Artificial Intelligence (UI, Artificial Intelligence, AI) is a computer science specializing in the creation of machines showing signs of intelligent behavior. The definition of "intelligent behavior" is still the subject of discussion, most commonly used as a standard of intelligence by human reason. John McCarthy came to 1955 for the first time.
Artificial intelligence research is highly specialized and specialized, and it is divided into several fields that can often not be linked. The whole research is also divided into several technical problems; some of the subfields deal with the solution of specific problems, some of them, for example, to use specific tools or to achieve specific applications. The question of whether it is possible to construct artificial intelligence is also closely related to the problem of consciousness, the question of calculations carried out by the human brain itself or the question of the evolution of cognitive abilities. Similar philosophies of artificial intelligence are similar dilemmas.
The main issues in artificial intelligence research include thinking, knowledge, planning, learning, natural language processing (communication), perception and the ability to move or manipulate objects. Achieving general intelligence is still one of the main goals of research in this field.
From the psychosocial point of view, artificial intelligence is one form of non-human intelligence.
The understanding of artificial intelligence, especially in the general public, is influenced by the lack of knowledge of what is possible and what is not possible in artificial intelligence, which is further supported by the boundless imagination used in "Science Fiction". In order for artificial intelligence to be considered as a discipline, it is necessary to determine these limits and to eliminate (sometimes even magic-casting) conjectures. These limits are determined by the properties of the tools available to artificial intelligence. First, it is necessary to say what is artificial intelligence. Artificial intelligence consists in constructing and using models of human activity (processes) that are considered intelligent. This activity is generated by the structures of the human brain, the real world. Since its inception (from the mid-20 century), this modeling has taken two paths:
1. Modeling the external manifestations of intelligent human activity
2. Modeling of the recognized human brain structures, currently neural networks
Artificial Intelligence has chosen a computer-based computer program for modeling (modeling). The computer can be supplemented by sensors (physical, chemical, biological, etc.) and actuators (tentacles, motion devices - wheels, belts, legs, etc.) and thus complex equipment can be created - cognitive robot, lunar vehicle, These computer add-ons are not essential now because we only need to monitor the processing capabilities of the computer program.
Programming languages that make it possible to build a computer program are among the artificial formal languages. The basic feature of these languages is the exact interpretation of all their language constructs and all operations above them. It's an artificial abstract design. It is understood that natural human vague, emotional and subjective interpretations, called connotations, must be removed and replaced by exact interpretations. It will achieve the prohibition (annulment), internal vagueness, which will also expel human emotionality and subjectivity, see Vagnost. With this interference, everything human (vagueness, subjectivity, emotionality) disappears, and the result of this obliteration is the arid desert of a soulless machine - an artificial formal language (computer). The meaning of each language construct (symbol strings) and each operation above these constructs is then accurately outlined (with zero internal vagueness), so that every knowledgeable person knows no matter what they mean. Objects with so precisely defined elements form a group that we call an exact world. This includes not only computer languages, Turing machine, but also mathematics, formal logic, exact games (chess, lady, card games, etc.), exact science. Another disaster for the exact world, caused by the prohibition of inner vagueness, is:
Loss of inferential momentum
By demonstrating inner vagueness beyond the boundary of the exact world, we have shown not only the human spirit of invention living in hypothetical vague imagination and sensory language but also the inventive ability of self-movement of thought. Thus, we have lost the inference of the inference in the exact world. The loss of self-mobility of inference, the impossibility of transmitting it to a world with forbidden internal vagueness, is a step from man to soulless machine; is a step from the living to the inanimate in the informative sense. For example, in mathematics, a human movable must look for a way to find out how mathematical relations can be chosen to obtain the desired (final) relationships. In a simple example, this is shown on the Exact Science page. What inference applies to mathematics is also true in card game or chess. The meanings of cards or chess pieces are known to the well-known man, as are the rules of the game. The player (movable) must apply his / her intellect to select the strokes in play according to the rules. If the movement in the mathematical derivation or the exact game is to be programmed for the computer, the activity of the initiator must be programmed. The programmer has to program the role of the engineer (mathematician or player) so that, after each game step or deriving, the (programmed) movable is able to generate the next step. It is not appropriate to search for a magic machine on the computer, its magic is as empty as the magic of stacks of cards, or figures on a chessboard, lacking a movable, missing a life. Hybatelem is in any case a person, his intellectual ability. Unlike other machines that process or process mass, the computer processes information, but this difference must not be misleading. Nor must he deceive the knowledge that the program can be created in such a way that it can change itself, because despite all the changes, it will be part of the exact world and can not leave it. The key feature of the machine is the quantity (sometimes precision, manually unattainable) of the processed entities (matter, information) and the use of external energy input for this activity. The wheeled excavator in the surface mines with its mass of coarse coal opens up possibilities that would not exist if its activities were to be supported by people (thousands of people) with pebbles and shovels. The machine with its amount of performance (with a speed of activity compared to a man) allows the realization of activities, otherwise unrealizable, whether it is rock mining or information processing.
That is to the tool - a computer that has artificial intelligence available for modeling. It is important to keep in mind that programming languages (and so on, computers) belong to the exact world, they are exact machines, and their exactness lies in the perfect connection of human psyches with the meanings of language constructions and operations over them.
Artificial Intelligence solves two types of problems, depending on whether they are related to the real world or not. Those that do not relate to the real world are from the exact world, for example, exact games (see Exact) or mathematical proofing (proof of theorems). This is the modeling of the exact world with the exact world, and the modeling is reduced primarily to modeling the motif. The most famous is the modeling of the chess expert's moves. Limiting obstacles can be the complexity of algorithms (see also Asymptotic Complexity) modeling movers, and possibly, for example, a number of combinations of chess disposition or other artificial intelligence models.
Moreover, the situation is complicated as soon as real world problems are concerned. The only bridge between an exact and a real world is a tool we call quantity (mechanical strength, ion concentration in solution, intensity of illumination, etc.). It is common to both worlds, because in an exact world it is precisely defined (with a zero internal vagueness of its interpretation), so that every educated person without any doubt knows its meaning and in the real world is an elementary measurable probe into this world, and thus his measurable elementary representative. It is the cornerstone of an exact science. Problems of artificial intelligence related to the real world need to be divided into two categories.
The first one is where the Artificial Intelligence Model uses knowledge about the real world, such that it derives knowledge from a certain set of (background) knowledge from hidden knowledge but deductible from it. Such a model is called an expert system. Since this model must be part of the exact world, knowledge must be written in an artificial formal language (mathematics, formal logic, programming language), ie in a language with an exact interpretation, and therefore it can not be a native language with a vague, emotional and subjective interpretations - connotations. This knowledge must therefore be acquired through the method of exact science. It should be noted that the knowledge gained from the natural vague, subjective and emotional human knowledge expressed in the native language is not transferable to exact knowledge, which can be described in the exact formal language, see Vagnost. If the natural language language constructs are embedded in the exact world (mathematics, formal logic, programming languages), we must give up their natural, human-assigned meaning, because in an exact world there is no agent who can identify, use and process it capable only of human psyche). Such language constructs can be processed as any string of symbols, but without their natural interpretation. Either ignorance or ignorance of the above-mentioned possibilities of entering into the exact world, mistaken attempts to use natural language knowledge gained from natural human knowledge have been discovered at the beginning of the thoughts of artificial intelligence.
The second category of problems related to the real world are the intra-psychic processes on which the human intelligence, that is, the processes generated by the real world. These processes are inherently connected with internal vagueness, and so there is no bridge to the exact world, for the above-mentioned bridge, which is a magnitude, requires a ban on inner vagueness, and that is inherently part of the inner-psychic processes. Therefore, the branch of artificial intelligence, which we have described above: Modeling the external manifestations of intelligent human activity does not even attempt it, it has no tools to do so. However, the second branch of artificial intelligence, led by modeling the activity of neural networks, must be noticed. Great hopes are being put into it, but it is necessary to determine to what extent this model can bring closer to the real human brain, ie processes with inherent inner vagueness. Historically, the first mathematical model of neural activity was presented by Warren McCulloch and Walter Pitts in 1943. What is important is that their (mathematical) model belongs to the exact world, and that all their followers go along this path when the original simple model of a neuron is transformed into increasingly complex forms (according to the creative imagination and professional experience of the author) but still as a mathematical (computer) model, a model of an exact world with forbidden internal vagueness. The real inherently vague processes running in the human brain can not be modeled by an exact world, that is, a mathematical language or a computer. A very thorough study of the structure and activity of neuron carried out under an electron microscope is presented by prof. Stuart Hameroff in his book. In it, the neuron is referred to as a very complex unit, having its own autonomy of behavior based on the processing of a vast amount of information used primarily to create its own (intelligent) decision-making capabilities for co-operation with other neurons. Such neurons are approximately 100 billion in the human brain. Since the inherently vague processes of the human brain can not be modeled by the exact world, other modeling tools, probably biological, must be sought. Intra-psychic processes with their inherent connection to intra-psychic vagueness differ from all the processes studied so far in the real world, such as physics, chemistry, etc. As a search for a new path (in this case, a step away from the path programmed by artificial intelligence) use live brain structures, such as rats instead of computer models of neural networks. These living brain structures are, through a suitable interface, involved in artificial (computer) processing systems that are part of, for example, cognitive robots.
In summary, therefore, we can say that the limits of artificial intelligence are outlined:
- Complexity of algorithms
- For real-world themes, it is also necessary to use only the knowledge acquired through the artificial knowledge of exact science, written mathematically (by the programming language) represented by the relations between the quantities. It is not possible to use inherently vague knowledge obtained by natural human knowledge, represented by vague, emotional and subjective natural language. Nor can they be translated into formal language. Since an artificial formal language is capable of representing only the real world knowledge that has been obtained by artificial knowledge of exact science, and this is only a tiny fraction of human knowledge, the applicability of artificial intelligence in this respect is very limited.
- Inherent vagueness of internal psychic processes. There is no language tool available to describe the intrinsic, inherently vague processes of the human brain so that they can be modeled by the exact world - the computer. Thus, even the neural networks modeled by the exact world cannot be a sufficiently adequate model of the real, inherently vague processes of the human psyche in the real world - the human brain.
Possible path of further development
The inherently vague processes of human psyche have their material carriers - biochemical processes, apparently at their core, which can be described by chemical and physical laws. If the principle of creating a process environment with inherent vagueness is recognized and determined, these processes can be imitated artificially, perhaps even in a nature other than the biological essence of the human brain. It is the discovery of the principles of the environment with uncontracted inner vagueness, and thus the understanding of the principle of life in the informative sense.
You can also find related information in the Turing test.
This comparison is also based on the idea of Turing's test, which Alan Turing wrote in 1950 in his article "Computing machinery and intelligence". In short, he claims that we can declare the machine intelligently, not recognizing his linguistic output from the linguistic output of man.
The argument of a Chinese room is often seen as a counter-argument to Turing's test. He thinks there might be a machine that would simulate intelligent behavior with a ready set of responses to all possible questions without "thinking" over anything.
You can also find related information in Neuron Network.
Artificial neural networks in artificial intelligence have a pattern of behavior of corresponding biological structures. They consist of computational models of neurons that transmit signals to each other and transform them through the function to transmit to other "neurons."
Related information can also be found in Genetic Programming.
Genetic programming, strictly speaking, is not a means of solving artificial intelligence problems, but a general programming procedure that instead of writing a specific task solving algorithm seeks this approach through evolutionary methods.
Related information can also be found in the Expert System article.
Expert system is a computer program designed to provide expert advice, decisions or recommend solutions in a particular situation.
Expert systems are designed to handle non-numeric and vague information to solve tasks that are not solvable by traditional algorithmic procedures.
Search for state space
You can also find related information in the Status Space Search section.
Particularly when creating algorithms for solving classic games (chess, ladies), it seems expedient to define a set of states in which we can get in the game, allowed strokes or transitions between states and the start and end positions. We then seek the way from the initial states to the ending states that represent our success.
Because the state spaces can be large (for example go go) and in some cases endless, smart methods of trimming inappropriate paths and positioning have to be chosen.
See also Data mining for related information.
Large data sets (often stored in databases) about a system are not usable and understandable even though they contain information and patterns of behavior of the monitored system. Knowledge mining methods convert data into a compact and explicit form describing a system that is more usable.
In a broad sense it is not only about the processing of elementary data (numbers, strings, categorial data) but also the processing of sound, images (Digital Image Processing) videa, native language (see natural language processing, corpus) and bioinformatics (bioinformatics).
The outputs are different for different tasks and depend on what we want to use them and what (and how well) we can extract.
Related information can also be found in Machine Learning.
1979 overtook the world's master computer in the game of the backgammon.
The royal chess game has been the subject of analyzes since the beginnings of computer science. The solution to the problem has been associated with intelligence from the outset, but winning does not have to mean greater intelligence. 1997 defeated Deep Blue from IBM, the incumbent world champion Garri Kasparov. Deep Blue, however, was a hybrid system with computational accelerators. It was more about brute force. The current AI is no longer so successful and it is more successful.
Chinook is a program for English ladies, whose creators in July 2007 said they can not lose. He had been defeating human opponents for several years before. This result was achieved by a combination of coarse force in mid-game position searches and a good start and ending database.
Computer programs playing go have often not done so well. This is probably because the goban (go-go) is quite extensive, and with each stone next step, the complexity of decision-making increases, but people have a chance to handle it because of its inborn shape recognition capability. However, the best programs using both brute force (more precisely tree search) and intuition, are able to defeat (2016) even masters.
ALPHA Air Combat Artificial Intelligence can lead air combat better than human pilots.
Certain tasks for intelligence tests are that AI can handle better than most people.
AI is also able to handle the mirror test of self-realization.
AI is able to determine the risk of heart failure better than the doctor.
AI makes it easy to mimic human voices.
The problem is that AI behaves like a black box. Man has to blindly believe the results that can be better (smarter) than his, because they do not understand them. It is called after an explanatory AI (XAI).
AI can eliminate human cognitive distortion. It may, however, introduce its own distortion. Both human and artificial thinking can be deceived.
Artificial Intelligence in Culture
Reasonable machines are a grateful subject for science fiction writers. Isaac Asimov devoted much of his narrative work to the robotic intelligence topics, his short story I, Robot, as well as the story of The Twelve Man, was filmed.
The Polish author, Stanisław Lem, dealt with the philosophical aspects of the intelligence of the inhumanity in his books Cyberdia and Solaris (which was once again filmed, even twice). Some aspects of machine intelligence were also discussed in Golem XIV.
Indeed, much of the publications of the current style of cyberpunk scifi are inherently related to the penetration of human and machine properties, so as to cope with the idea of an intelligent machine. As an example, let's mention Neuromancer William Gibson.
The mid-audience movie audience at the beginning of the century most affected the trilogy Matrix, which tells of a world dominated by artificial intelligence originally created by man. Among the influential older works are Terminator or Blade Runner.