ALAN TURING AND HIS IMPACT ON SOCIETY
“Sometimes it is the people no one can imagine anything of who do the things no one can imagine.”
THE TURING TEST
The Turing test is an artificial intelligence test invented and described by Alan Turing (Cf. Who was Alan Turing) who’s goal was to evaluate a computer on his ability to have a “normal” conversation with a human, this one not discerning the mechanic nature of it. As you might know, the “Turing Test” was originally called “Imitation Game” and Turing’s main aim was to try to respond to this problematic: “can a computer think?”
The Turing test is very basic. It consists in one person (a tester) having a conversation with two different “people” but the tester doesn’t know whether he will talk with a human or to a computer . To do so, the conversation takes place on a computer and, during a limited time, the tester will get to ask questions. The computer or the other human will get to respond and react to these. The fact a limited time is imposed is because it will prevent the tester of knowing how long it takes for the other to respond, especially with questions related to mathematics.
Here, the tester is “C”, the AI is “A” and the other human is “B”
Then, as you might expect it, the tester has to say if the two conversations he held were with humans or machines. If he says he was talking to two humans, the computer / Artificial Intelligence is approved but if he says one of the conversations was held with an artificial intelligence or computer, the AI is disqualified and seen as non-working for what it was designed for.
Alan Turing made some predictions about his Imitation Game and once hypothesized that after the year 2000, a computer would be able to pass 30 per cent of his “Turing tests”. He also thought that working on how computers could learn by taking in information would be a major case of research, which it is.
Although it is still used today to test different types of AI’s, there are some aspects of the test that can be questioned. The biggest problem is that testers compare the human behavior to the computer’s, therefore implying that people will never have the same reaction if there are testers. The value of that comparison is mostly problematic as it doesn’t measure precisely the degree of “artificial intelligence”. Furthermore, the results of the test are based on the aptitude of an intelligence not to seem that intelligent, as humans are and, the assumption that testers are capable of judging “intelligent” behaviors even if these are not human.
This leads researchers to look for a way that would enable the AI to understand what she is saying which is much more difficult than it seems as “understanding” is quite abstract…