by Diego Maria Cappiello
Sometimes conflictual and sometimes collaborative, the relationship between art and technological progress is probably as old as civilization itself. In recent decades, there has been no field in the visual arts that has not had to contend with the electronic revolution in all its forms: from the home personal computer to corporate digitization, from digital photography to vector design. In very recent times, generative artificial intelligences (AI) have gained center stage. They have actually existed and operated for about two decades, and their first theorization goes back even further 〈1〉.
However, their recent and sudden improvement (occurring mainly in the last two years), has drawn the attention of the international public, specialist and non-specialist alike, to the opportunities and dangers of this technological revolution. A revolution that is having a very considerable impact on markets, means of production and culture, and on art itself, conceived as a product of specifically human skills.
In the wide range of AI proposals designed with a creative function in mind, it’s not easy to determine the actual adherence of the medium to the decorator’s craft. The ability to generate a good concept in seconds and with a small number of inputs can certainly raise questions and fears in the field of graphic design and illustration. But when it comes to decorating, the contextual functionality and the practical and productive translatability of an idea are indisputable facts, so that, at least for now, it is unimaginable that AI can go beyond the rough outline of the starting intuition. At first, for those who, like me, have always used the traditional pencil as a drafting medium, integrating AI into design was not easy, despite my growing confidence with digital languages. However, after going through many trials and errors, I now cannot give up a medium that is as useful as it is at times – I must admit – inconclusive.
Without going in too much technical detail, it can be said that an AI generates images using advanced learning models trained on a very large visual database to create new images – similar to those already in place but original – based on complex patterns and logical relationships. These models are able to generate new visual representations, from random input as well as from targeted commands. Of course, the “intelligence” of AIs is largely determined by the databases on which they are trained. This means, given the size of current generative AIs, that most of them are trained on images belonging to every branch of art (historical period, style, artist…), all of which can be freely drawn upon.
Let’s start with the advantages. First among them is the efficiency and speed given by the computational power, which allow the AI to produce a variety of design options, and of all kinds, in a very short time. This means that as an artist, I can focus more on refining ideas and compositional choices, saving valuable time and resources. Moreover, AI is not a merely passive tool, but actively contributes to the process, like a valuable creative assistant. There is also the observation that even the most precise AI tends spontaneously to generate additional elements not anticipated by the user, thus stimulating inspiration and, thus, experimentation with new possibilities. Finally, through its ability to explore vast creative spaces, AI encourages the designer to break out of conventional patterns. This exploratory freedom can lead to unexpected and unique solutions.
Let’s come to the disadvantages. First among them, and extremely practical in nature, is the inability of AI to correctly interpret the function of the element being designed. See, for example, the chess images (made with the Midjourney platform) illustrating this article: while interesting in form, the pieces are impractical on the chessboard because of the double coloring. Indeed, the AI used had no small amount of trouble separating blacks from whites, appearing profoundly bewildered when faced with such a trivial alternative. In proper aesthetic terms, however, there is a strong risk of running into trivial and repetitive solutions. In other words, there is a real possibility that certain dominant styles or trends, will be perpetuated excessively, causing a loss of originality and a surfeit of projects similar one to another. Human creativity, capable as such of breaking the mold and embracing the unexpected, risks being stifled by excessive and homogenized production. AI can thus run into problems of overfitting, that is, the tendency to generate images based solely on the source data on which the AI has been trained, without introducing new and unexpected elements. This limitation can fatally lead the artist to design in an excessively “safe” and uninnovative way.
At present, and specifically in decoration – a field of the arts characterized by definition by the attention to detail applied in space – the advantages that AIs can grant to the artist’s productivity are potentially enormous: thanks to them, a ubiquity and complexity of definition, previously unthinkable in such a short time, becomes attainable. However, the partial reading of elements, the risk of repetitiveness and the general tendency to visual overload, make AI a double-edged sword, no more or less than any other tool that the professional must know how to use wisely and consciously.
In the current state of knowledge, I would like to state the following. It is true that most of the AIs on the market have neither a database, nor the corresponding training, designed with strictly decorative composition in mind. But the results to date portend largely positive outcomes within a few years or even months. Perhaps more than any other art form, decoration relies on archives and formulas, making it an ideal interlocutor for hypothetical dedicated AIs. However, even if AIs become the new design standard, given their nature it would be difficult to imagine them competitive without a skilled designer, who knows how to bend them to a precise and consistent intent.
〈1〉 Bibliographical directions: 1) J. Shane, You Look Like a Thing and I Love You: How Artificial Intelligence Works and Why It's Making the World a Weirder, Voracious Books, Boston 2019; 2) M. du Sautoy, The Creativity Code: Art and Innovation in the Age of AI, Harvard University Press, Cambridge (Massachusetts) 2019 ; 3) O. Theobald, Generative AI Art: A Beginner's Guide to 10x Your Output with Smart Text Prompts, Amazon 2023. The first is a brilliantly popular text introducing AI to a wide audience. The second, wide-ranging and more analytical than practical in nature, deals with the medium itself: it is a mathematician's view of the ability of AIs to assist in creative fields. The third is a technical but fluent text, ideal for learning the basics of generative art: it compares the main models of AIs and also the cross-method (text generators used to generate images, as proposed in this article). Homepage: Pompeo Borra, Knights (part.), oil on canvas, 1948, cm. 81 x 96, Modena, Museo Civico (photo © Paolo Pugnaghi/Museo Civico di Modena). Below: renderings of chess redrawn from the concept generated with AI.