What is intelligence. Two directions of development of artificial intelligence. Person. Mechanisms of artificial intelligence.

Mind, intellect, consciousness. . . Consider, first of all, how these categories define the various sources accessible to the reader. We need this in order to more clearly present the diversity of those aspects of intellectual activity that are subjected to scientific study, to more fully perceive the place that model-cybernetic developments in it are aimed at creating "artificial intelligence" systems

Usually, however, the term “artificial” is used only in the sense of “made by man”. It is noteworthy that in the explanation of the word "artificial" explanatory dictionaries reflected "a deep distrust of man to the creations of his hands": "Created by the hands of man, and not by nature; not real, not natural, feigned, not relevant to the essence of the matter." Synonyms: fake, made, insincere, hypocritical, ostentatious, deliberate, fake, fabricated, unnatural.

Mr. Simon outlined four signs by which the “artificial” differs from the “natural”:
" 1. Artificial objects are constructed (though not always quite intentionally) by man.
2. Artificial objects may look like natural, but significantly different from the latter in one or several aspects.
3. Artificial objects can be characterized by their functions, goals and degree of adaptation to the requirements of the environment.
4. Artificial objects are often, especially when they are designed, considered not only in descriptive terms, but also from the point of view of the category of obligation. "

As for the second item of this classification of G. Simon, it should be noted that the opposite situation occurs more often in cybernetic practice: a certain mathematical or physical system (model) is created, which “outwardly” is completely not similar to the “natural” one, but is so organized that in some frames it detects behavior that is almost the same or similar to the functioning of a natural object, under the same environmental conditions.

Three basic conclusions can be deduced from the definitions and statements about "artificial intelligence":
a) the term "artificial intelligence" is used in two different senses: as a designation of a specific research direction and as a name for systems, the development of which this direction is aimed at;
b) among scientists there are disagreements about the possibilities - both fundamental and real - of "artificial intelligence" as a research area;
c) a functional approach is typical for discussing the problem of “artificial intelligence”: a machine is recognized as “intellectual” - in one or another, perhaps very limited sense, if in the considered field of intellectual activity it produces results similar to those given by people; and preferably far beyond those results.

Intellect must be distinguished from imitation of intellect. Imitation is a decision on the model. The task is subjected to human analysis. They also identify the essential features (parameters) of the simulated activity. Further, an artificial system varies the parameters for the production of variants. Read more ...

Imitation is easy to distinguish from intelligence, if you give the created system a task from a fundamentally different area. But keep in mind, imitation is an indispensable tool for creating intelligent systems. Intellect imitates reality and cannot be done without imitation. Only a simulation interpretation system has the ability to self-development. The intellect has to build itself (to complete, to teach), we have the right to create only the core.

Two directions for the development of artificial intelligence

The development of intellectual systems is observed in two directions - expert systems and imitation of rational activity, expedient behavior. First: expert systems are output systems, the basis of which is mathematical logic. Despite the fact that these systems can be useful in areas with a small dimension, they have quite serious drawbacks: when outputting, there are many side effects, combinations of which reach by the number of a huge number, which cannot be processed on a single computer. The second drawback can be considered the monotony of mathematical logic that does not correspond to human logic. The mechanisms used in expert systems are recommended to be used for embedding in software products, giving the latter features of intelligence. The world of expert systems should change, develop, but this mechanism is implemented here is extremely difficult.

Another way is imitation of intellect, when models are used for imitation.

fig.1.1

Should an artificial intelligence system be like us? Undoubtedly, it should be and will be! A reasonable creature, no matter how it is not related to another, will have the common features of reasonableness. Dolphin understands Man, Friday - Robinson, and so on. Understanding occurs on the basis of commonality, it means the similarity takes place.

But how much? What is the level of similarity?

If the similarity in imitation is 100%, then, firstly, it is too expensive financially and, secondly, we built the same thing, but artificially, was it necessary to try so hard? If the similarity is 0%, then this is a useless product.

Truth between these two numbers.

A person and what is meant by a modern computer are similar to many others, for example, have a similar basis (electrical processes), but are not even more alike, the main thing - the principles of “thinking” are completely different. Therefore (and by the results of activity) a computer cannot be called an intelligent machine, although in some cases such an illusion can be created. Our computer will be considered as a good (by the way, carriers are better and better) a carrier for the intellect, but not the intellect itself. A computer is the most convenient medium (relative to other known environments) for automating the problem of intelligence.

For this direction, a serious task is to find an answer to the question: what is the person who is going to imitate? Unfortunately, making attempts to build intellect, we hardly know what we want to build, what is intellect. You can automate only what can be clearly presented, described. Unfortunately, the sciences of man (biology, psychology, medicine, sociology, philosophy, anthropology, and others) do not give us information about the object necessary for conducting AUTOMATION.

In addition, man does not exist outside the environment. If the environment disappears, the person immediately degrades. If you immerse a person in a pool (imitating weightlessness), block all channels of communication (smell, hearing, sight, touch). The brain will refuse to work in a few tens of minutes. Hallucinations, madness is inevitable. A person without a constant flow of information, without processing information does not exist.

The environment, that is, NJ, has a complex structure for the individual. I am not divided into society and the world. Society - public and family. The world is natural and artificial. That is, according to the degree of closeness and distance from the Person associated with the scale . Personality interacts with non-IT information.

For an adequate interaction, the personality builds a MODEL within itself. What model?
First of all, a model is a system. The system is the Elements (that is, the concepts and their properties) and Connections. If the model is not brought into the system, it is mostly useless, since tasks cannot be solved on it. A model that is not a system cannot perform a predictive role.
Model adequately describing the surrounding world, similar to the world. We understand the world, if a model of this world is built inside us - there are concepts that make up the world, and they are connected, if we can match the inner world and the outer one, imitate, reflect the world in ourselves. Model ready for a mental experiment on it. Due to this, we accept, assimilate and store information.
Model building model of the world. Model building itself.
A model that solves problems coming from the outside, on a constructed model of the world.

Person

Living can be considered as capable of evolutionarily self-organizing, aggressively and adaptively interacting with the environment and enhancing its structural entropy system, the internal processes in which are cooperative, the combination of components obeys the rule of nonlinear superadiative addition.

Principal points:
something must rebuild itself;
something must "communicate" with the environment;
something must "consume" the products of the environment, realize the activity, produce results;
something does not have a single center that commands all;
something or their combination realizes evolution for the better.

The level after which they talk about living is DNA.

A conscious subject is recognized if the following conditions are met:
the ability to receive and store (perceive) information, and not bits, but a structure;
the ability to learn on the basis of the information received;
to logically and emotionally generate new information within oneself;
at will give orders to action.

Artificial Intelligence Mechanisms

1. Training
ability to learn in a stationary environment;
the ability to learn in a non-stationary environment;
tuning to the optimal learning system (the intellect optimizes itself);
ability to learn in a team;
a clear understanding of the concepts "I", "not me", "the same as me", "not the same as me";
development and conflict resolution.

2. The ability to classify
Recognition - assignment to one class;
self-education of new classes;
concept highlighting and concept education.

3. Reproduction, including self-reproduction
there must be an evolutionary development;
inheritance of useful knowledge;
providing a mechanism for mutation and selection;
collective perception;
collective security.

4. Primitive reactions (reflexes - conditional and unconditional)
fear, want, need, should, possible;
measure of desires;
when interacting in a team, a potential difference arises, which should cause a response — an action;
feelings: touch, smell, sight, taste, vistibular feeling, sense of three-dimensional space, feeling of a magnetic field (in birds), feeling of another's look, etc.

5. Language as a way of perceiving information
as a way of thinking, storage, formation of structure, organization of information;
how to learn a language from scratch, how to enter a language;
how to teach intangible concepts that can not be shown;
the formation of a metalanguage, on the basis of which any concepts are explained, since it contains global concepts;
real language;
specific ways of communication (hypnosis, suggestion, visualization, information guidance).

6. Knowledge Base
organization of information within your self, the structure of knowledge;
the role of laws (known from other sciences) for thinking, hidden meta laws, basic (simplest) laws;
knowledge of space, of time.

If at least one mechanism is missing, then it is difficult to talk about artificial intelligence. Although the components of AI are much more, for example, a number of mechanisms are not mentioned here (for example, the synthesis of emotions, psyche, input and identification of visual and sound information).

Consider the possible implementation of each of these mechanisms. Separately, it's easier to deal with each of them. Complex behavior is composed of a set of simple acts.