Trevor Paglen: From Apple to Anomaly Catalogue
Move your mouse over image or click to enlarge

Trevor Paglen: From Apple to Anomaly Catalogue


View Wishlist

For his new Curve commission, American artist Trevor Paglen invites us to take a critical look at how artificial intelligence networks are taught to ‘perceive’ and ‘see’ the world by engineers who provide them with vast training sets of images and words.

Paglen has drawn on ImageNet, a research project based at Stanford University that has created a dataset of over 14 million images. These images are subsequently used to train digital networks for a wide range of applications including surveillance systems and driverless cars.

To enable these networks to understand what these images mean, the entire ImageNet dataset has been categorised by over 25,000 human workers from the crowdsourcing platform Amazon Mechanical Turk. Images are tagged with over 21,000 terms taken from WordNet, a database or ‘map’ of English nouns, verbs, adverbs and adjectives. When given to an AI (Artificial Intelligence), this combination of image and text forms the basis by which the network will interpret and judge any new visual information it sees.

Following on from previous projects such as Machine Readable Hito and A Study of Invisible Images (both 2017), From ‘Apple’ to ‘Anomaly’ continues Paglen’s preoccupation with the social and political implications of the often-concealed processes behind digital technologies.

For his commission Paglen has created a vast mosaic of approximately 30,000 of these images across 100 categories that charts ImageNet’s labelling of the world, moving from uncontroversial nouns such as ‘cloud’, ‘anchovy’ and ‘apple’, to terms that make problematic and prejudiced visual judgements such as ‘schemer’, ‘traitor’ and ‘anomaly’.

Presented as individually printed and pinned photographs, this resolutely analogue glimpse into a digital mind is a stark reminder that the forces behind these artificial intelligence networks are rife with the hidden politics, biases, stereotypes and philosophical assumptions of their human programmers. While their application for tasks such as surveillance presumes objectivity, these networks simply reinforce and further problematise our existing ways of seeing.

Featuring a commissioned text by the academic, writer and curator Sarah Cook, and an interview with the artist by curator Alona Pardo.


Dimensions: 28 x 24cm