You are here

Is an AI-generated Icon an Icon? Theological, Ecological, and Communal Considerations

Meeting Preference

In-Person November Meeting

Only Submit to my Preferred Meeting

The ability of generative AI programs such as DALL-E, Midjourney, and Gemini Ultra to produce creative and stunning images has raised controversy in the art world (Coeckelburgh, Benedikt).  Much of the criticism questions the ability of a computer to be truly creative, since generative AI programs base their output on training sets comprised of the work of previous artists.  Eastern Orthodox iconographers, however, are expected, indeed required, to base their renderings on previous icons.  The formulaic nature of icon painting would seem to make it ideally suited to generative AI.  Nor would the virtual world seem inimical to the veneration of icons.  Icons are meant to be immersive, to draw the viewer into their world, one in which our usual perceptions of space and time are suspended.  The icon, like virtual worlds, “imposes its own principles on the spectator” (Evdokimov). 

Yet most Orthodox theologians remain suspicious of any relationship between icons and digital technology, whether that be finding existent icons and printing them (Evdokimov, Justinian), venerating them on-line (Suslov, Ricardi-Swarz), or most recently, generating them through AI (Sheen, Albia et al.).  Theologically, this skepticism marks a return of the iconoclastic controversy of the eighth century.  Orthodox doctrine based legitimization of icons on Jesus’ incarnation in material form.  The immateriality underlying both digital reproductions and AI generations raises fears of a new Docetism that separates spirit and matter (Justinian, Suslov).  Ephesion1:10 promises that in " the fullness of time, [God will] unite all things in heaven and things on earth.” The icon is a graphic expression of this union, “firstly because it depicts God become flesh (Christ) and flesh become god (saints), and secondly, because the icon is itself a material bearer of uncreated grace” (Aiden). 

An icon made from egg, wood, and mineral paints, represents the eschatological redeemability of the animal, vegetable and material worlds and their union with the spiritual.  An icon essentially of numbers and probabilities diminishes or disregard the material realm.  Indeed, rather than honoring the natural world, an icon generated by an AI program does hidden damage to that world, since generative AI programs require extensive amounts of fossil fuel and water to power both their training and their usage. 

A third is rooted in the role of both icon and iconographer within the temporal community and the larger community of saints.  The iconographer joins the temporal and divine as they approach their subject through prayer and contemplation, establishing a triune community between iconographer, saint, and viewer facilitated by the icon itself.  Clearly an AI, sui generis, cannot fully take on this role of iconographer as it cannot engage in prayer nor participate in this relational triad.  Considering the AI as merely a tool for the human iconographer still raises issues, as the program must be prompted with words and words are notoriously poor at expressing holy visions, requiring an emphasis on the superficial characteristics of the icon rather than the spiritual communion with the saint.

While the AI might sever the relationship between the iconographer and the saint, that between the venerator and the saint could remain. Digital dissemination and printed reproductions have made icons more widely available, and it cannot be denied that one can have a spiritual experience viewing an icon on paper or screen.  AI-generated icons thus could have the potential to educate people and lead them to meaningful experiences of the divine.  But they could just as easily mislead.  AI programs are loose cannons, embedded in neither the community of the faithful nor the larger community of saints.  Anyone with access to these programs can generate their own icons, as, ultimately, these programs could generate icons on their own.  While a digitized print of an existent icon might be lacking materiality, it still represents the work of a human iconographer.  An icon produced by a machine runs the risk of misrepresenting the faith.  Without the blessing of either iconographer or icon, such works are merely religious pictures.  Should they become common, they risk denigrating the very meaning of icon and serving as a wall rather than a window between the faithful and the community of the saints. 

This presentation will be illustrated with examples of both human produced and AI produced icons.

Abstract for Online Program Book (maximum 150 words)

Images created by generative AI have raised controversy in the art world, where critics question the ability of AI to be creative since generative programs base their output on that of previous artists.  The formulaic nature of icon painting would seem to make it ideally suited to generative AI.  Nor would the virtual world preclude the veneration of icons.  Yet most Orthodox theologians remain suspicious of digital technology.  Theologically, AI generation raises fears of a new Docetism that separates spirit and matter.  A second concern is rooted in the role of both icon and iconographer within the temporal community and the community of saints.  Through prayer and contemplation, an icon establishes a relational triad linking iconographer, saint, and viewer.  AI severs the relationship between iconographer and saint and has the potential to mislead viewers, since it is embedded in neither the community of the faithful nor of the saints. 

Authors