Paper Short Abstract:

Today, many digital platforms are powered by various forms of machine learning, colloquially referred to as AI. To understand the principles of machine learning, this paper draws on a study of Pinterest and the ways in which digital versions of museum objects and archival photos circulate and are curated on this platform. To understand how machine learning models impact how we know of online culture, this paper suggests cross-readings of technical papers and new media philosophy.

Paper Abstract:

Today, many digital platforms are powered by various forms of machine learning, colloquially referred to as AI. To understand the principles of machine learning, I draw on a study of Pinterest and the ways in which digital versions of museum objects and archival photos circulate and are curated on this platform. On Pinterest users share more than three billion images by adding them to thematic collections. As soon as an image is made available, it becomes part of the platform’s vast image collection and accessible in searches as well as for training Pinterest's machine learning models. At the same time the agency of display is dispersed.

Machine learning models are created in human-machinic constellations of engineers and statisticians, mathematical functions, software, hardware, and data training sets. Once trained, the models operate and mutate in abstract mathematical space to classify and organize cultural material as neighbouring data points in processes that are, as pointed out by Beatrice Fazi (2021), too complex, abstract and layered to be compared to human meaning making. To understand how their operations impact knowledge of online culture, I suggest cross-readings of technical papers and new media philosophy.