Exit the "Chinese Room" or whether the machine can think


I have long been interested in the problem of consciousness and its connection with the brain. One day, I came across a very interesting thought experiment called the “Chinese Room”. I don’t remember in which publication I first read about him, but he was very interested in me, so I began to study the question. Only after reading the original article [1] did I understand that in most sources the ideas of the author are presented simply, distorted and not fully understood. The experiment with the Chinese room poses questions and tries to give answers. The questions are very interesting, and the answers, in my opinion, are unsatisfactory. Therefore, I thought for a long time and decided to formulate myself a solution to this problem, which will be my article.

The essence

In 1980, the philosopher John Searle first presented his thought experiment in the article “The Minds, Brains, and Programs” of The Behavioral and Brain Sciences. The experiment has caused a lot of controversy among people involved in the problems of the philosophy of consciousness and artificial intelligence. The stated purpose of the article was to prove two statements: 1) The intentionality of man (and animals) is the result of the causal properties of the brain; and 2) Computer execution of a program cannot be a sufficient condition for the emergence of intentionality.

Three decades before Searle’s article was published, Alan Turing formulated his well-known criterion for a reasonable machine, also known as the Turing Test: a computer capable of supporting dialogue (textual) and, therefore, for a person, should be considered reasonable. By the end of the 70s, computers and AI programs had reached the level where some of their developers began to declare that their machines were capable of understanding English. Searle’s article, containing the experiment described below, was largely a reaction to such statements and the general opinion that a computer is in principle capable of understanding human speech.

What was the experiment? Searle proposed to present the room in which the person is located. Cards with Chinese hieroglyphs can be drawn inside the room. The person inside the room does not speak or read Chinese at all, so the meaning of the hieroglyphs is not clear to him. At the same time, a person is given detailed instructions on how to operate with incoming cards and which cards to answer. Thus, a person outside the room can ask various questions using cards and get intelligible answers. The person asking the questions should get the impression that the person inside the room speaks Chinese and understands the meaning of their dialogue, but in fact he only mindlessly executes instructions and does not understand a word. Moreover, a person in a room is not even able to gradually learn a language, since meaningless cards for him in Chinese are the only way to communicate with the outside world.

In this example, it is easy to see a parallel with the computer. The computer also operates data in binary code, which does not carry any meaning for it, the machine simply performs arithmetic and logical operations according to the instructions it contains.

According to Searl, a computer (or a Chinese room) is not able to achieve understanding, because operates only with syntactic constructions. The person, in turn, perceives and operates with meanings, and not just with the symbols themselves, which the computer is not capable of. To summarize Searl’s ideas in the article, he is not against the idea of ​​the existence of artificial intelligence as such. Its main idea is that for the existence of consciousness there is not a sufficiently formal program, but a carrier with properties that give rise to intentionality is necessary, but a formal program does not carry meaning and cannot serve as a carrier of understanding or consciousness: “... only a machine could think, and It has been found that it is a product of causal power of the brain ... of intentionality ”[1]. Thus Searle says that intentionality is a necessary property for the emergence of thinking, and this intentionality is a product of some intrinsic properties of the brain. A machine with the same properties is capable of thinking. A classic computer does not possess such properties, therefore, no matter what program is written, it is not capable of leading to the emergence of thinking, and, as a result, to the emergence of consciousness.

“It’s a funky way to try it.” The formal program carries no additional intentionality. It adds nothing, for example, to understand Chinese. ” [one]

(The purpose of the example of the Chinese room was to try to reflect this, by showing that as soon as we put something that really possesses intentionality (a person) into the system and program it with a formal program, we can be sure that this formal program does not carry any additional intentionality. It does not add anything, for example, a person’s ability to understand Chinese).

The ideas set forth by Searl cannot be called complete, which caused quite a few disputes around them. However, these disputes did not lead to a final conclusion. Neither Searl’s opponents, nor his supporters, nor the philosopher himself, came to the solution of important questions, the answers to which should lie at the basis of thoughts about consciousness and AI: “What properties of the brain give rise to intentionality?”, “What is the meaning? And how does it arise? ”,“ What is the process of formative understanding? ”.

In this article I will try to give answers to these questions.

First of all, it is necessary to deal with intentionality - an intrinsic property of the brain - which, according to Searl, is so necessary for the existence of consciousness. To do this, I want to deviate somewhat from the main topic and consider briefly the origin and evolution of living organisms.

Biological evolution can be viewed as a continuation of chemical evolution, which in turn is a continuation of physical evolution. According to modern cosmological ideas, after the big bang, a huge amount of energy / substance was released, which, as the universe expanded and cooled, condensed into a huge number of various particles. The life time of most particles, due to their instability, turned out to be extremely short, and as time passed, only the most stable particles remained, which became the basis for the formation of atoms of known chemical elements. The atoms of the substance interacted, forming various compounds. The most stable of them were accumulated, the unstable disintegrated and merged into stable ones. As a result of the unique conditions on Earth, compounds appeared that were not only fairly stable, but could also duplicate their molecules, these were the ancestors of DNA. Gradually, such replicating molecules, by random recombination, evolved to a state where they could not only reproduce their own kind, but also create favorable conditions around themselves for stable existence and reproduction. This created favorable environment has become the internal environment of bacteria and unicellular organisms, and the control molecules, which can already be called DNA, have acquired the function of supporting the dynamic balance of the organism (internal environment) - homeostasis. Next were the mutations and natural selection that created the living world that we know. The idea that all living organisms are merely complex mechanisms for the propagation of a DNA molecule is not new, and is considered in detail, for example, by Richard Dawkins [2]. Thus, it is possible to trace how the origin of life stems from the fundamental laws of the existing Universe. Roughly speaking, the whole evolution of the Universe is just a chaotic movement of energy, subject to probabilistic laws. What happens is most likely, what remains the most stable in time. A living cell, the smallest unit of life, can be viewed as a huge chemical plant with many complex chemical processes that, individually, each have no internal purpose, but all together maintain a stable internal cell environment that is necessary for the existence and replication of DNA. We say that the goal of all life is survival and reproduction, this is undoubtedly so in terms of biology, but it must be noted that the person himself "assigned" these goals to the living to describe the general nature of the behavior of living organisms. If we are consistent with the above-described reflections on evolution, then we must recognize that living organisms have no purpose or intrinsic intentionality in principle, just as a falling apple has no purpose - it falls obeying fundamental physical laws (we can say that an apple seeks fall, but this does not mean that the desire to fall is due to some inherent intentions of the apple). A living organism is a way of existence of a substance due only to the fundamental physical laws, there is no need for additional magic. I am writing about this to show that in terms of intentionality there is no fundamental difference between the living and the nonliving. DNA is a molecule, it does not have and cannot have goals, only directions are physical and chemical laws, following which it exists and controls the cell.

Later in this article, when the words of goal, need, intentionality with respect to living organisms are used, it is necessary to remember that these concepts are only convenient abstractions used to describe behavior, followed by real processes due to fundamental objective laws, and not obscure subjective factors.

As already mentioned, the main distinguishing feature of living organisms from non-living physical structures is the ability of the former to replicate and maintain a stable internal environment — homeostasis. These are precisely properties (!), But for convenience we will call these properties as goals and return to the fact that the goals of living organisms are reproduction and maintenance of homeostasis. Consider the standard behavior of a living organism.

A living organism cannot passively exist waiting for favors from the outside world, it has to actively influence the environment in order to achieve its goals - to survive and multiply. These goals are the starting points for his behavior. But how exactly to behave? The main question that confronts any living organism at any given time is “What to do next?”. Consider a hypothetical simplest single-celled organism. In the scheme of his work there are three main elements: the external environment, internal state and program (1). Each of them can be the initiator of the movement. For example, in order to answer the main question, an internal state of the body is checked, which determines current needs (for example, nutrition). In a cell, this check is due to the chaotic movement of the molecules of the substance of the internal environment. Separate organelles and DNA interact with chemicals and react accordingly. In response to the internal state, a program is launched that is determined by DNA, and in more complex multicellular organisms, the program is determined by the nervous system. The program starts a process that changes the internal state, which in turn changes the external environment (for example, the organism moves in space). The external environment affects the internal state, and the cycle closes. This scheme is applicable to the simplest organisms as well as to the most complex, including humans. Moreover, this scheme is applicable to different parts of the body, its various systems of different levels. "

Let's remember the considered scheme and we will return to the Chinese room. As I have already said, the principle of its functioning is essentially identical to the device of a standard computer, and the experiment itself perfectly demonstrates the limitations of the machine’s ability to understand the surrounding world. But hey, isn't the brain in the same situation as the person inside the Chinese room? In fact, the brain deals with the same Chinese literacy. The brain sits in the dark of the skull, isolated from everything except the meaningless code that it receives from the senses. If we connect the electrode to the nerve cell, any nerve cell, whether it belongs to the brain or the retina, all we will see is a series of power surges. For the convenience of observation, such pulses are often connected to the loudspeaker, then they are heard as uneven chatter. The nerve cells of the organs of perception — the cells of the retina of the eye, the hair cells of the ear, etc. — emit such a chatter. This chirps the brain and receives from the senses. All that the brain can know about the outside world is that it looks like a chirping rumble of billions of cells. In this buzz there is absolutely nothing indicating the presence of space and time, the existence of objects or people around. Sounds, shapes, colors, smells - it's all quite different from a strange and meaningless chatter. From our own experience, we know that the brain still managed to decipher this code and use it quite successfully for orientation in the world. Why does the brain succeed, but is it fundamentally beyond the power of a person in a Chinese room (or computer)? After all, they are in the same position! In fact, the last statement is wrong - their position is fundamentally different.

The fact is that the method of posing a mental experiment deprived the person inside the room of one crucially important opportunity - the person cannot ask questions outside. The conditions of the experiment say that he can take the cards, manipulate them according to the instructions and thus answer the questions put before him, although the person himself does not understand what these questions are, he does not even understand what these questions are. In addition, a person, operating in the Chinese room, does not pursue any goals with his own manipulations - there is no intentionality. Cards do not affect the state of a person inside a room, they are neither useful nor harmful, for example, for survival. Although it can be said that the cards influence the internal state of the room as a whole, since they force to perform manipulations, but this influence is neutral, it does not bring the system closer to the target and does not remove it, since there is no objective at all except manipulation for the sake of manipulation There is no feedback. If we consider the work of the nerve cell and the signals it receives (chemical, by its nature), then it becomes obvious that they directly affect the work of the cell. They change its internal state. At the same time, the nerve cell has an optimum mode of operation for it and an internal state, the optimum chemical composition. Therefore, the signals do not just change the work of the cell, but also shift its internal state relative to the target (optimal) value. The target internal state is defined by the program, which is recorded in the DNA. DNA is in a position similar to the brain and to CC, because it does not contact directly with the outside world, but only through the medium of the intracellular medium. Thus, the optimal internal state, which is the goal of cell functioning, is both a condition and a cause of the inherent intentionality of this cell, given by the program (DNA).

We can say that the program itself is a carrier of intentionality. But where did the intentionality come from there? As already mentioned, DNA exists and functions according to physical / chemical laws. Replication and other functions related to maintaining homeostasis are DNA properties arising from these laws, and not the goals or needs of DNA. Just as redness is a property of a red apple, not its intention. It follows, as I have already stated at the beginning of the article, that what we call the intentionality of living organisms is not in its essence. Therefore, the question “where does intentionality come from?” Does not make sense. Intentionality is not there, as there is nowhere else. There is only a chain of causes and effects that follows universal physical laws. Intentionality is a human invention and a convenient abstraction, however, I also wrote this. But we will continue to use this concept, since it is convenient for describing the complex behavior of organisms. To summarize what we have written, let us build a logical chain of the emergence of "intentionality":

Physical laws → Chemical laws → DNA + evolution → Replication and Homeostasis → Target internal state → Intentionality (target behavior)

Let us return to the question of "understanding." What is it? When can we say that the Chinese room “understood”? Where is the boundary between the understanding system and the system that is not capable of understanding? Can a single cell have any "understanding"? And if it is capable, then how is a cell different from an uncomprehending one?

You can consider animals, as their central nervous system decreases and is simplified, and you can ask yourself if this animal is capable of understanding? For me personally, it is extremely difficult to grope the line between understanding and not understanding in this way, so I decided to go from the bottom up, starting from a single nerve cell.

The nerve cell, in contrast to the "Chinese room", is capable of not only reacting passively, receiving signals, but also actively influencing their source. Methods of such exposure can be, for example, a change in the number of mediator receptors on the postsynaptic cell membrane, or a change in the frequency and strength (number of mediator molecules) of the generated signals. The latter method is all the more obvious, the simpler the considered neural network. On the basis of the universal scheme of functioning of living organisms (1), we consider the simplest neural network (2) that performs the work of this scheme.


Suppose an external stimulus, in this scheme, is a light source, the receptor (P) is the light receptor on the retina of the eye, (H) is the control neuron, and the actuator (A) is the muscle responsible for constriction of the pupil. Obviously, the scheme works as follows: the light hits the receptor; it generates pulses arriving at the control neuron; which in turn generates control pulses to constrict the pupil; the pupil reduces the amount of light falling on the receptor; as a result, the frequency of impulses generated by the receptor changes, and so on in a circle. The controlling neuron in this scheme does not know anything about the outside world, it does not communicate with it directly. All that is “known” to this neuron is how its “behavior” affects its internal state, i.e. the pulses generated at the output affect the character of the pulses received at the input. The neuron, like the brain as a whole, as well as the person in the Chinese room, is in isolation, all that it “cares about” and is accessible to perception is its internal state. The external environment has an impact, changing the internal state, and, starting from this change, the neuron changes its behavior. At the same time, the goal of behavior is to achieve an internal state or mode of operation optimal for the cell. Going back a little, it should be noted that the target state is not the final goal, which, as already mentioned, is the survival and reproduction of the organism.

Just for the Program information is available only about the internal state of the body. Achieving the target internal state and is a condition for survival and / or reproduction. The target state is not something permanent, but depends on the external conditions and personal experience of the cell. It is also worth noting that for the neuron itself it is not worth the task of sole survival and reproduction, since it is only part of the whole organism. However, to pursue any other goals other than the “personal” neuron (like any other cell in the body) is not capable, due to the limited information it receives. , «» , .

, , - . , : . . , , .. . , , , , - , . , , .. , . , . , , , . , .

. , , , . , , , .

, , , , . « » . «», .

. (, ) ? :

1. . – / ;
2. . , , ;
3. . , , . , .. ( – ).
4. , . / . , , , , ( ) .

- .
, . « »? , . . , . () , , , . , . , , , .

, .. 4- ? , – . , , , . , . – .

– . . , / . , 25 . , . 25 , . . , ( 23 ), . .

, 4- () . – . «» 25 . . – . , , , , , «» «». «» – . – . – . , . . – . , .

, . - , - , - . , , [3]. , , , . .

, , . , . (1) .

, , . , , , . (, ). (1) : . → . → → . → . Wednesday. «», . , , , .

, () , , () , . , « ?». , , (1): « ( )?».


, , .

1) ( ) .
. , . . , – . , . , . . . .

, , , , . – . , . , <-> <-> . , , . , . . — . , , , .

2) .

, , . , , . , , . – , , , . , . , , .

? - . () → . → → . → . (). , , , , . , , , , , . «» . , – , – . . , , , , , . , , .. «» , , , . , .

PS , , . . . . , . , , . . « » , . , . , . , , , .

, . , . , , . , , . , , :

→ → →

, , , – . : « , — , , ». , , , , . , . , , (, , , ..). , . , , , , , , , , , . , , . , . .

1. Searle J. Minds, Brains, and Programs. The Philosophy of Artificial Intelligence / Boden M (ed.) Oxford. 1990. : «The Behavioral and Brain Sciences», 1980, № 3, pp. 417–424.
2. , . . — . from English .: :CORPUS, 2013
3. . : . .: URSS, 2013

Source: https://habr.com/ru/post/405941/

All Articles