Although they belong to different generations, resonances can be observed in their professional practices, particularly in how both artists begin with a concrete, existing context as the foundation of their methods. For Tomo Savić-Gecan, this context is the exhibition space itself, including its walls, floor, display cases, lighting, temperature, humidity, plans, or website, as well as, in a broader sense, all components encompassed by the term “art system.”
For Goran Petercol, the given context of a location implies a highly ephemeral aspect, which he, through exploring new possibilities (and constraints) of its representation, subordinates to his own logic of structuring. “The forms that arise”, he often reiterates, “are the result of a process.” This process unfolds in a persistent tension between embracing chance and establishing order, only to deliberately evade conventions.
Tomo Savić-Gecan, with his arsenal of absurd and logically disjointed actions, has long provoked not only the art system but also questions about the flexibility of art’s boundaries and definitions. His works, consistently titled Untitled, with only the year of creation and minimal details (often imperceptible to viewers), are left deliberately open to others’ interpretations. This extends to the point where the curators’ interpretations themselves become the work — one that is, paradoxically, absent from the exhibition.[1]
For instance, just as he may have toyed with the ““white cube”” aura by laying parquet flooring in a gallery, he similarly introduces irony in this project for the Diocletian’s substructures.[2] Here, the curatorial text itself is submitted to artificial intelligence analysis, transforming it into a generator of the presence or absence of Petercol’s works, and the combinations in which these appear. This irony extends to the art world, where curators, as Renata Salecl notes, have become indispensable mediators between artworks and viewers—figures whose interpretations visitors often passively accept, particularly when unsure of what should be percieved as art. Moreover, they not only entrust the curators with interpretation and judgement on artwork, but also generously leave them the task of mediating the enjoyment of the artwork itself.[3]
In contrast, Goran Petercol willingly describes and writes about his artistic processes. Yet,even with his precise explanations, he consistently avoids adding extraneous layers of meaning.[4]
For Savić-Gecan, process is also significant, but it unfolds in unpredictable sequences of singular projects, some of which bear similarities—specific gestures recurring in variations.[5]
For Petercol, the process is the core of the work, whether part of a cycle or a single piece. His artistic decisions on the path to completion, however, remain as unpredictable and arbitrary as Savić-Gecan’s. In the initial description of his project for Diocletian’s Cellars, Petercol explains the framework of his processes:
“I turned the temporal succession of thematically unrelated works over time into the structure for the exhibited pieces. Within five spaces, I created six consecutive works with light. In each, a light form is projected, to which I add a bulb as a new light source. The purpose of the lit bulb is to double the first light source, introducing it into the work as something past and peripheral.”
From this description and those of individual installations, we see that Petercol creates new connections between the space and its existing lighting with each work. At times, he carries over experiences from one work to the next. He may repeat the same dimensions or shape, achieved with a simple tool — a profile spotlight with four shutters. Occasionally, he transfers prior experiences of projecting light forms into spaces using the preexisting architectural coordinates and edges.
The locations Petercol chooses for his works are, as in much of his practice, ephemeral interstitial spaces. For this project, they were selected for their proximity to existing lighting, which he incorporates into the work. This creates an ambivalent situation where we are unsure which light sources contribute to the form. In addition to the existing lights, Petercol adds another source to amplify the intensity of the light form while simultaneously, as he explains, marking the presence of the original light source.
What happens, however, when artificial intelligence, analyzing this text, triggers the deactivationof Petercol’s lights in certain spaces, returning them to their original state? Does the empty exhibition space remain part of the artwork? Whose artwork? Could this situation be likened to the erasure of another artist’s work, as Robert Rauschenberg did in 1953 when he approached Willem de Kooning, an artist he deeply respected, and requested a drawing he intended to erase?[6] The disruption of de Kooning’s work was not an act of disrespect toward the artist or an attack on his artistic vision but rather an experiment in creating through erasure. However, in Diocletian’s Cellars, nothing has been erased—only temporarily suspended. The exhibition spaces where the light installation is momentarily absent paradoxically become spaces of double presence, as Goran Petercol writes:
“When my projection is switched off, then, in that emptiness of the exhibition space, we are both present—I, who have left, and Tomo, who has entered. And then, conversely, I return, and he leaves. (…) By turning off my work’s projection, the space of the work remains within the context of art. It confirms that it is not nothing; on the contrary, it materializes as the interspace of the entirety of my sequence, additionally, as if illuminated by Tomo’s work.”
Radmila Iva Janković
[1] The curators communicate with the audience about Tomo Savić-Gecan’s nonexistent work. Untitled, 1998, 2020, 2022 (HDLU Zagreb; MSU Zagreb; SKD Prosvjeta, Zagreb).
[2] The AI analysis of the accompanying curatorial text determines the presence and combination of Goran Petercol’s works within the framework of Tomo Savić- Gecan’s project, Untitled, 2024, in the substructures of Diocletian’s Palace, Split.
[3] Renata Salecl, Against Indifference, arkzin d.o.o. & Što,kako & za koga, Zagreb, 2002., str. 182–183.
[4] Meticulous descriptions of all site-specific light installations in this project, accompanied by photographs and sketches of future works taken on-site, were sent to everyone involved in the project.
[5]
[6] Jessica Davido, Erasure as an Act of Creation: Rauschenberg’s Erased de Kooning & the Presence of the Past, Academia.edu (accessedNovember 20, 2024)
I accepted Tomo Savić-Gecan’s proposal to use my works during the exhibition to perform his work because it allowed me to approach reality differently with my array. The fact is that my work will disappear after the exhibition ends. The duration of the exhibition was not determined by me but assigned to me. I accepted this convention. It is an external factor that determines the duration of my work. This demand for unconditional execution is essentially no different from Tomo’s termination of my work’s duration. Tomo’s work, too, is an external factor that unconditionally demands execution. His work was necessary for me to double the convention of the duration of the work within the exhibition.
The selection of light installations to be activated and the moments in which they are activated are determined through a series of queries to a large language model and the analysis and interpretation of its responses.
In the first step, we request the large language model to break down each of the existing texts about Petercol’s exhibitions into six meaningful paragraphs. Each of these six paragraphs is then converted into a numerical representation in the form of a vector of real numbers (known as an “embedding vector“). The values of each vector represent the semantic content of the paragraph in numerical form.
The main text we use for comparison is the text by Radmila Iva Janković, specifically about this exhibition. This main text is also represented as a numerical vector, and we calculate the similarities between the vector representations of the paragraphs and the vector representation of the main text. The semantic relationship between these vector representations is calculated using the cosine similarity measure. The result of this step is six values between 0 and 1, which are used to rank the light installations. Depending on their ranking, a light installation will either be activated or not.
The goal is to have between three and six light installations activated at any given time. To determine the number of installations to be activated, we generate text summaries using the large language model. Summaries of 3, 4, 5, and 6 sentences in length are created for both the existing texts about previous exhibitions and the main text.
For each pair of summaries (one from the existing text and one from the main text, with lengths of N sentences where N equals 3, 4, 5, or 6), we calculate their relationship using cosine similarity. The result of this step is a set of similarity metrics for cases where the summaries are 3, 4, 5, or 6 sentences long. The length of the summaries (3, 4, 5, or 6 sentences) with the highest cosine similarity determines how many light installations will be activated.
With the ranking of light installations from the first step and the desired number of activated installations from the second step, the final selection of activated lights is made by choosing the N installations with the highest values from the first step, where N is the desired number determined in the second step.
In the final step, we use the existing texts to determine the day and time when the activated light installations will change. The existing texts used in the system for determining which installations will be activated are arranged in a specific sequence. To determine the day and time of the change, we extract summaries written in 6 sentences for two consecutive texts in the sequence. We then calculate the relationship between the consecutive texts using the cosine similarity of their vector representations, and based on the obtained values, we select the moment of change, which can occur after 1, 3, or 5 days, within the gallery’s working hours.