INFLUENCE OF THE LINGUISTIC RELATIVITY ON LEARNING AND WORK OF ARTIFICIAL INTELLIGENCE

Abstract

The article analyses infl uence of linguistic relativity on existing neural networks, as well as to the fore-cast of the potential infl uence of this principle in the future on more advanced neural networks. The investigation provides an analysis of language situations when working with neural networks and a demonstration of specifi c examples of the infl uence of language on the work of AI, as well as possible situations of obtaining a positive eff ect when training neural networks considering the infl uence of linguistic relativity. Experiments conducted on neural networks demonstrating the existence of such an infl uence are described. The principles that allow to begin work on the formation of a framework for interaction with AI are formulated. The principle of engagement, which assumes that the factor of AI involvement in working with natural language not only in meanings, but also in connotations is to be taken into account. The principle of Kaja: to avoid endowing AI with consciousness and private world and not to use such metaphors.