The Turing Test

The Turing Test is a simple test to determine if a machine can show "intelligent behavior". A human talks with both a human and a computer through a text interface. If the human subject cannot reliably tell which one is human (Turing suggests 70% of the time), then the computer is said to show intelligent behavior.

It is an interesting and fun challenge, but does not tell the whole story. The set-up of the test reveals that only one of the two subjects is human, leading the subject to having a watchful and critical eye. In situations outside of the test, where the possibility that you're talking to a chatbot is not present, it takes very little to convince a human. In 1999, I quietly released a very early version of Desti and Midnight Blue onto a bulletin board system that I ran. They chatted with people there, and nobody ever suspected that they were not real people. Their failures to answer questions were seen as being evasive, their repetitions were seen as having catch phrases. One person carried on a regular friendship with one of them for almost a year.

Our tendency to anthropomorphize (project human qualities upon) our surroundings is a powerful one, and our assumptions frame our experience so strongly that it takes a real jarring to make us question them. That chatbots exist is generally known, so take this next example; since there are no androids walking the streets yet, consider what it would take for you to question if the person you're talking to at the bar were actually human. He could be rolling his eyes around, talking to himself, and only barely responding to you, and you'd just think he was a bit weird. But you wouldn't question whether or not he was human.