








In recent years the use of text-to-image artificial intelligence systems, such as Stable Diffusion, Midjourney and Gencraft, has exploded. These generative AI systems use prompts typed by users to generate images in seconds, and while millions have enthusiastically embraced the use of these tools, their application have also raised deep concerns around ethical use, questions of originality, the nature of creativity and the materiality and aura of artworks. It is such questions that I explore in my project titled Iconologai through an examination of the use of prompts within AI systems and the use of analogue emblem books by artists in the past.
The title of my project is a misspelling of Cesáre Ripa’s emblem book titled Iconologia which was first published in 1593. This publication offers detailed textual descriptions, but no illustrations, of personifications such as Sloth, Folly and Confidence, listed in alphabetical order, and became widely referenced by artists (Osborne 1978:554, 840). Later, numerous illustrated editions were printed as the book’s popularity and influence grew.
Ripa’s handbook formed part of a tradition of emblem literature, which Osborne (1978:369) describes as “books of symbolic pictures accompanied by explanatory texts”. These books resulted from attempts by Italian humanists to modernise Egyptian hieroglyphs following the discovery of the Hieroglyphica of Horapollo in 1419. The first emblem book, Andrea Alciati’s Emblemata (1531), influenced the production of large numbers of emblem books during the 16th and 17th centuries. Such books drew inspiration from a variety of sources, including the Bible, medieval lapidaries and bestiaries, fables, classical myth and legend (Osborne 1978:369-370). According to Osborne (1978:370) emblem books formed part of a “vast system of codification” from which Ripa’s Iconologia emerged.
The dissemination of emblems, their descriptions and illustrations through emblem books, and their repeated use by artists across time, contributed to the vast amount of visual material on which generative AI is trained today. Just as artist drew inspiration from emblem books, AI draws from the mass of visual material created by artists in the past and present. In a way, the “prompt” written for AI acts like the explanatory texts found in emblem books and it is this connection that I explore in Iconologai.
The starting point for the work was self-portraits which I generated in 2021 with, what now appears to be a quite rudimentary AI image generator, PortraitAI, which generates portraits in a variety of painterly styles from a selfie. I linked seven of the portraits to seven personifications listed in Ripa’s Iconologia and then used the descriptors of the personifications in the 1709 illustrated English edition as prompts to generate images. I used a variety of AI systems, including Midjourney, to see how different systems interpret the same prompt. I used the seven PortraitAI images as the basis from which to create what I term “augmented painting”, the combination of AI generated imagery with traditional oil painting techniques. The results of my exploration are presented in the form of this artists’ book consisting of the seven paintings and booklet contained in wooden boxes hinged together.
Deirdre Pretorius April 2024
Sources
Osborne, H (ed). 1978. The Oxford Companion to Art. Oxford: Clarendon Press.
Ripa, C. 1709. Iconologia. London: Benj. Motte.
Ripa, C. 1976. Iconologia. Garland edition reprinted from 1611 Padua edition. New York: Garland Publishing.