PT

Decoded records: the official past


Old images from the official archive of the Fundação Bienal are here interpreted by a set of artificial intelligences: Amazon Rekognition, Darknet YOLO, Facebook Detectron, Google Cloud Vision, Microsoft Azure, IBM Watson and APIs of the services Deep AI and Clarify. The latter is officially used by the United States Department of Defense. The aim is to analyze how the understanding of the past is transformed when read by contemporary tools for the reading of images.

Like what happens in the action Decoded Records: The Amateur Present, some of the results reveal problematic interpretations – images with women tend to increase the “raciness index” of the AIs, and many of the elements of the Biennial's exhibition space are read as products, for example.

Here, as these are records of the past, there are also temporal transformations – like the identification of cell phones and computers in periods when these devices did not exist. Not even the past can resist the current technological consumption.

 


Records Methodology


The results of the images of the Biennial (of this one and of the past editions) presented here were arrived at by a “methodological” mode with technological/scientific bases, always with the aim of highlighting the unexpected results. We are interested in the responses of the artificial intelligences that expand the meanings of the images and/or reveal the categories of the structures of these platforms.

If you are interested in consulting the complete results, without our filter, visit the Biennial Art Decoder, an internal platform constructed exclusively for the project and which presents all the results of the images produced by the different artificial intelligences.

In the "Content" category there is a grading of the results in the descriptors “adult,” “physician,” “racy,” “falsification” and “violence” in the following order: very likely > likely > possible > unlikely > very unlikely. For its part, the term “Safe for Work” is an Internet expression used for contents appropriate for viewing in public places or at work.