{"title":"This Person Does Exist","authors":"M. Schäfer","doi":"10.46467/TDD37.2021.214-225","DOIUrl":null,"url":null,"abstract":"This Person Does Exist is an artistic approach to exploring a large dataset of photographic portraits in a randomised manner. The dataset was originally created by Nvidia Research Lab, which has scraped and analysed creative commons images from the popular image hosting platform Flickr. These pictures were then used to train a machine learning model which can create new stochastic images of faces. \nIn contrast to a popular website that showcases the computer generated images, I am displaying random faces from the dataset with their corresponding metadata. \nThis essay looks into extractivist mechanisms in current machine learning techniques, using the internet to populate and refine databases, while focusing on artistic approaches that expose them. I make the case for Dataset Art as an emerging field which reframes scientific corpora by placing them into galleries and exhibiting them as found objects online. Finally, I argue that this artistic practice is a legitimate way of opening up a larger public discourse, although artists working with human data must be aware of ethical issues and responsibilities regarding privacy and consent.","PeriodicalId":34368,"journal":{"name":"Temes de Disseny","volume":"43 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2021-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Temes de Disseny","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.46467/TDD37.2021.214-225","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"Arts and Humanities","Score":null,"Total":0}
引用次数: 1
Abstract
This Person Does Exist is an artistic approach to exploring a large dataset of photographic portraits in a randomised manner. The dataset was originally created by Nvidia Research Lab, which has scraped and analysed creative commons images from the popular image hosting platform Flickr. These pictures were then used to train a machine learning model which can create new stochastic images of faces.
In contrast to a popular website that showcases the computer generated images, I am displaying random faces from the dataset with their corresponding metadata.
This essay looks into extractivist mechanisms in current machine learning techniques, using the internet to populate and refine databases, while focusing on artistic approaches that expose them. I make the case for Dataset Art as an emerging field which reframes scientific corpora by placing them into galleries and exhibiting them as found objects online. Finally, I argue that this artistic practice is a legitimate way of opening up a larger public discourse, although artists working with human data must be aware of ethical issues and responsibilities regarding privacy and consent.