Nikolay Amosov    WORLD OUTLOOK

Life is complex, man is complex, and it is difficult to understand even oneself. The truth seems to be elusive when we speak of living systems.

A model is "necessarily inaccurate"; it can distort the truth incredibly and will still seem logical enough. This is true of verbal model first and foremost. Mathematical and physical models are more reliable. If they are active, however, they simplify an object with the loss of its most significant properties. The true models are those that I call "complete" ones, those that allow the reproduction of an entire system. This is feasible now only in physics, chemistry, and engineering. In biology, the models are not "complete" but generalized ones. It is impossible to make a frog using photographs alone. However, even this incomplete truth is necessary and useful. We are not able to make a frog, but we can control some functions using its scientific models — namely physiological data.

This is how sick people are treated, for example.

Truth is produced by the mind; the latter employs truths (or models) in its activity.

Intellect. This word contains so many elements and is so difficult to define. The mechanisms of the intellect are studied by the scientists of various types. These include theory of their respective branches of science for psychologists and physiologists and ways to create artificial intellect for cyberneticists.

In cybernetics, the mind is an apparatus of optimal control of an object through actions with its models. An example: a physician treats a patient using the physiological models of the sick organism that represent the essence of knowledge. He tries to control the situation in the best way, of course. One criterion of this optimal control is that the patient get well faster and more effectively.

Since electronics embodied in computers now allows us to create complex models beyond the human intelligence that can control an object, mind in this sense can exist without man, it can be detached from him. This is called artificial intellect.

Philosophers, cyberneticists, and people of other occupations often discuss whether artificial intellect can be cleverer than man. This question is always asked at the lectures. An idea is presented: "If an artificial intellect is created by man, how in can be cleverer than the creator?" These doubts are groundless. A walking excavator will excavate much more rock than all the people who designed it. This very powerful machine is the result of the efforts of many people joined in a certain way. The same is true of mental efforts. Men create models, realizing them in objects, words, or formulas. If a team of scientists composes its models in a certain way, an artificial intellect that is more intelligent than any one of its creators or all them together will be the result. The problem lies in "composing the models in a certain way," in the technology of realization and summing up these models. In the same way as with the excavator. Nowadays it is a reality, but fifty years ago, this machine was impossible. At present, we do not have any technology which can "sum up ideas" into a work model of the mind, but there are some prerequisites... Artificial intellect will be a reality — cyberneticists have no doubts about that.

The technology of the creation of an artificial intellect is a serious matter. It has two aspects: what should be used for creating models, and how they should be combined to work well.

"What to use" is the code for models; it is elementary physics. Triplets come from a single cell; there are 100 million of them; there are over 10 billion neurons in the brain. So what about the artificial intellect? This question still has not been solved. Only realistic solution is available today — electronics. I do not want to discuss this problem, but there is no effective substitute for a neuron yet. This fact raises numerous problems for the sceptically-minded.

How can the models be summed up to ensure that these "actions" will be effective and reliable?

Sequence of actions is denoted by the capacious word "algorithm." Sequence in working on a problem. What should be done after what? What information should be used and where should it be collected? The algorithm of intellect presents the order of actions along with the models.

I have been interested in this subject for many years, since my college days. I have wanted to understand the structure (model) of the process and how it all proceeds: ideas, desires, creative activity, will — i.e., everything about which psychologists write in words. Now, I think, I know how it proceeds. Naive boasting will not convince anybody: evidence is needed. It can be obtained from neurophysiology and from creating models of the intellect — i.e., cybernetics. I think the most realistic way is the latter. Physiology will be able to demonstrate only individual "fragments" of psychology for many years to come, since the cortex contains over 10 billion cells. It is impossible to measure and to record even a small part of this. Impossible for now.

However, building an artificial intellect equal to the human one in its capacity does not yet provide sufficient evidence that our brain works in the same way as a machine, but it will advance the interpretation of psychology. Or at least I think so.

Early in the 1960s, I attempted to create a hypothesis about the general mechanisms of the intellect. It was the basis of a large series of models of the simplified intellect created in our department of biocybernetics. We generated many publications on this topic, and dissertations were even defended. The latest book in this area is Algorithms of the Mind.

I will try to enumerate the basic ideas of my hypothesis.

The intellect serves to control objects. It requires auxiliary devices to perform this: receptors — sense organs — or sensors, needed to perceive both the object and subject of intellect — its body. The intellect needs something to influence the object: organs of movement, offectors, muscles, or machines. Their purpose is to amplify the infinitesimal capacity of control pulses, nervous impulses, for example. "A body" is required to ensure the needed energy. Some power installation that will ensure energy for the intellect and the auxiliary devices. Energy is obtained from various sources. Animals get energy from food. The artificial intellect utilizes electricity...

The intellect is discrete: it functions "in portions" — "yes" or "no." But not exactly in this way. Continuous, or slow-going processes are observed at the same time, and they change the "intensity of the response."

The intellect controls an object through models as we have already emphasized. But the models have a structural character. In other words, they should be represented by physical spatial structures: geometrical structures characterizing spatial objects. The problem of determining a method to make this characterization is very complex. Figures in computers or words in descriptions put "the geometry" straight, but for a model to begin to work, the lines must again be transformed into a spatial structure.

The notion of conductivity (or resistance) in communication is important: it determines the strength and stability of the model itself. This is especially significant in the case of the brain where there are tremendous numbers of surplus links between neurons. One group has intrinsic conductivity and consists of models of unconditioned reflexes, while others are formed in the process of education — Pavlov's conditioned reflex. They make up all the models of attitudes, everything what an individual learns for living.

The principle of response training: they get stronger from training but vanish if not used.

Responses of fluctuating conductivity provide complex combinations of models of various sizes and strengths. Phrases made of words are an example. Words are remembered well; their models are strong. Phrases are formed of unstable responses. But the habit of connecting them reflexively can be strengthened by repeating exercises. This is how poems, sayings, and quotations are fixed in memory. These responses provide us with models of various degrees of generalization and clarity.

They are united by various indices.

Any long-term memory is characterized by strong conditioned reflexes or responses.

The activity of models: this principle is equal to the principle of the structural character of models. The activity of models in the cortex is expressed by the number of nerve pulses that are produced by nerve cells and circulated through the links of a model. Any usage of a model is possible only via increase of activity. When it is not active, it is sleeping. All operations with models are expressed in terms of changes in activity. In order to control models, it is necessary not only to increase their activity rapidly but also to decrease it. Therefore, not only should activation — irritation exist in the mind, but its active reduc­tion — inhibition should also be present. Physiologists have been aware of these brain processes for a long time.

One of the most important properties of a living cell is its ability to be trained. The "trainability" of a nervous cell is extrapolated to the trainability of the entire model.

Neurons in various parts of the brain performing different functions have various levels of initial activity and various trainability. Therefore, certain parts of the brain become sources of activity or inhibition for others: it has activating and inhibiting systems.

Finally, about the mechanism of intellectual activity.

Nowadays, everyone is accustomed to automatic vending machines: insert a coin in the slot and get a soft drink or a newspaper; but you must insert the correct coin and not a metal slug. A sensor must recognize the coin and turn on the action mechanism. A simple scheme that calls forth a reflex is rigidly programmed: "irritant — action." The objective is programmed in communication. The stimulus is in the coin. All intellect is limited by recognition and turning on. These automatic coin-operated machines have gradually become more complex: various coins, various requirements on the part of the user, and various actions. The complexity lies in the system of switching over; the "intellect" has remained the same.

A characteristic feature: an automatic vending machine acts in "sequence": irritant — final action. A simple reflex acts in the same way. Further process is expressed in development of inhibitors; the mechanism of action is impeded if the conditions are not favourable, i.e., if inhibition is present. But inhibition can be replaced by activation: conditions control actions.

Now we already have four components: irritant (a coin), action (soft drink). The stimulus is the condition that activates this mechanism, while the inhibitor activates a blocking mechanism. If, to examine this process further, we imagine that one and the same irritant can turn on various and even controversial actions, everything depends on a com­bination of stimuli and blocks.

I referred to a portion of action with the use of mentality as a Functional Act (FA). This is a chain of obligatory elements, of actions with models where each consecutive step is triggered by the previous one, but with the participation of stimuli and inhibitors. The same picture as in the case of a simple reflex (irritant — action) but one response is replaced by several. The responses link up as follows:

Perception — analysis — objective — planning — decision — action.

The most important fact is that stimuli and inhibitors are present in each, and models are used.

Perception is the tuning of the sense organs — the receptors. To look, to listen, to sense. As a result, a time model of the object and of the entire situation are reflected in the brain.

Analysis means switching off the receptors and investigating what has been seen. In the first place, to identify "What is this?" — which means to compare a model in the short-term memory with models in the long-term memory. How this procedure takes place in the brain is a complex question, so let us not go into details, but it occurs approximately like this: a similar model is selected. Upon identification, it is necessary to assess the model in terms of potential usefulness. If the model is suitable, it is necessary to make some prediction concerning what will happen to the object taking into account conditions and concrete situation. Several models are activated as a result. Feelings prompt what can be done in this situation, defining the request and objective. The objectife is a model of a future state of an object which provides maximum satisfation of feelings (requirements). The request is the elementary action necessary to reach the objective. Memory (experience) will prompt the options: what the object and situation can offer, and the feelings choose the most pleasant option.

When the action is simple and desire strong, it follows immediately after the analysis. If the objective is remote and conditions are complex, the next stage is required — and that is planning. In other words: selection of options for action — how to reach an objective faster, with minimum efforts and troubles... In order to do this, the various options for actions divided into stages exist in the memory, as do the cor­responding difficulties for each option. These options are assessed by a balance of feelings: how strongly this is desired and how difficult it is to obtain. If advantages prevail over disadvantages, the next and most dramatic step is taken — that of decision making. In fact, this step is the shortest one: it is necessary only to trigger the performance of the previously selected plan of action. But some effort is required to do this. Indeed, in the analysing and planning stages, there is almost no responsibility — everything can be cancelled and no action taken. The effort, i. e., the activity of models at this stage is not too great — just to recall, to retrieve data from the memory. Deciding to act is a different matter entirely. In this case it is necessary to overcome the resistance of the object, to brace the muscles. Therefore, a decision is required to begin the action.

Action itself implies implementation of the plan decided upon despite all expected and unforseen difficulties. It requires obligatory mobilization of strong enough feelings to ensure activity.

The notion of stimulation and activation of the models takes place at each consecutive stage. The Functional Act (FA) goes as smoothly if the "route" is well known as a result of repeated exercises, i.e., if the response has been properly engrained. In this case, the FA occurs without a hitch and does not require many stimuli. But we must not forget that everything new is much more difficult.

Another important notion connected with modelling is reality. Reality excites feelings, depending on its significance for us in the sense of meeting our needs, that is. Feelings are stimuli for actions: the stronger the feeling, the greater the tension it can provide.

The future also excites feelings and is also a stimulus for action. The major part of our actions are caused by future events, awards, or difficulties. We find ourselves in the present, but at the same time we live, act and aspire to the future. We study in order to work, to obtain, to give, to insure other pleasures in the future. We do research, expecting results and satisfaction in the future. When this "future" comes — when it becomes a reality — it seems quite short as compared to the lengthy time we spent waiting. And satisfaction when the goal is reached is usually much less than we thought when the effort was made... This is how strangely our minds work.

The significance of future events — rewards or punishment in the sense of excitation of feelings and desires (stimuli) — is defined by two indices: probability of realization and expectation time.

Probability of realization in the future we define by experience and knowledge, but taking into account features of the particular personality, i.e., "the degree of optimism." The pessimist always exaggerates the probability of the negative and underestimates the positive. The optimist does the opposite.

"The time factor": how long the wait will be and how it effects the stimuli also depends on the personality. Impatient individuals are capable of making efforts if the reward is visible. Stubborn persons are ready to work for years when they see chances for success.

It so happens that these very personal features, such as optimism and tolerance have as much of a place in the algorithm of the mind as feelings do, by the way. They are put not in verbal, but in digital form — time coefficients and probability of realization. They are used to calculate simuli and inhibitors when the planning of future actions is under way prior to decision-making.

The Functional Act. It can be very simple and short — it can blink, for example. It can be extended for many years in the case of the most persistent and goal-oriented individuals. Long FA are comprised of shorter ones, the latter are comprised of even shorter ones, and so forth down to discrete motions. This is called an FA hierarchy (or storeys). Each short FA has its own goal; this goal itself is part of the general objective of the "main" FA. The same picture can be observed with stimuli: short-time actions have special stimuli, but, moreover, they share part of the "main" stimuli of the common FA.

Man is motivated by needs-feelings and by so-called convictions. More details about them can be found below. Yet I can say: they are different and even contradictory. They produce different FAs that interfere with and contradict each other. Each of them has a storey-like structure.

As a result, the plans of many Functional Acts are permanently contained in the human brain: some of them have a vertical stratification, some of them co-exist peacefully, and others are constantly counteracting or conradicting each other. If you observe your thoughts and actions carefully, you will see this complex mosaic from which one FA at a time must be chosen. Because only one thing can be done at a time, one FA should be completed, or at least one separate stage should be.

Let me turn back to these stages once again. They are of two types. In one type, the models depend on the external factors and switch on the receptors and effectors (muscles) — these are receptions and actions. Other stages, such as analysis and planning are entirely "theoretical." They present models for actions in the memory alone, without receptor or muscle involvement. These stages also require stimuli: they also require energy to break the resistance of the inter-model links or reflexes, however much less energy than is required for motions to overcome the resistance of the outer world.

Observe this process in yourself, and you will see that the major part of your FAs remained incomplete: they stop in the "theoretical" stages. First you saw or recollected something and assessed it; then followed desire — an objective — and planning may even have started, but when all the pros and cons were evaluated, it became clear that the stimuli were insufficient. So you drop the objective... The FA came to an end without any result. However, it remains in the memory for some time and it can be used in other circumstances.

How can we orient ourselves in this tremendous number of models? The models are interconnected according to very different principles. For example, models belong to one object but are of various degrees of generalization. Or perhaps they belong to the same FA or are linked by the unity of time, actions, space, feelings. Each model has its own activity; its primary impulses try to penetrate other models and overcome the resistance of the reflexes. Strong generators of excitation, such as feelings and desires, are observed among minor, half-forgotten subject-models with low activity. Or among subjects linked to them, or plans for important actions or the images of their objectives. All these models interact with each other and try to reach the action stage through a corresponding decision... Unfortunately, actions are limited and rigidly bound by their models: there are trained motions and nothing more. When one sequence of actions is in operation, others cannot interfere with them: to do so, one must stop and design another Function Act.

I don't know whether my explanations are clear or not. I've tried, at least, to show what a jigsaw puzzle of models exist in the cerebral cortex. To enable a man or an animal to choose a Functional Act, which is the most important at the moment and to complete it — i.e., to reach the ultimate goal — there must be a certain mechanism that protects it against all other models which are nothing but white noise in this case.

I invented this mechanism for our models of artificial intellect some twenty years ago. We called it a System of Amplified Inhibition (SAI).

The SAI provides for the additional amplification of one model of the intellect which is the most active and for the subsequent inhibition of all the other models. A selected model is amplified for a short period of time; then another, the most active of the existing models is amplified. During the short interval when the SAI is switching from one model to another, amplification does not exist, as it were, and the activity of all models corresponds to their condidtions — i.e., to the level of their independent activity (excitation). A SAI-amplified model increases its own activity which can be transmitted to other models via the existing, well-established linkage. Thus, SAI switching affects a number of subsequent cycles and provides additional training for the model in question.

The most active (excited) model is the one that plays the most important role in a given moment. It assumes this role, as a rule, due to its links with strong emotions. The second factor that accounts for this situation is the model training in the course of its previous activity. In short, the most active model is enhanced by the SAI; all others lose some of their activity. The secret lies in the fact that amplification is switched on for just a short period; this process can be repeated only after some cycles. Thus, only one model is subject to amplification at a given time, but it is not fixed forever. Amplification is continuously transmitted from one model to another, and a new model may move to the "forefront," acquiring prominence among the rest.

The diagram shows the System of Amplified Inhibition (SAI). At the given moment, "consciousness" operates with model P₂. The next one will most probably be model D₂, since it receives energy from P₂ via a well-established channel and, when the SAI reorients inself, it will become the strongest.

The SAI enables the simulation of psychological concepts and notions in the following fashion:

A thought is the model being amplified at a given moment. Try to follow your thoughts, and you'll see that they can be represented by images, emotions, actions, sensations and words which denote these concepts. All of them are models. A thought cannot be fixed on one object, therefore, the SAI is switched to other things. However, one may return to the initial thought after a number of "digressions." The more frequently we return, the stronger the thought becomes until it is transformed into an idee fixe as a result of model training. We forget it when life forces us to direct our attention to other, more significant things.

Attention is mechanism which ensures that the SAI will switch to the model which is enjoying priority at the given moment. That is why we say "to attract attention." Anything can attract attention: things, their qualities, recollections, any thought, etc.

The conscious level represents a chain of SAI-amplified models which are the most prominent and have been embodied in the form of thoughts. They are held in the memory in the sequence in which they originated. This chain reflects, as a rule, the orientation of man in question in space and time, his relations with other human beings, an assessment of his actions and even thoughts.

The subconscious level. How many disputes have been centred around it! The meaning of the subconscious is now commonly known: some actions are seemingly performed without any preliminary plans. We discover them only after these actions have been completed or perhaps we do not even notice them. They are, as a rule, short and automatic actions.

The subconscious is easily reproduced in our models of the intellect: it is shown as the circulation of activity between models which are not amplified by the SAI. On the contrary, they are somewhat inhibited but not to the level at which all activity stops. The subconscious level, therefore, provides for the performance of integral functional acts if the participating models, and the linkages between them are well trained. These are usually simple, short actions which are based on the principle, "see and do." For instance, we may walk and think about something complex and important: we walk automatically, and the spot to place a foot is chosen subconsciously. All memorized simple movements are controlled by the subconscious; the conscious level switches on a Functional Act that consists of a sequence of similar (homogeneous) simple actions — steps, for example. Just have a look at your early morning pattern of actions... I doubt that you ponder on every movement. Pavlov called a similar phenomenon in dogs "dynamic stereotypes."

The subconscious level plays a tremendous role. Not only due to the fact that it controls memorized movements. What is more important is that the subconscious prepares models for the conscious level. It is the SAI-inhibited models that exchange energy among themselves: energy is accumulated on the "crossroads" of linkages, making a certain model so active that the SAI chooses it over the course of the next energy redistribution. Thus a model becomes a thought. Everyone is familiar with wild thoughts which originate totally outside the framework of the preceeding ones and the current situation. We are amazed when we suddenly recall a forgotten word or image: "Where does it come from?" More often than not, we can't find an answer to this question. But it is the subconscious that accounts for this phenomenon: the word in question acquires the necessary level of activity and captures the SAI, as it were. These thoughts seem strange to us; however, they are predetermined by the situation, by the pattern in which activity goes from one model to another. We are unable to trace the activity of movements: they are like an iceberg —90—95 per cent of them are inhibited or hidden under water, as it were. These models and activity movements are not amplified by the SAI. Most emotions also operate on the subconscious level, supplying energy to other models and guiding behaviour.

Moreover, the subconscious automatically monitors certain objects from the external world, and the conscious level is switched on when changes in these objects reach certain limits.

In short, the subconscious is a marvelous thing!

A question may arise in this connection whether it is subconscious that determines all our actions. I doubt it. A chain of the SAI-amplified models guides the evolution of thought, because these models send energy ahead of them through associations. Therefore, most thoughts are quite logical and interconnected: we can trace their origins. These logical thoughts form our conscious behavior and thought. The latter also relies on the subconscious, but the subconscious in this case does not act as the main motivating force.

All the phenomena I've described above in outlining my hypothesis about the mechanism of the reason are visible both in man and animals. What I'm going to present now will show the differences between them. Perhaps, they explain the specifics of the "new" cerebral cortex in the frontal lobe of the human brain. Its cells and the linkages between them are characterized by enhanced training. Thus, memory has been expanded. Physiologists have learned that man quickly forms conditioned reflexes. We are capable of retaining in our memories a long sequence of sounds: a phrase, for instance. We will regard this phrase as a single whole. A dog memorizes only words or short combinations at best. Moreover, man has a perfect locomotive system — the larynx and arms. They are not just instruments, for there are corresponding large models of complex movements. Hence, the develop­ment of speech and labour.

The "new cortex" accounts for three important properties of the brain: the ability to learn, adaptability and creativity.

The ability to learn implies the ability of humans to retain in their memories what they see or do and to transfer models from short-term memory to long-term memory and to accumulate them in large numbers.

Adaptability means the ability to vary the activity of innate requirements — i.e., the emotions — and to create new centres of higher activity which guide behaviour. In other words, this quality of the brain enables us to transform congenital stimuli and inhibitions. Both properties are based on the enhanced training of cells and the linkages between them.

Creativity is the basis of everything. This parameter may be interpreted as the generation of new models in the cerebral cortex, either by producing new combinations from simpler and known elements or by summing up and generalizing a number of similar models. The latter process results in the emergence of a new notion. The simplest form of creativity can also be found in animals: they develop by chance new techniques or skills (combinations of movements) and memorize them if they are helpful. The trouble is that the new skills acquired thusly are not transmitted to other animals and die with their "inventor." It is true, however, that animals possess a mechanism of imitation. This enables them to learn from their parents various techniques of self-protection and food acquisition. Creative "discoveries" are also transmitted, but only partially, and their share in terms of total knowledge is negligible.

Man is completely different. Spontaneous creativity is typical of him. An "accidental" model-thought is retained for some time at the conscious level; and if it is followed by other models, their sequence is recorded and retained in the memory. Thus, new and Useful combinations of models can be generated and memorized in a spontaneous fashion, and a creative act can be performed. Higher levels of societal development have produced more effective special algorithms of creativity, for instance, designing which can be learned.

Spontaneous creativity helped man develop speech — a combination of sounds which represent the models of images and can be transmitted to other people. Signals are typical of all animals, but human speech diversified them and increased their informativeness.

The means of labour and struggle, together with speech, transformed a herd of prehistoric humans into an integrated society with a special code and set of models. Society is a self-sustaining entity: new members learn first through imitation and later, according to a special syllabus. The latter is introduced once a society has indentified itself. Progress accelerates with the advent of written language: a chain of models is no longer disrupted by the death of an individual, since it is embodied in texts.

The importance of speech is enormous: it is an economic code to express complex image-based models, a means of comunication and information storage, a tool of education and training. Inner speech is a mighty lever for self-improvement.

Step by step, we have now come close to stimuli and inhibitions. We have already approached this issue but only in abstract terms: there are stimuli which supply positive energy for activity, and there are inhibitions which decrease this energy and even annihilate it. Let us examine the inner workings of this mechanism.

Energy for operations with models — to turn on a new model or to sharply raise its activity — is derived from four sources.

The first is the outside world. It excites the nerve cells of the sense organs (i.e., the ears, eyes, skin, etc.); this activity goes along the innate tracks to the brain cells and further, through the stages of functional acts. The pictures of the world are recorded in the form of temporary models; they in turn excite models of the long-term memory, which help to identify objects, forecasts and evaluate actions, etc.

The second source is the innate activity of the nerve cells or of their models. This activity is typical of every neuron; however, if it does not receive energy from other cells its activity is reduced to the minimum. On the contrary, when a cell receives outside stimuli for action, it is trained, and its inner activity increases drastically. An increase may be so high that the complexes of these cells which form a closed circuit can maintain enhanced excitation together.

The third source stems from the previous one: it implies "beliefs." In model terms, beliefs are active wordings which are well-trained due to frequent use. They define good and evil and serve both stimuli and inhibitions, acting as the coordinates of conscience: man checks all his plans for functional acts against ethical norms, authoritative recom­mendations, etc. Moreover, beliefs per se may become a source of functional acts, compelling him to take active steps in the name of his ideas.

Finally, the fourth and the most powerful source of "brain energy" is provided by biological needs. To avoid possible accusations of biologization of man, it is better to couch this concept in the following terms: "biological needs modified by social education." Man is a social animal. At the same time, he is a part of nature and a biological creature.

Needs (feelings) are structurally present in the nerve centres. Their neurons are very active by nature, but they receive excitation from the body — from receptors and the outside world. These irritants may inhibit the centres — fulfilling needs and satiating centres — or, on the contrary, the irritants may excite them. The simpliest example: the need for food and feelings — hunger and satiation. If you do not eat for a long time, the glucose in the organism is spent and its level in the blood is reduced. This signal is received by the hunger centre and activates it. The centre demands food, as it were. The receptors of the empty stomach also help to enhance the feeling of hunger. Once he has eaten, the individual feels satisfied, the glucose level goes up, his belly is full, and the satiation centre is excited, while the hunger centre is inhibited.

It is more difficult to explain complex needs — communication, for instance. It seems that there are appropriate centres; however, they are excited not directly by the receptors of the eyes or ears, but by information about living objects which belong to a given species. This information is processed in the brain. To put it in simpler terms, the communication centre is excited when a man does not see other people, does not hear them or communicate with them over a long period of time: it is similar for eating. Hence, a need for a company. This need is expressed in a vague feeling of loneliness. A certain "dose" of com­munication will inhibit the centre, and a man may enjoy solitude. To simulate this situation, one should assume that some image of man, but not of an animal or any other object, is innate in the brain. This image is congenital like the cells which process glucose. I don't yet know how nature contrived something so complex.

Let us try now to understand needs and feelings. We will begin with the nomenclature: if man has invented several hundred words which designate feelings, this does not imply that the numbers of feelings are equally great. Most words describe combinations which are treated as a single whole.

For the sake of modelling, we use three types of feelings: instincts, complex reflexes, and integral feeling ("pleasant-unpleasant"). The first two categories will be discussed in the passage about the personality, and the last one should be reviewed here.

The integral feeling of "pleasant-unpleasant" is represented by nerve centres in some areas of the subcortex. It seems to me that they embrace all specific manifestations of the feelings: positive feelings are governed by the "pleasant" sectors, and negative, by their "unpleasant" counterparts.

The algorhithm of the reason implies that this integral feeling acts as the main regulator of behaviour; man always tries to achieve the maximum of positive sensations or reduce negative ones, at least.

The "pleasant-unpleasant" centres sum up the state of all specific manifestations of feelings or, to be more precise, the state of their positive and negative components. The result is a certain sum total with a positive or negative sign. In models, we call it a "level of mental comfort," LMC.

The feelings have a friend and an enemy combined in one phenomenon — adaptation. Not a single feeling stays at the peak level for long: it inevitably gravitates towards zero or equilibrium. A mechanism of adaptation is unclear yet (at least for me) but its existence is obvious, and adaptation should be present in models. Thus, there are two opposing processes: training that increases activity, and adaptation that reduces it.

Now, some words about emotions. They are often confused with the feelings, since it is difficult to distinguish between them. Where does fear end and horror begin? Similar quiestions can be asked about irritation and anger, a bad mood, sadness and grief or sorrow. The difference seems to be a qualitative one, but emotions add something to feelings, as it were. I tend to believe that this addition implies an enhanced production of hormones (adrenalin?) which change the regulation of both body and brain. The SAI is excited; the conscious level is tense and operates rapidly, and the subconscious is suppressed: extreme subjectivism dominates all. Operational is the feeling that switches on the emotion. Nature has created emotions for emergencies, for extreme situations when the very existence of the animal is threatened. In other words, emotions are designed to concentrate power in the neccessary major direction, switching off secondary areas: there is a pressure limit that the organism can withstand. However, this picture is true in the case of anger and horror. Happiness and grief are manifested in a gentler way, but even there, the selected models are very excited and the rest are inhibited.

Does any type of intellect, including an artificial one, needs emotions? I doubt it... Feelings are more than enough.

Yet there is another interesting issue: character. In fact, the notion of character is a very vague one, while models must be precise. We define character as a human ability to endure tension, both in scope and duration. Tension represents the level of activity of models; their activity is connected with the properties of neurons and the SAI. Strength of character is measured and described in terms of its weakness. Weak people tire easily, they are unable to develop a high level of tension and maintain it. Strong people are insensitive to tiredness, as it were, therefore, they are capable of performing larger amounts of work. Likewise, bravery can also be described in terms of insensitivity to fear. Well, these are only models and nothing else. What is the situation with character in the "living" intellect, I don't know.

Here, I may finish my simplified presentation of the hypothesis about the Algorithm of the Intellect. Unfortunately, it is much more difficult to translate a hypothesis into reality than to invent it. At the same time, hypotheses help us in our quest for the truth.

We in the department of biocybernetics — Ernest Kousul, Lora and Alexei Kasatkins and myself — are not dabblers in the field of development of the artificial intellect. We have developed several operational computer models and two self-propelled carts equipped with "intellect." The inverted commas are a real must here, since tremendous imagination is required to use this word in connection with an electronic device capable of finding its way among the trees and pits in a park and locating the target or destination of its journey. However, this device has "muscles" or an engine and wheels, to be more precise, and a minimum of psyche: models of objects, properties and actions, feelings (deprived of emotions!) and receptors. What is more important, it has the System of Amplified Inhibition that reproduces the conscious and subconscious levels. Who else can boast this? Just imagine: Conscious and Subconscious! Our man-made "intellect" is short of models (from 20 to 200 in various types) and linkages — up to 2,000. These numbers are extremely low, as compared with the millions and billions in the human brain.

There's the rub! One must reproduce numerous models capable of generating energy and transmitting it through linkages to other models, forming new linkages and ensuring training. This is needed to maintain the links between all models and the System of Amplified Inhibition that controls and regulates them. Yet, it is impossible to create an intellect with the required number of models. Otherwise, it will remain a substitute that can be used only to illustrate the main tenets of the hypothesis about the algorythms of the reason. Is the situation hopeless or not?

Well, there might be a way out, although, I'm not overoptimistic about it.

We have tried two main approaches: an artificial intellect as a programme for digital computers and as a special electronic device that may be called an "analog intellect." We are convinced now that a digital computer cannot produce a fairly complex universal intellect, only pseudo-intellectual programmes. An analog machine provides broader possibilities, but it requires special elements and linkages. I'm sure that an artificial intellect can be created. Most experts on cybernetics are also of the same opinion; however, few of them realize what it means.

For instance, let us take feelings. Why does the artificial intellect need them? Maybe it can do without them? But the thing is that the intellect, which is designed to control complex processes and to attain the desired objectives, must be governed by optimization criteria, and these criteria act as the feelings of "living" intellects. Once the intellect outgrows the level of an automation, it simply cannot exist without feelings.