Artificial intelligence. Machine learning

摘要

Problem. In this paper the problems and risks of introducing the provisions of artificial intelligence (AI) into the civilization of humanity are considered. Also the stages of the development of artificial intelligence from the game of checkers and chess through machine learning to deep learning (from 1950 to the present) are considered. Goal. The aim of the work is to review and evaluate the features of machine learning, including deep learning, since these methods of artificial intelligence most actively develop and most fully characterize it. Methodology. Methods of machine learning with and without a teacher, problems of machine learning and a family of algorithms for solving them are considered. Results. It is shown that the current state of development of artificial intelligence in terms of the number of equivalent to neurons, which is used in this case, corresponds to the level of a mouse. Mankind has several decades left to prepare for the ubiquitous spread of robots with artificial intelligence. The difference between a regular program and machine learning is shown. The analysis of the features of machine learning under various schemes has been carried out. Examples of the learning process of the algorithm, types of machine learning, classification of tasks and algorithms are given. The distinction between the problems and the family of algorithms is shown. Comparison of different machine learning algorithms is presented. The scope of machine learning is defined. Examples of the use of Google’s cloud machine learning services are given. It is concluded that instead of creating a program manually using a special set of commands, the algorithm is prepared using a large amount of data. The examples of the use of artificial intelligence in business processes, such as manufacturing and, in particular, engineering, are provided. Originality. The dangers of introducing artificial intelligence are formulated. The areas of applicability of artificial intelligence and machine learning, health and education, preferred for relative safety reasons, are proposed. Practical value. The attention of specialists is drawn to the features of artificial intelligence, which may be important in various areas of human life and activity.Key words: machine learning; artificial intelligence; Industry 4.0; deep learning, logistics