Full Text at http://databasecultures.irmielin.org/how-to-hack-artificial-intelligence
This 2018 essay explores the weaknesses of automated pattern recognition (aka artificial intelligence) and discusses, how artists, scientists and activisits have been able to exploit them, using adversarial patterns. Arguing about human agency in relation to machine agency it discusses a range of adversarial methods:
- abuse scope
- change the color
- altered / trap iconography
- perturbation
- sensor blast
- camouflage mask
- camouflage surroundings
- arms race: machinic un/fake
- input trolling
- tag trolling
- be obscure
A version was republished as How to Hack Artificial Intelligence. Artistic projects and current research on the (dis)abilities of machine learning techniques. In: State Machines – Reflections and Actions at the Edge of Digital Citizenship, Finance, and Art. Edited by: Inte Gloerich, Marc Garrett and Yiannis Colakides. Institute of Network Cultures. Amsterdam 2019
A version of the essay was presented as a talk at transmediale festival 2018.