Objects in Limbo
A series of explorations that closely investigate and dramatize the quantitative behaviors of machine learning
Year: 2019 | Thesis Project | Keywords: Machine Learning, Performative Adversarial Attack, Machine Vision
Machine learning is taking computer intelligence to a whole new level. One of the tasks machine learning claims to excel at is image classification. Nevertheless, cases where machine vision fails are not uncommon. On one hand, moments of misclassification by machine learning indicated computer inability of quantitative nature. On the other hand, they point out human stereotype induced by a qualitative mindset. Objects in Limbo is a series of daily objects that are manually deformed in an incremental and performative manner until they stop being correctly recognized by an algorithm. The human-machine dialogue throughout the process generates a poetic layer of gestural meaning around the epistemology of what we see and what machine interpret.