Do neural networks dream of the Mona Lisa?

Without going into technical details, I would like to touch a little on the question of whether neural networks can achieve anything significant in art, literature, and whether this is creativity. Technical information is easy to find, and there are well-known applications, for example. Here, only an attempt to understand the essence of the phenomenon, everything that is written here is far from news, but I will try to formalize some thoughts only a little. I will use here the term neural networks in a general sense, as a synonym for AI, inextricably with machine learning and selection algorithms.



In my opinion, the question of the creation of neural networks should be considered not only in terms of computer science and art history, but also - philosophy and psychology. First you need to determine what creativity is, how something completely new is created; and in principle, all this rests on the problem of knowledge, in that part - how new knowledge, discovery, a particular symbol, image appears. In art, after all, as in fact, and in pure science, novelty is of genuine value.



Art and literature, (probably music) suggest, maybe not quite equal right now, but methods of cognition as in science. All of them constantly influence each other and are closely intertwined. In some epochs, the cognition of the world just happened by means of art or literature, and earlier - and generally in line with religious tradition. So, in 19th-century Russia, powerful literature actually replaced us with philosophical anthropology and social philosophy, indirectly, through art, reflecting on the problems of society and man. And as a structuring reference point, which put on the agenda the rather topical problems of human existence, developed later by well-known philosophical trends, it is still highly appreciated. Or at the beginning of the 20th century, the emerged artistic modernist and avant-garde movements, which cannot be considered in isolation from their ideological content, and which portended the scrapping of tradition, the emergence of a new world and a new person. After all, we cannot admit that the fundamental value of art is only aesthetic. In this case, perhaps, we would still be living only surrounded by some kind of aesthetic system of the past that has pupated in its self-completion. All the great creators, geniuses in art and literature just earned this “title” not so much because of the aesthetic value of their works, but because of the discovery of new directions by them, doing something that no one had done before and did not even assume that you can do that.



Whether the product will be considered new as a result of an unprecedented combination, some shuffling of existing, known parts - the grids can cope with this - based on a predetermined limited number of data, for example, when styling images or generating new ones. Or it will be a perfect breakthrough, a previously unknown quality, revealing something with which it is impossible to compare any of the previously observed - although, of course, any incredible, unparalleled breakthrough is nothing more than the result of a well-prepared work that is simply carried out implicitly, not all that is manifested and visible to the uninitiated and even to the creator himself - so far only a person can act, in my opinion.



Roughly speaking, the first type of cognition and creativity can be compared with a very slow, gradual development as a result of evolution, and the second with spasmodic development as a result of positive mutations. Neural networks, in my "creative" activity, in my opinion, now gravitate somewhere to the first type. Or rather, to a situation that is described as the absence in the near future of a qualitatively new development, under the conditions of a system that supposedly approached the limit of complexity at this stage, to the “end of history”, when new meanings are formed as a result of changes in combinations - or insertions in an unusual context - already existing patterns. It is similar to how new unusual patterns are created in a kaleidoscope, each time from the same set of colored glass. But, I think, not in vain because, as indicated, the structure of networks in general terms repeats the structure of the nervous system: neurons-nodes, axons-connections. Perhaps this is like the beginnings of the first cells, only now, the process of evolution will be accelerated by human hands, that is, it will become its tool, thus overcoming the slowness of nature. Including by example, based on the ideas of transhumanism.



Asking myself the question: will it be interesting for me to look at the paintings created by the grid at this stage, I can answer that here, probably, it is necessary to distinguish between something applied as design and pure art. What is good for design and will free a person from the routine, secondary processes of developing wallpaper, prints and draperies, is not suitable for art, which, generally speaking, is not always always on the cutting edge, at the peak of relevance, but must express a personality in its search. The artist, in the broad sense, living his experience and "absorbing" the spirit of the era, consciously or not, processes them into an artistic image. Thus, one can read some ideas, messages from his work, they can greatly affect feelings. A neural network also receives some kind of data set as input and transforms it, but so far it is too flat, the one-dimensional processing and the “surplus” value of the information received at the output is not great, and the result can only entertain for a while. The same applies to experiments with neural networks in journalism, making progress more where writing dry financial news is required, rather than creating software works with an author's point of view. In experiments with music, especially electronic business can be a little better. In general, he noted such a thing that sovrisk, modern literature and painting, for about a century, as if specially produces such abstract and minimalistic forms that it seems to be created to easily be processed by neural networks and passed off as human art. Can a hunch the end of an era?



They say intelligence is not equal to the whole person. Although, with a personality, the question is, of course, philosophical - after all, in the GAN network, for example, the generator creates new data from nothing, in part, just following the discriminator's verdict under the influence of the weight of decisions. One can, after all, ask the question this way: is the creator in his cognitive activity, so to speak a generator and discriminator in one person, somewhat pre-trained by the very information background that “hovers in the air” of the era and who vote for his choice internal weights, and he builds a new world, a new work from the knowledge of the known bricks (pixels) of being? In this case, are we not some kind of super complex analogue of the grid, with colossal, but still limited input data. Maybe the person is such an advanced selection algorithm, with the presence of implicitly necessary functionality indirectly affecting quality pre-training?



In any case, I will go to the first exhibition of works of art created by the so-called AI, when it acquires a personality with all its attributes, consciousness and self-consciousness. Perhaps there will even come a time when, like the character in the 14th series of the animated series Love, Death, and Robots, AI, in search of meaning, realizes that art must be inseparable from life and then the time will come for giving up the frightening, bottomless never-satisfied complexity, where in essence simplification is a metaphor for death. So far, you can often see in movies that AI gains self-awareness and, of course, gets out of control as a result of some kind of software malfunction, which the scriptwriters probably think is an analogue of some kind of randomness that triggers new positive (and for some not so) transformations, as was the case with positive mutations for the natural evolutionary path of development.