Search

The paper presents a thorough practical study of the generative neural networks phenomenon from the position of possible substitution of essentially creative human labor. The functionality and peculiarities of graphic image reconstruction according to the user's textual request were analyzed. From the economic, market and ideological points of view, and as a result of visual analysis of a typical generation result we found that this phenomenon appears as a «similarity machine», which has the opposite meaning in relation to the essence of human creative potential. This relation opens new dialectical avenues to understanding the manifestations of individuality in the visual arts.

In this paper, we present an original tool for continuous semantic processing of an arbitrary graphic image in the «human–machine» mode. Its essence lies in the automatic smooth rendering of the visual with a tone monoline, with guaranteed progress from random noise to full detail. We show that, having previously been used exclusively for natural science purposes [I. V. Kashin, AIP Conf. Proc. 2313, 030048 (2020)], the algorithm also claims to be authentic in the palette of generative design and generative art methods. The flexibility of its practical implementation suggests that it can be useful for stimulating the possibility of dynamic decomposition of the drawing and its subsequent creative rethinking.