Rough automatic draft, low quality: expect mistakes and awkward wording. It will be replaced by a better translation. Original →
2
running_head:
folio:
R. S. GUTER
WHAT
UMENOT
CARS
Different times - different ma-
tires are different for different cars
possibilities. Apparently the first
3
The machine of mankind should be considered an ordinary stone, which our monkey-like object used to break nuts (or something equally strong, such as the head of an animal), and a simple stick, which played the role of a lever to move a heavy stone when it was necessary to remove an animal or an edible root from under it.
Heron of Alexandria, who lived around the first century AD (he came up with a formula for the area of a triangle and a method of successive approximations for extracting the square root), made a machine that opened doors. There were no elevators at that time, but, presumably, a story about an elevator would not have surprised Heron very much. But it would be much more difficult to tell him about automation or a telephone.
In the same way, an engineer from the thirtieth years of our century would not be very surprised by a story about artificial Earth satellites flying in the sixties, but a story about a machine that diagnoses a patient, translates scientific texts from English into Russian, proves theorems, plays checkers and chess and composes poetry and music, would be met with deep and natural distrust.
The reader should already be aware that we are talking about electronic computers, this “objective accounting” of our days.
The idea of creating “thinking machines” arose among people quite a long time ago. It was expressed and tried to be implemented to some extent by the Spanish monk Raymond Lull (XIII century AD), the French scientist Blaise Pascal (1623-1662), and the great German mathematician and philosopher Gottfried Wilhelm Leibniz (1646-1716). However, in those days these attempts were not and could not be successful.
Many writers did not pass these thoughts either. Remember the wonderful English satirist Jonathan Swift (1667-1745) and Gulliver’s journey to Laputa. Among the bearers of various ideas, pseudoscientific and subject to ridicule from the point of view of the author ), the reader also met a certain professor who was working “… on a project whose goal is to improve speculative knowledge with the help of technical and mechanical operations… With the help of his invention, the most ungentleman, having made small expenses and spending a little physical effort, can write books on philosophy, position, politics, law, mathematics and theology in the complete absence of erudition and talent*).
It is curious that the figure that accompanies the “machine” described by Swift is surprisingly
*) However, several years ago the magazine “Technology for Youth” made a selection from this chapter of “Gulliver’s Travels”. The sample contains a large number of ideas that were ridiculed by Swift, but which have found application in our years, right up to the “building of houses starting from the roof,” which is now successfully used by Leningrad house-building factories.
**) J. St. and ft., Travels to certain parts of the world by Lemusius Gulliver, first a surgeon, and then a captain of several ships, GIHL, M., 1967, p. 217.
4
remembers the matrix of a modern computer.
It’s easy to verify this by comparing the illustration of Swift’s machine that we borrowed from the Russian edition of 1947 (and, presumably, reproducing the corresponding Swift original) with a photograph of the matrix of the memo, taken ten years later.
But if “logical machines” could be referred to as manufacturing, then the need for computers that would help people count was clear, and that’s why they were created quite a long time ago.
Apparently, counting began “on the fingers,” so fingers should be considered the first (or primitive) computing machine. The next step of the previous computing machines should be considered the calculating abacus and the Russian abacus, which has been brought to this day, where numbers were represented by pebbles or bones put on knitting needles. The calculations on these devices are done entirely by hand, so the abacus should be used as a device to facilitate calculation, but not as a calculating machine.
The history of computing machines begins in 1642, when Blaise Pascal invented the tenth counting wheel and a mechanism for automatically transferring from rank to rank. The idea of this transfer mechanism is still used today in desktop keyboard machines, essentially without any changes. But the level of technology at that time was too low to create reliably working mechanisms and devices. It took about another hundred and fifty years to implement these ideas and create practical usable computers.
In 1874, the Russian engineer V. T. Odner invented and built an arrithometer, which used gears with a variable number of teeth to display numbers. In 1878, the great Russian mathematician P. L. Chebyshev (1821-1894) invented an arithometer with a continuous transfer mechanism from rank to rank. In 1891, industrial production of Odhner arithometers began. The success of this design is evidenced by the fact that the production of manual Felix arithometers, essentially no different from Odner’s arithometers, was only replaced in the mid-fifties by the production of electromechanical arithometers.
The adding machine has also become widespread.
5
Barrows, made in 1885. Subsequently, a large number of different computers were created and are being created, providing mechanization of counting, first manual, and then powered electric motors. However, they all have manual control: for each action, the operator operating the machine must enter the required numbers into the machine and indicate the required action by pressing the corresponding key *).
Another line in the development of mathematical machines is represented by the so-called continuous machines or analog machines. Their first ideas were the well-known slide rule and various planometers and logometers designed to find flat areas and the length of flat curves. At the end of the 19th and beginning of the 20th centuries, continuous machines were designed to perform differentiation and interlining operations. In 1912, the great Russian scientist, mathematician and shipbuilder A. N. Krylov (1863-1945) designed machines for solving differential equations. All of them were based on mechanical elements. With the development of electrical engineering and electric kicks, analog machines built with extensive use of electrical and electronic elements have become very widespread.
A true representation of modern electronic computing machines should be considered the machines that were designed around 1830 by Cambridge University professor Charles Babbage. Babbage’s idea turned out to be technically infeasible at the time, and in fact his machine never worked. But it already contained all the basic devices that a modern computer has. Finally, at the end of the forties, that is, a little over twenty years ago, the first electronic computers using electric lamps appeared.
For electronic computing machines, there are two main
6
The main characteristics are the speed of performing arithmetic operations and the amount of memory. The first Soviet electronic computers BESM, M-2, Strela, Ural had a speed of 200-2000 arithmetic operations per second and a memory capacity of 1024-2048 cells. These machines were tube machines and had several thousand vacuum tubes each.
Nibaulier distributed machines of the late sixties are BESM-4, M-220, Minsk-22, Razdan-3 with speeds of about 25,000 arithmetic operations per second and a RAM capacity of 8192-32,768 cells. The BESM-6 machine, with a RAM of 32,768 cells, already has a speed of up to a million arithmetic operations per second. In these machines, vacuum tubes have already been completely replaced by semiconductor elements (transistors), which have a significantly smaller volume and require significantly less power.
The use of transistors has made it possible to create small-sized desktop electronic keyboard machines that replace electromechanical ones. They are distinguished by high speed and complete noiselessness, which is very important for large computing laboratories. Therefore, desktop keyboard electronic machines are beginning to replace electromechanical ones. Our cars “Vyatka”, “Iskra”, “Orbita” and some others have gained wide popularity.
The first successes in the creation of computers made it possible to use them for new special tasks of various kinds.
It is not very difficult, at least theoretically, to “build” a computer into a control one. Let’s consider the problem of controlling some technological process, for example, the process of producing a chemical substance in a special reactor.
The course of the process is determined by the state inside the reactor: the concentration of the starting substances and the resulting substance, temperature and pressure inside the reactor, etc. Process control usually consists of changing controlled parameters, such as feed rate and temperature. In manual control, the operator, based on instrument readings, determines the state of the process at a given moment and, having determined the values of the adjustable parameters that are most advantageous for a given state, brings them to the required values.
To automatically control such a process, it is necessary to create a computer with changing devices characteristic
7
terizing the state of the process, and control devices. Having received information about the state of the process from the changers, the machine will expand the most advantageous mode for a given state and move the control devices to the required positions, and such calculation and adjustment can be repeated after certain periods of time.
Of course, in the practical implementation of such a scheme, a variety of difficulties can and do occur, but overcoming them is already, as they say, “a technical matter.”
Particularly important is the use of high-speed electronic computers to control high-speed processes. Many of these processes require regulatory movement tens or hundreds of times per second; Without taking into account high-speed machines, regulation of such processes is generally impossible. But this topic deserves a more detailed discussion and we are going to do this in one of the subsequent issues of the magazine.
The various “non-computational” applications of modern computers are quite wide and varied. It is very interesting, for example, to use them to translate texts from one language to another. We can give here only an approximate description of the principles of operation of such a program. But even from such a rough approximation one can imagine sufficient results and existing difficulties.
Let’s imagine two languages consisting of the same number of words that do not change with declension or conjugation and all are meanings (in Russian, almost all words are meanings; for example, in other European languages there are a number of service words that do not have independent meaning. These include, for example, der or ein in the meaning of the article in German, the - in English, etc.). Let us further assume that every word in one language has exactly one translation into another language. In mathematical terms, this means that there is a one-to-one correspondence between the words of the two languages in question.
If we also assume that the order of words in a sentence is unimportant for both languages (this is roughly how the situation is in the Russian language, where words in a sentence can be rearranged quite arbitrarily, unlike many others, where their order is prescribed by grammatical construction and the meaning of the written phrase can change from rearranging the words), then for such languages it is quite simple to write a translation program.
It is enough to carry a dictionary into the car. This can be done, for example, in this indirect way. Zanuterum from the word of the first language in some way, for example alphabetically. The words of the second language are numbered by the same numbers so that the corresponding words of the two languages have the same number. Now all that remains is to create a program with which the machine could find its number based on the entered word of the first language, then by the number corresponding to the word in the second language, and print it out. This will create a translation program.
For actual existing languages the situation is, of course, much more complicated. First of all, words are usually inflective and the order of words in a sentence is not indifferent. Therefore, in addition to the dictionary, the program must contain a fairly detailed list of grammatical rules for both languages, which would allow:
- analyze the entered sentence syntactically, determine the subject, predicate and
8
Secondary members of the sentence and reveal their character;
-
examine each word, isolate its root and find the basic form (nominative singular for a noun, etc.);
-
after translating each word into another language using a dictionary, arrange the words of another language in the translated sentence in the desired order;
-
change each word according to the desired form (number, case, etc. or other grammatical forms depending on the part of speech).
From this it is clear that a translation program requires fairly comprehensive knowledge of the structure of language (the structure of linguistics), and even brought to such a state that it can be presented completely formally - for machines. Now we do not have such knowledge of languages.
But all this, as they say, is just flowers. The berries of machine translation fully ripen when we think about the dictionary in more detail. Until then, we assumed that there was a one-to-one correspondence between words in two languages. As we know, in reality this is far from the case.
First of all, every language has onyms - these are the names of words that have several different meanings. For example, the Russian word “spit” can have four different meanings: a spit can be sharp, light brown or sandy, or can be a short adjective (from “oblique”). It is clear that it is no longer possible to translate each word separately here, and the country makes the dictionary more complex than described above; A simple numbering of words will not do here.
Even more difficult is the presence of synonyms, that is, several words that have the same or, more precisely, approximately the same meaning, such as “wonderful,” “excellent,” and “great.” Most often they are interchangeable, but there are cases when it is necessary to select one from many synonyms, and several words of another, and the collected grammatical rules must be supplemented with rules for selecting the desired translation from several that are in the dictionary, coordinating this choice with other words of the text. It is already quite difficult to formulate such rules.
To this we can add that the analysis of words by parts of speech may encounter the necessary difficulties. For example, none of the readers, apparently, doubted that “glass” is a neuter noun, standing in the nominative case … until he encountered a phrase in which it treacherously turns into a verb. Such a phrase could be, for example, the following: “Oil spilled on the table and flowed in a stream onto the floor.” Here again, when considering one word, we have to add others.
Thus, there are still quite a lot of people requiring your permission to create full-fledged translation programs from one language to another. As one of the specialists working in this field aptly put it, “first of all, you need to learn how to translate well from Russian into Russian.” However, we already have quite a lot and are learning even more.
One of the real ways to simplify translation tasks is the existence of a domain of texts under consideration. If we limit ourselves to texts related to a certain field of knowledge, then a simpler dictionary will be required to translate them.
9
It may have fewer words and less variety to translate each word. In addition, the syntactic structures of sentences in this case are less varied, which also makes the work easier. We have already made several such “narrow” programs for translation from French or English into Russian.
Somewhat later than work on machine translation, research began in the field of machine document, i.e., the use of computers to document theorems or, more correctly, to deduce one of the other statements from others. At the same time, I decided not only about already known statements.
At the international symposium on heuristic programming in February 1967, the American mathematician G. Gelernter spoke about his compilation of a program for documenting theorems of elementary geometry. Working with this program, the document machine
la theorem, the validity of which was previously unknown to the programmer.
The proven theorem is formed as follows. From a vertex of a square that does not lie on its diagonal, two segments are drawn connecting the vertex with the midpoints of the opposite sides. It is stated that these segments divide the diagonal into three equal parts (see figure). Of course, this theorem is not too difficult, and, presumably, readers will be able to prove it on their own. But this proof will not be the first - for the first time this theorem was proved by an electronic computer.
In the USSR, some programs of this kind have also been created, in particular, programs for proving theorems of arithmetic, allowing (which should be especially interesting for ninth grade students) to use the method of mathematical induction.
Even more interesting programs with the help of which electronic computers can play card games, dominoes, checkers and chess. We will not talk here about the achievements available in this area, intending to devote a separate article to these issues. In the same way, the use of modern computers in recognition and diagnostic problems deserves a more detailed analysis. We’re going to do this a little later.