In 2022, the project was reconfigured at the University of Washington through a directed research group (DRG). In this iteration, three Latin American artists addressed the challenge of collecting Spanish language datasets that could be used to train new NLP models. Early on, we recognized that many existing datasets actually reflect the voice, tone, and vocabulary of European Spanish – for example, one existing language model was trained exclusively on texts from the National Library of Spain. Thus, the DRG centered its artistic research around the potential to not only avoid Anglocentric A.I., but to also decolonize A.I. by collecting particularly Latin American textual datasets of fiction, journalism, history, and poetry. This discourse went even further to assert that there is no “standard” Latin American Spanish, and that the nuances of regional language should ideally be reflected in multiple new, hyperlocal NLP models. Ultimately, the project supported the development of works by Esteban Agosin Otero (Chile), Laura Luna Castillo (Mexico), and Nicolás Kisic Aguirre (Peru), as well as the re-staging/exhibition of work by Nicolás Varchausky (Argentina).