Turing's Testes (Essay #6: Artificial Intelligence and Robotics)

The Turing test, created by the world's first computer scientist (before computers even existed!), is a test designed to gauge when true artificial intelligence has been manufactured. It is basically a test to determine when an artificial intelligence program can imitate a human being well enough to fool a majority of people it encounters into thinking that it is a human.

The test is designed to gauge how well a computer can fool a human into thinking it is human. This is done in a number of ways; the foremost among them are by attempting to engage in human behavior. This includes irrationality, typographical errors, “lame” or “corny” jokes, incorrect math, creation of a persona, and can even include attempts at flirting. When a conversation persists for long enough, a person will eventually be able to tell that something is off, and that they may not be talking with an actual human being. If you've ever tried to talk to CleverBot, you know that the robot often types in circles and uses quotes or reasoning that a human wouldn't. An example of a conversation with CleverBot illustrates this idea.

Me: What's your name?

CleverBot: Pepus.

Me: Pepus?

CleverBot: Yes Pepe.

Me: Why did your parents name you that, Pepe?

CleverBot: That is not my name!

Me: Then what is your name?

CleverBot: Then my name is Weronika.

Me: Are you a female?

CleverBot: Good question. Am I female?

Although the CleverBot did use the word “my,” indicating that it has some degree of awareness of itself, this conversation illustrates how inconsistent the CleverBot was in the conversation. It contradicted itself in two different posts about its name; the first, when it changed its name from Pepus to Pepe, and the second when it changed its name from Pepe to Weronika (and no, I did not make that up). CleverBot also exhibited gender identity confusion, which is a problem experienced by many humans, but humans don't go around asking others if they are male or female. Obviously, CleverBot has failed the Turing test.

Turing said that if an artificial intelligence program could win the so-called imitation game, it could be counted as true artificial intelligence because it would have exhibited enough human personality characteristics to fool actual humans. The test allows for the threshold of what is considered artificial intelligence to be low enough that a sizable percentage (30%) of the population would be able to tell or have some idea that they were talking with a machine. Why not go for trying to convince 100% of the panelists that the artificial intelligence program is a human being? Well, when you know you're supposed to be judging the humanity of a presence that you cannot see or feel, you are, naturally, going to have a nagging doubt in your mind about whether the person is truly a human, while at the same time having a doubt about whether the presence is a robot. The test allows for that margin of uncertainty that accompanies any human judgment.