“We can only see a short distance ahead, but we can see plenty there that needs to be done.”
By Alan Turing.
The Turing Test is born as a method to determine if a machine can think. Its development is based on the game of imitation.
The original idea is to have three people, an interrogator, a man and a woman. The interrogator is separated from the other two, and can only communicate with them by writing in a language that everyone understands. The objective of the interrogator is to discover who is the woman and who is the man, while the other two is to convince the interrogator that they are the woman.
The variant introduced by Turing consists of replacing one of those interrogated by a computer. Two cases can be given, that man be substituted, with which only the computer would have to appear to be a woman, or that the woman be substituted, with which both the man and the computer would be imitating. Although this last option could be an interesting experiment, there is no attempt to check the ability to imitate a woman, so Turing changes the goal of knowing sex by recognizing the machine. The purpose of these changes is to make the game as fair as possible. The first thing is that it does not have to consist of a cheating contest, so one of those involved would not have to pretend to be something else. Another detail is that Turing cares little if the computer uses pre-established tricks to evade or manipulate the answers (for example, making mistakes in arithmetic questions or taking longer than necessary to answer them). It assumes that the interrogator will also use them to recognize him, so the important thing is what results from the game, not the methods used to play or the internal mechanisms of reasoning, which, among other things, are also unknown in the human being.
A machine could pass the Turing test when the interrogator failed to recognize it on a significant number of occasions.
As soon as the Turing Test appears, the first criticisms also come to light. Most of them were based on ethical and religious issues, and many of the most critical positions came from people who considered that the human being was very special and that no machine could even approach the capabilities of this.
One of the first objections is mathematical. Gödel’s theorem states that in a logical system with sufficient power, sentences can be created that can neither be proven nor refuted within it. Nevertheless, Turing affirms that of the errors or confusions the human mind is not free either, and this diminishes the intellectual capacity.
Another difficulty is the lack of conscience. It was stated that in order for a machine to be mentally active, it must be aware of both itself and others, and generate positive or negative feelings about the information it receives or the actions it performs. Solipsism is a radicalization of this idea, which holds that the only way to know if a machine thinks is to be that machine. The problem is that, following this idea, the only way to know if another human being thinks is to be that human being, what is known as the problem of other minds. Turing affirms that, if among human beings it is considered politically correct to obviate solipsism, it should also be done with machines. And how the only way to solve the problem of lack of consciousness is solipsism, the most appropriate thing is that it is not considered either.
With the objection of Lady Lovelace, the idea is that the machines could never generate anything new, surprising or different. As Turing says (and as anyone who has used, for example, a program of structural calculation or simply known operating systems of windows, could ratify), the computer, being a machine, can surprise continuously. Although this can not be considered as a creative mental process, creativity may be realized in the mind of the observer, and not in the generator. For example, both a book and a person or a car can surprise you.
To the problem that the machine is a discrete system while the human mind a continuous system (problem of the continuity of the nervous system), Turing responds that any continuous system can be discretized with sufficient resources so that the difference between one is not noticed and other.
Finally, we can talk about the problem of informality of personality. Human behavior can not be described with a set of useful rules in any situation. Turing’s answer is that there are differences between rules of conduct (for example, with the red light, stop) and rules of action. The rules of conduct can be listed, but not the rules of action, because, among other things, many are unknown. But Turing also says that even with a few rules of action in a discrete system the answers can be totally unexpected and different, so that, as in a human being, can not be foreseen.
Turing’s article gathers many audacious comments about the possibilities of the intelligence of machines, which at that time many seemed to be science fiction. Turing believed that computers capable of human tasks and in a human way, that the difficulties of designing thinking machines were mainly programming and that the “feats” he expected from the machines would be realizable in the foreseeable future (such as adjusting their own program or predict the effect of alterations in its own structure).
What in 1950, in terms of speed and capacity in computers was unimaginable, is now reality. However, Turing’s predictions about machines and the Imitation Game are still a challenge (Turing thought that in about 50 years there would be machines that would “play” the Imitation Game so well that an interrogator would not have a probability greater than 70% to make the proper identification after five minutes of questions).