Historical continuum and normatization in art collections and datasets: experiments with artificial intelligence at paulista museum

Abstract

This article presents a set of experiments in the field of Art History with Artificial Intelligence technologies (computer vision) carried out within the scope of demonumenta, a research and university extension project that seeks to critically and creatively tension public memory policies. From the understanding that datasets and artistic collections are analogous practices, we questioned how to work in this intersection to subvert the normative assumptions that characterize the organization of art collections, databases and the discourses that their tools enunciate. We believe that part of this answer lies in the critical activation of public holdings to reshape the way we train machines today. Therefore, we created a dataset, based on the works of the Museu Paulista of the University of São Paulo (USP) available in its GLAM (Galleries, Libraries, Archives & Museums) in the Wiki projects. The systematization of this dataset was the basis for carrying out five analytical experiments with Artificial Intelligence algorithms. They are: Numerical Natures, Possible Landscapes, Color Archeology, Affirmative Album and Animated Ignorance. Such experiments evidence the colonialist continuum that elaborates the historical narrative based on normative visual parameters and patterns.