Features of the development of artificial intelligence at the present stage

Abstract

The features of the intelligent logical systems have been discussed. Their comparison with the activity of neural networks has been made.  The processes of forming new knowledge in the logical systems and accumulating data in neural networks by organizing a system of concepts have been explained. It has been noted that filling the base of data and rules in logical systems is a process that requires significant resources and therefore has a limited application range. On the contrary, the accumulation of information in extensive neural networks by connecting it to a huge number of agents could provide for effective machine learning and opens up new perspectives for creating artificial intelligence systems. Although teaching the formal descriptions, the usage of abstractions and quantitative calculations to neural systems can present a difficult problem. The prospects for the planetary mind formation and the further development of the planetary intellectual system are considered. The process of increasing the number of elements of neural systems was held back by the ambiguity of the solution before. But, the multiplicity of solutions did not bother the creators of giant neural networks associated with a huge audience. First of all, they were hardly aware of the importance of the solution uniqueness because resolving the problems a person always develops a lot of different scenarios. The possibility of the emergence of artificial intelligence comparable to a natural one in his capabilities has been discussed. The problems of teaching the expert systems and neural networks, how they were defined in the past and what has changed later are explained. The usage of fuzzy logic which is able to form a language of communication with neural networks, free from the need for double translation into natural languages and to simplify the direct transfer of information between a person’s brain and a network may be of a particular interest. Neural networks based on fuzzy neurons which are able to combine the capabilities of expert logical systems and neurocomputers are discussed.