Video Abstract
In describing the essence of the process of generating images using artificial intelligence, we will not focus on the technicalities, which are constantly evolving, but on the revolutionary characteristics of AI image generation in relation to other image generation processes developed over millennia of human history. Indeed, the new process has peculiar properties that propose it as revolutionary both in its relationship with the author, the nature of the work generated, and its positioning in society. Let us analyze the main features of this innovative technology:
1 – Text-image correlation
The image has a strong correlation with a text, both in the training phase of the model, through the association of text labels (TAGs) to images, and in user activity, which generates it mainly through texts called prompts.
2 – Boundless availability of images from a universal archive
The image maker has global imagery as his palette, digitized, labeled, and used to train the model, so the image maker with artificial intelligence appears to be a modern-day Demiurge capable of manipulating a boundless iconographic universe containing images created by countless authors. As for videos, the huge archives of film, television and digital videos produced in recent decades are also made available.
3 – Accessibility of the production process
In contrast to previous civilizations in which image makers and artists were a tiny part of the population, educated in professional schools, today everyone can easily create complex images without limits. This extreme accessibility to image production has been experienced recently, thanks to the ease with which in recent decades anyone with a smartphone has the ability to produce visual content themselves, taking photographs and “shooting” videos to be widely shared via social networks, activities that have given rise to the new profession of the “creator.” But while still linked to the UGC (user-generated content) trend, AI image generation actually opens up much more revolutionary scenarios in the history of creative disciplines.
4 – Conceptual character of creative activity
Unlike the mass production of photographs and videos, which is focused on filming real places, situations, and characters, AI image generation does not directly reproduce the visible world, but generates synthetic content from an idea. Like abstract art, it “makes the invisible visible,” visualizing the author’s thoughts so to speak. And this is done not by photographing or drawing, but by generating from a pure textual description . Images take shape directly from words. So the author of a book could generate the illustrations simply by describing them with a prompt. This new possibility of generating images therefore introduces authors to the visual arts sector: even those incapable of drawing can use their imagination, translate it into words, and then into images. The writer potentially becomes the most influential illustrator of his works.
5 – Priority of automation in the production process
In our culture, the artistic act is deeply personal, linked to an artist’s bodily actions, which are consolidated into an individual style clearly recognizable in the finished product. In contrast, in AI image generation, most of the process works automatically, managed by algorithms that apparently do not allow the author to control it or conduct it at will. The author’s effort, at first glance, is minimal, the bodily input even nil, and the result appears impersonal, indistinguishable from similar results created by other authors. This feature would place AI image generation outside the domain of art. This requires the consolidation of a new form of authorship that can overcome the standardization imposed by the automation of the production process. An effort is needed to raise the author above automatism, to arrive at results that are distinctive, original, and relatable to the concept of a work of art.
6 – Unreal nature of the visual result
As the great success stories of AI image generation, which can be traced back to the deepfake phenomenon, show, the image it produces is ontologically unreal. Compared to photographs, which by their nature depict real worlds with real people, AI-generated images depict unreal worlds and unreal people. This nature that pertains to the unreal and that stands out in the great iconographic archives that now combine real images with artificial ones, can create confusion and for that very reason undermine this new creative methodology. But this disempowerment is effective only in those compartments where “real” images are required, such as news documentation, scientific photography, product catalogs… In other compartments, such as film fiction or art, the concept of unreality loses meaning, and the use of image-generative AI simply opens up new scenarios to the imagination.
Italiano