From Apple to Anomaly
Artist Trevor Paglen’s new Curve commission takes as its starting point the way in which AI networks are taught how to ‘see’ and ‘perceive’ the world by taking a closer look at image datasets. Paglen has incorporated approximately 30,000 individually printed photographs, largely drawn from ImageNet, the most widely shared, publicly available dataset. This dataset is archived and pre-selected in categories by humans, and widely used for training AI networks. In some cases, the connotations of categories are uncontroversial, others, for example ‘bad person’ or ‘debtors’, are not. These categories, when used in AI, suggest a world in which machines will be able to elicit forms of judgement against humankind.