Artificial Intelligence

Artificial intelligence (AI) is a research and applications discipline that includes the areas of robotics, perception systems, expert systems, natural language processing, fuzzy logic, neural networks, and genetic algorithms. Another area, artificial life, is the study of computer instructions that act like living organisms.

Behind all aspects of AI are ethical questions.

Artificial intelligence (AI) is a group of related technologies that attempt to develop machines to emulate human-like qualities, such as learning, rea­soning, communicating, seeing, and hearing. Today the main areas of AI are:

· Robotics

· Perception systems

· Expert systems

· Natural language processing

· Fuzzy logic

· Neural networks

· Genetic algorithms

Robotics

Robotics is a field that attempts to develop machines that can perform work normally done by people. The machines themselves, of course, are called robots. A robot is an automatic device that performs functions ordinarily ascribed to human beings or that operates with what appears to be almost human intelligence.

Expert Systems

An expert system is an interactive computer program that helps users solve problems that would otherwise require the assistance of a human expert.

The programs simulate the reasoning process of experts in certain well-defined areas. That is, professionals called knowledge engineers interview the expert or experts and determine the rules and knowledge that must go into the system. Programs incorporate not only surface knowledge ("textbook knowledge") but also deep knowledge ("tricks of the trade"). An expert system consists of three components:

· Knowledge base: A knowledge base is an expert system's database of knowledge about a particular subject. This includes relevant facts, infor­mation, beliefs, assumptions, and procedures for solving problems. The basic unit of knowledge is expressed as an IF-THEN-ELSE rule ("IF this happens, THEN do this, ELSE do that"). Programs can have as many as 10,000 rules. A system called ExperTAX, for example, which helps accoun­tants figure out a client's tax options, consists of over 2000 rules.

· Inference engine: The inference engine is the software that controls the search of the expert system's knowledge base and produces conclusions. It takes the problem posed by the user of the system and fits it into the rules in the knowledge base. It then derives a conclusion from the facts and rules contained in the knowledge base.

Reasoning may be by a forward chain or backward chain. In the forward chain of reasoning, the inference engine begins with a statement of the problem from the user. It then proceeds to apply any rule that fits the problem. In the backward chain of reasoning, the system works backward from a question to produce an answer.

· User interface: The user interface is the display screen that the user deals with. It gives the user the ability to ask questions and get answers. It also explains the reasoning behind the answer.

Natural Language Processing

Natural languages are ordinary human languages, such as English. (A second definition is that they are programming languages, called fifth-generation lan­guages, that give people a more natural connection with comput­ers. Natural language processing is the study of ways for computers to rec­ognize and understand human language, whether in spoken or written form. The problem with human language is that it is often ambiguous and often interpreted differently by different listeners.

Still, some natural-language systems are already in use. Intellect is a prod­uct that uses a limited English vocabulary to help users orally query data­bases. LUNAR, developed to help analyze moon rocks, answers questions about geology from an extensive database. Verbex, used by the U.S. Postal Service, lets mail sorters read aloud an incomplete address and then replies with the correct zip code.

In the future, natural-language comprehension may be applied to incom­ing e-mail messages, so that such messages can be filed automatically. How­ever, this would require that the program understand the text rather than just look for certain words.

Fuzzy Logic

A relatively new concept being used in the development of natural languages is fuzzy logic. The traditional logic behind computers is based on either/or, yes/no, true/false reasoning. Such computers make "crisp" distinctions, lead­ing to precise decision making. Fuzzy logic is a method of dealing with imprecise data and uncertainty, with problems that have many answers rather than one. Unlike classical logic, fuzzy logic is more like human rea­soning: it deals with probability and credibility. That is, instead of being sim­ply true or false, a proposition is mostly true or mostly false, or more true or more false.

Neural Networks

Fuzzy logic principles are being applied in another area of AI, neural net­works. Neural networks use physical electronic devices or software to mimic the neurological structure of the human brain. Because they are structured to mimic the rudimentary circuitry of the cells in the human brain, they learn from example and don't require detailed instructions.

To understand how neural networks operate, let us compare them to the operation of the human brain.

· The human neural network: The word neural comes from neurons, or nerve cells. The neurons are connected by a three-dimensional lattice called axons. Electrical connections between neurons are activated by synapses.

· The computer neural network: In a hardware neural network, the nerve cell is replaced by a transistor, which acts as a switch. Wires connect the cells (transistors) with each other. The synapse is replaced by an electronic component called a resistor, which determines whether a cell should acti­vate the electricity to other cells. A software neural network emulates a hardware neural network, although it doesn't work as fast.

Computer-based neural networks use special AI software and compli­cated fuzzy-logic processor chips to take inputs and convert them to out­puts with a kind of logic similar to human logic.

Neural networks are already being used in medicine and in banking business.

Genetic Algorithms

A genetic algorithm is a program that uses Darwinian principles of random mutation to improve itself. The algorithms are lines of computer code that act like living organisms. A hybrid expert system-genetic algorithm called Engeneous was used to boost perfor­mance in the Boeing 777 jet engine, it involved billions of calculations.

Computer scientists still don't know what kinds of problems genetic algo­rithms work best on. Still, as one article pointed out, "genetic algorithms have going for them something that no other computer technique does: they have been field-tested, by nature, for 3.5 billion years.

Genetic algorithms would seem to lead us away from mechanistic ideas of artificial intelligence and into more fundamental questions: "What is life, and how can we replicate it out of silicon chips, networks, and software? " We are dealing now not with artificial intelligence but with artificial life. Artificial life, or A-life, is a field of study concerned with "creatures"—com­puter instructions, or pure information—that are created, replicate, evolve, and die as if they were living organisms.

Behind everything to do with artificial intelligence and artificial life—just as it underlies everything we do—is the whole matter of ethics. Many users are not aware that computer software, such as expert systems, is often subtly shaped by the ethical judgments and assumptions of the people who create it.

We must take into consideration that human create such technologies, use it, and have to live with the results.

Дата добавления: 2015-11-14