The Turing Test Meets Real Life

One of my pet peeves when I was writing my Reflections Beyond Technology blog was the proliferation of people who should know better asserting that the Turing Test was a definition of intelligence. This came to a head when someone named Jeff Hawkins (who hopefully has been forgotten after having his fifteen minutes of fame) wrote a book entitled On Intelligence. Hawkins was indicative of those who cited Turing without ever taking the trouble to read him, thus getting things wrong from the very beginning of his argument.

What I liked most about Turing was that he was honest enough to recognize that any useful definition of intelligence was probably beyond our current grasp and could well remain so for some time. Like any good scientist, he believed that we should focus on questions that we had a fighting chance of answering. In the interest of eating that particular form of dog food, he introduced a concept he called The Imitation Game. This was a two-player game where all "moves" involved exchanges of typed sentences, meaning that each play could be kept from seeing and hearing the other. The goal of the game was for one person to guess the sex of the other strictly on the basis of these exchanges of printed text. He then asked whether it would be possible for a computer to play the game of the person whose sex had to be determined without the other player realizing that (s)he was exchanging sentences with a machine.

I thought about this "real" Turing test the other day when I was reading "The Programmed Prospect Before Us" by Robert Skidelsky in the April 3 edition of The New York Review of Books. (Yes, I know how I get behind on my reading.) Near the end of the article, Skidelsky offering the following anecdote:

Recently, Michael Scherer, a Time magazine bureau chief, received a phone call from a young lady, Samantha West, asking him if he wanted a deal on health insurance. After she responded to a number of his queries in what sounded like prerecorded fashion, he asked her point-blank whether she was a robot, to which he got the reply "I am human." When he repeated the question, the connection was cut off. Samantha West turned out to be a system of recorded messages that were part of a computer program created by the brokers for health insurance.

This may be the best real-world account I have seen of an instance of The Imitation Game in which, for all intents and purposes, the computer "won." I should also point out that this exercise did not take place at an artificial intelligence laboratory or even one for experimental psychology. It was a piece of software deployed by a health insurance company, presumably as a result of an effort to improve "efficiency" by following those principles outlined by Frederick Taylor in his study and promotion of practices of "scientific management." Given that I have already written about how these principles have been applied to heart surgery, it seems ironically inevitable that they are now also being applied to health insurance.