Can cars think?

Can a car think? "- perhaps the most famous article by A. Turing . Even now, almost 50 years after writing it, it, which in its time caused a great number of serious research and pseudoscientific speculations, has not lost its meaning The article is written with humor and irony ("there are emoticons between the lines, according to E. Hodges, Turing's biographer), but behind the humorous tone of the presentation are some of the most original and profound ideas expressed in the past century.

The “imitation game” described in this article was called the “Turing test” (which became the standard theoretical test for “machine intelligence”), which, in addition to cybernetics experts, was interested in some psychiatrists who perceived a deep psychoanalytic meaning in the game goal (" guessing the floor ").

Imitation game

I'm going to consider the question: can cars think. But for this you first need to determine the meaning of the terms "machine" and "think". It would be possible to construct these definitions so that they reflect as best as possible the usual use of these words, but this approach carries some danger. The fact is that if we figure out the meanings of the words "machine" and "think", exploring. As these words are usually defined, it will be difficult for us to avoid the conclusion that the meaning of these words and the answer to the question, "Can machines think?" should be sought by statistical surveys like a questionnaire survey conducted by the Gallup Institute1. However, this is ridiculous. Instead of trying to give such a definition, I will replace Our question with another, which is closely connected with it and expressed in words with a relatively clear meaning.

This new form can be described with the help of games, which we call the "imitation game." In this game, there are three people involved: a man (A), a woman (B) and someone who asks questions (C), who can be a person of either sex. The questioner is separated from the other two participants in the game by the walls of the room in which he is located. The goal of the game for the questioner is to determine which of the two other participants in the game is a man (A) and who is a woman (B). He knows them under the designations X and Y and at the end of the game either says: "X is A and Y is B", or: "X is B and Y is A". He is allowed to ask questions of the following kind, for example:

S: "I will ask X to tell me the length of his (or her) hair."

Suppose now that in reality X is A. In this case, A should give an answer. For A, the goal of the game is to induce C to come to the wrong conclusion. Therefore, his answer may be, for example, as follows:

"My hair is cut short and the longest strands are about nine inches long."

So that the questioner could not tell by voice which of the two other participants in the game is a man and who is a woman, the answers to the questions should be given in writing, and even better - on a typewriter. The ideal case would be a telegraph between the two rooms where the participants are located. If this cannot be done, then some mediator should transmit the answers and questions. The goal of the game for the third player - the woman (B) - is to help the questioner. For her, probably the best strategy is to give truthful answers. She can also make such remarks as “A woman is me, do not listen to him,” but by this she will achieve nothing, since a man can also make such remarks.

We now ask the question: "What will happen if a car takes part in this game instead of A?" In this case, will the questioner err just as often as in the game, where the participants are only humans? These questions will replace our initial question "Can machines think?".

Criticism of the new formulation of the problem

Just as we ask the question: "What is the answer to the problem in its new form?", You can ask:

"Does the problem in its new formulation deserve consideration?" We will consider this last question, not postponing the case, so that later we will not return to it.

The new formulation of our problem has the advantage of allowing a clear distinction between the physical and mental capabilities of a person. No engineer or chemist claims to create a material that could not be distinguished from human skin. Such an invention, perhaps, will ever be made. But even admitting the possibility of creating material that is indistinguishable from human skin, we still feel that it hardly makes sense to try to give the “thinking machine” a greater resemblance to a person, putting it on such artificial flesh. The form that we have given to the problem reflects this circumstance in the condition that prevents the questioner from contacting other participants in the game, seeing them or hearing their voices. Some other advantages of the entered criterion can be shown if we give examples of possible questions and answers. For example:

S: Please write a sonnet on the topic of a bridge over the Fort2.

A: Dismiss me from this. I never had to write poetry.

S: Add 34,957 to 70,764.

A (silent about 30 seconds, then gives the answer): 105 621.

S: Do you play chess?

A: Yes.

S: I only have a king on e8 and no other figures. You only have a king on eb and a rook on h1. How will you play?

A (after 15 seconds of silence): Hh8. Mat.

It seems to us that the method of questions and answers is suitable to cover almost any area of ​​human activity that we want to introduce into consideration. We do not wish to blame the car for its inability to shine at beauty contests, nor to blame the person for being defeated in a contest with an airplane. The conditions of the game make these drawbacks irrelevant. Those responsible, if they find it appropriate, can brag about their charm, strength or courage, as much as they like, and asking questions cannot demand practical proof of that.

Probably, our game can be criticized on the grounds that in it the advantages are largely on the side of the machine. If a person tried to pretend to be a car, then, obviously, he would look very pitiful. He would immediately impersonate sluggishness and inaccuracy in the calculations. In addition, can not the machine do something that should be described as thinking, but what would be very far from what a person does? This objection is very compelling. But in response to it, we, in any case, can say that if you can still implement such a machine that will play imitation satisfactorily, then you should not worry too much about this objection.

It might be noted that with the “game of imitation” it is possible that simple imitation of human behavior would not be the best strategy for the machine. Such a case is possible, but I do not think that it will lead us to something essentially new. In any case, no one tried to explore the theory of our game in this direction, and we will assume that the best strategy for the machine is to give answers that people would give in the right situation.

Machines involved in the game

The question posed in Section I will not be completely accurate until we indicate what exactly is meant by the word "machine." Of course, we would like the game to use any kind of engineering technology. We also tend to admit the possibility that an engineer or a group of engineers can build a machine that will work, but they cannot give a satisfactory description of the work, since the method they used was mostly experimental [trial and error]. Finally, we would like to exclude people born in the usual way from the category of cars. It is difficult to construct a definition so that it satisfies these three conditions. It is possible, for example, to require that all designers of a machine be of the same sex, in reality, however, this is not enough, since it seems that one can grow a complete individual from a single cell taken (for example) from human skin. To do this would be a feat of biological technology that deserves the highest praise, but we are not inclined to regard this case as “building a thinking machine.”

This leads us to the idea of ​​abandoning the requirement that any type of technology should be allowed in the game. We are even more inclined to this thought due to the fact that our interest in "thinking machines" arose thanks to a special kind of machine, usually called "electronic computer" or "digital computer". Therefore, we allow only digital computers to participate in our game.

At first glance, this restriction seems very strong. I will try to show that in reality this is not the case. To do this, I will have to give a brief overview of the nature and properties of these computers. You can also “say that the identification of computers with digital computers - as well as our criterion of“ thinking ”- should be considered completely unsatisfactory if (contrary to my conviction) it seems that digital computers are not able to play imitation well.

A whole range of computers is already in operation, and the question naturally arises: "Why would we, instead of doubting the correctness of our reasoning, not set up an experiment? It would not be difficult to satisfy the conditions, we could use many different people, and the statistics obtained would show how often the questioners managed to come to the right conclusion. "

In short, this question can be answered like this: we are not interested in whether all digital computers will play well in imitation, and not whether those computers that we currently have will play well in this game; the question is whether imaginary computers exist that can play well. But this is only a short answer. Below we will look at this question in a slightly different light.

Digital computers

What we have in mind when talking about digital computers can be explained as follows. It is assumed that these machines can perform any operation that a human calculator could perform. We believe that the calculator adheres to certain rules once and for all and has no right to deviate from them in anything. We can also assume that these rules are collected in a book that is replaced by another when the calculator starts a new job. The human calculator also has unlimited. the stock of paper on which he performs the calculations. In addition, he can perform the operations of addition and multiplication using an adding machine - this is not essential.

If the explanation given above is taken for definition, then there is a threat that our reasoning will be moving in a vicious circle. To avoid this danger, we will list the means by which the desired effect is achieved. We can assume that a digital computer consists of three parts:

1. storage device
2. actuator,
3. control device.

A storage device is a warehouse of information. It corresponds to the paper that a human calculator has, regardless of whether the paper is the one on which the calculations are made or the one on which the rule book is printed. Since the calculator person does some calculations in his mind, a part of the memory of the machine will correspond to the memory of the calculator.

The actuator is a part of the machine that performs various individual operations that make up the calculation. The nature of these operations varies from car to car. You can usually do very cumbersome operations, for example: "multiply 3 540 675 445 by 7 076 345 687" - however, on some machines you can perform only very simple operations, like these: "write 0".

We have already mentioned that the "rule book" available to the calculator is replaced in the machine by some part of the storage device, which in this case is called the "command table". The duty of the controlling device is to ensure that these commands are executed correctly and in the correct order. The control device is designed so that it happens without fail.

Information stored in the storage device is divided into small parts, which are distributed in memory cells. For example, for some machines, such a cell may consist of ten decimal digits. Those cells in which various information is stored are assigned numbers in a certain order. A typical command might read:. "Add the number stored in cell 6809 to the number stored in cell 4302, and place the result in the cell where the last of the numbers was kept."

Needless to say, if all this is expressed in Russian [English], the machine will not execute such a command. It would be more convenient to encode this command in the form of, for example, the number b 809 430 217. Here, 17 indicates which of the various operations that can be performed using this machine should be done with the numbers stored in the indicated cells. In this case, the operation described above is meant, i.e. operation "number ... add to the number ...". It should be noted that the team itself takes 10 digits and, thus, fills a single memory cell, which is very convenient. Usually the control device selects the necessary commands in the order in which they are arranged, but sometimes such commands can also occur:

"Now execute the command stored in cell 5606 and continue from there" or:

"If cell 4505 contains 0, execute the command contained in cell 6707, otherwise continue in order."

The commands of these latter types are very important, as they allow repeating again and again a certain sequence of operations until a certain condition is fulfilled, and to repeat this sequence of operations it is not necessary to resort to new commands. The machine simply performs the same commands over and over again. We use the analogy of everyday life. Suppose mom wants Tommy to go to the shoemaker every morning on her way to school to see if her shoes are ready. She can ask him again and again every morning. But she can also hang out once and for all in the hallway a note that Tommy will see when he goes to school, and which will remind him that he should go for shoes. When Tommy brings shoes from a shoemaker, mom should break the note.

The reader should consider it firmly established that digital computers can be built on the basis of the principles we have described above, and that they are really being built by adhering to these principles. It should be clear to him that digital computers can in fact quite accurately imitate the actions of a human calculator.

Of course, the rule book described by us, which the calculator uses, is merely a convenient fiction. In fact, real calculators remember what they should do. If we want to build a machine that imitates the actions of a human calculator when performing some complicated operation, then we should ask the latter how he performs this operation and present the answer in the form of a table of commands.

Creating a table of commands is usually called "programming." "To program the execution of machine operation A" means to enter into the machine a suitable table of commands, following which the machine can perform operation A.

An interesting kind of digital computers are "digital computers with a random element." Such machines have commands that contain the throwing of a die or some equivalent electronic process. One such command may be, for example, the following: "throw a dice and put a number when throwing it into a cell 1000". Sometimes they say that such machines have the freedom of choice (although personally I would not use such an expression). It is usually impossible to establish the presence of a “random element” in a car by observing its action, since if you make, for example, the choice of a command dependent on the sequence of digits in the decimal decomposition of the number p, the result will be quite similar.

All digital computers that exist in reality have only finite memory. However, it is theoretically easy to imagine a car with unlimited memory. Of course, at any given time, it is possible to use only the final part of the storage device. In the same way, a memory device that can be physically implemented always has finite dimensions, but we can think of it in such a way that, as the need arises, more and more new parts are attached to it.Such computers are of particular theoretical interest, and henceforth we will call them machines with infinite memory capacity.

Сама идея цифровой вычислительной машины отнюдь не является новой. Чарлз Бэббидж3, занимавший с 1828-го по 1839 г. Люкасовскую кафедру по математике в Кембридже4, разработал проект вычислительного устройства, названного им "Аналитической машиной"; создание ее, однако, так и не удалось завершить. Хотя у Бэббиджа были все основные идеи, существенные для создания такого механизма, его машина не имела перспектив. Скорость вычислений, которую позволила бы достичь машина Бэббиджа, оказалась бы, разумеется, выше скорости, достигаемой человеком, однако она была бы почти в 100 раз меньше, чем у той вычислительной машины, которая в настоящее время работает в Манчестере5 и которая является одной из самых медленных современных машин. Запоминающее устройство в машине Бэббиджа было задумано как чисто механическое, с использованием карт и зубчатых колес.

The fact that the Babbage Analytical Machine was conceived as a purely mechanical apparatus helps us to get rid of one prejudice. The fact that modern digital machines are electrical devices and that the human nervous system in a sense can be identified with an electrical device is often attached importance. But, since the Babbage machine was not an electrical apparatus and since, in a certain sense, all digital computers are equivalent, it becomes clear that the use of electricity in this case cannot have a theoretical value. Naturally, where fast signal transmission is required, electricity usually appears, so it is not surprising that we meet it in both these cases. For the nervous system, chemical phenomena play at least as important a role as electrical.In some computers, the storage device is mainly acoustic. Hence it is clear that the similarity between the nervous system and digital computers, which consists in the fact that in both cases electricity is used, reduces only to a very superficial analogy. If we really want to discover deep connections, we should rather look for similarities in mathematical models of the functioning of the nervous system and digital computers.we should rather look for similarities in mathematical models of the functioning of the nervous system and digital computers.we should rather look for similarities in mathematical models of the functioning of the nervous system and digital computers.

The versatility of digital computers

The digital computers discussed in the previous section can be classified as “machines with discrete states”. This is the name of the machine whose work is made up of abrupt shifts of their state that occur successively one after another. The states in question are quite different from each other, so you can neglect the opportunity to take by mistake one of them for the other. Strictly speaking, these machines do not exist. In reality, every movement is continuous. However, there are many types of machines that are conveniently considered as machines with discrete states.