remind me of the Turing test!
No mobile phone?
Some time ago I got rid of my mobile phone, and replaced my laptop computer with a desktop computer. Now when I walk out of my “office” at home, even into the rest of the house, I am largely out-of-touch. Bliss!
My voicemail on my office-phone simply says “This is Barry Pearson’s answerphone“, with no assertion that I am not around, no promise that I would ring anyone back, and no instructions about what they should do “after the tone”. I may be around. If they don’t know what to do after the tone, it is unlikely that they are worth speaking to. And I may choose not to ring them back. So I avoid patronising and lying to the caller.
(I may get a mobile phone for emergency use. Hardly anyone will know its number).