Self-training control block. No brain is able to hold enormous “knowledge bases” onall possible conditions of the entire world around. Therefore, one of the reasons why each species of animals occupies corresponding biosphere niche is the necessity to limit the volume of “knowledge base”. Antelope knows what the seal does not, and vice versa. In each separate ecological niche the quantity of possible situations is much less, than in all ecological niches all together. Therefore, relatively small volume of necessary knowledge is required in separate ecological niches. However, if one tries to somehow input /in the brain/ all the information currently available onall the situations which have already been occurring in the world, it would not help either, because the world alters continually and many situations have never ever arose. The “knowledge base” basically may not have information on what has not yet happened in the world. Naturally, the “base of decisions” cannot contain all the possible options of decisions either. “Genetic knowledge” contains only what the ancestors of animals have experienced. They materially cannot have knowledge of what is going to happen. When new situation arises, the system cannot identify, classify it and make decision on it. Even if this situation will occur repeatedly, if the system is unable of self-training it will every time fail to correctly identify a situation because such situations are not contained in its “knowledge base”. The ant runs along the fence, going up and down, and cannot guess that it is possible to easily bypass the fence. Millions years ago, when its genetically input “knowledge base” was formed the fences were non-existent. If one tries to sink a thread on the web the spider will leave this web and will weave a new one because it is not familiar with such situation and it does not know and cannot learn that it is possible to make a hole in a web so that the thread does not interfere. All this is due to the fact that insects as a class of animals are not capable of learning anything. They may be perfect builders amazing us with their sophisticated and fine webs, nests and other creations of their work. But they can only build based on their innate knowledge. They do have “knowledge base” (instincts), but they do not have cerebral structures (elements of control block) capable of supplementing their own “knowledge base” with new existential situations. They do not have reflexes on new stimuli/exciters/. To be able to identify and classify new situations the control block should be able to enter the descriptions of these situations in its “knowledge base”. But at first it should be able to identify that it is a completely new situation, for example, by comparing it to what already exists in its “knowledge base”. Then it should identify the importance (the value worth) of this particular situation for the achievement of its goal. If there is no any correlation between the new situation and the fulfillment of the goal of the system, there is no sense in remembering this situation, otherwise the brain “will be crammed with trash”. By singling out and classifying external situations (identifying them) and finding interrelation (correlation) between these situations, by decisions made and the achievement of the goal of the system the control block learns to develop appropriate decisions. Thus, the self-training decision-making block continually supplements its “knowledge base” and “base of decisions”. But under the conservation law nothing occurs by itself. In order for the control block to be able to perform the above actions it should have appropriate elements. The major element of the kind is the analyzer-correlator. It is the basis whereon reflex on new stimulus/exciter or a new situation may emerge. Its task is to detect a new situation, identify that it is new, determine the degree of correlation between this situation and its own goal. If there is no correlation between this new situation and implementation of the goal by the system, there is no sense in remembering and loading its limited “database” memory. If the degree of correlation is high it is necessary to enter this situation in the “knowledge base” and develop a decision on the choice of own actions for the achievement of its own goal and thereafter to define whether there is correlation between the decision made and the achievement of the goal. If there is no correlation between the decision made and the fulfillment of the goal by the system it is necessary to arrive at other solution and again determine the correlation between the decision made and the achievement of the goal. And it should be repeated in that way until sufficiently high correlation between the decision made and the achievement of goal is obtained. Only afterwards the correct computed decision should be entered into the “base of decisions”. This is the essence of self-training. Only the analyzer-correlator enables self-training process. As a matter of fact, the system’s self-training means the emergence of reflexes to new stimuli/exciters or situations. Consequently, these are only possible when the control block contains analyzer-correlator. Biological analogue of the analyzer-correlator is the cerebral cortex. The presence of cortex determines the possibility of emergence of reflexes to new situations. Cerebral cortex is only present in animals which represent sufficiently high level of development. Non-biological analogues of systems with such self-training control block are unknown to us. Computer self-training systems are built by man and the process of self-training at the end of the day always involves human cerebral cortex. There exist various so-called “intellectual” systems, but full-fledged intelligence is only inherent in human being. Let us specify that there are no self-training systems, but there are their self-training control blocks, because executive elements cannot be trained in anything. There may be systems with simple executive elements, but with control blocks of varying complexity. In order for the control block to be a self-training structure it should contain three types of analyzers: the analyzer-informant with “database”; the analyzer-classifier with the “knowledge base” and “base of decisions” (which is able of classifying external situation on the basis of the information from the “C” informant); the analyzer-correlator (able of identifying the interrelation – correlation between various external situations and the resultsof actions of the given system and transferring the knowledge obtained and decisions to the analyzer-classifier to enter them in the “knowledge base” and the “base of decisions”). Thus, the system with self-training control block is an object which can learn to distinguish new external influences and situations in which such influence may be exerted. For this purpose it has the analyzer-correlator. In other respects it is similar to the systems withcomplex control block. It can respond to specific external influence and external situation and its reaction would be stipulated by type and number of its SFU. The result of action of the system is also graduated. The number of gradations is determined by the number of executive SFU in the system. It also has analyzer-qualifier with “knowledge base” and “base of decisions” and the analyzer-informantwith “database”, DPC (the “X” informant) and NF (the “Y” informant), which operate the system through the stimulator (efferent paths). In inorganic/inanimate nature there are no analogues of systems with self-training control blocks. Biological analogues of systems with complex control block are all animals with sufficiently developed nervous system in which it is possible to develop reflexes to new situations (should not be confused with conditioned reflexes). The analogue of analyzer-correlator is only the cerebral cortex.