In collaboration with Aron Rossman-Kiss
Based on a huge audio archive of testimonies from participants and statistics of victims from the Yugoslav Wars, the work questions the modalities of presenting this material as well as the power structures that condition their use and dissemination. Ironic and grave, it is both a memorial in the making and a reminder that describing events simply as “war” risks glossing over the imbalances, intricacies and causes of the conflicts. In conjunction with the uncanny ability of algorithmically-generated “fake” text to function as testimonies, the work serves to highlight the complexities inherent in the use of such material within the public sphere, education and documentary research.
Papers: A deep learning algorithm was trained on interviews with participants of the Wars. New texts are generated based on what it has learnt, auto-translated for other languages and printed at regular intervals during the exhibition. Visitors are encouraged to take home a printout scattered on the floor. Deep learning relies on huge quantities of data for training; when more transcripts are made available to the algorithm, the quality of the generated texts improves.
Walls: a grouping of statistics, analyses and reports pertaining to the Wars. Although by no means exhaustive, they are an attempt to provide the context to the interviews and to draw a more balanced picture of the conflicts.
Sound: Many of the veterans’ interviews were tagged with keywords indicating their content. For non-Serbo-Croatian speakers, they provide the only way of accessing the material. The tags have been combined, rearranged and supplemented by additional words, becoming what appears to be a poem.