Machine vision is already part of how moving images are made and shared: footage gets tagged, sorted, and retrieved; read by machines rather than looked at; and fed to users by algorithms. And it's not just that machines read images now, images are increasingly made for machines to read, footage staged and filmed only to train surveillance software. Following Harun Farocki's idea of the operational image, this seminar introduces AI-based image analysis as a hands-on, open-source practice, and builds practical and critical foundations for moving-image practices with AI. We start with archival footage from the internet. We'll learn the basics of Python and how to automatically cut, process, and sort video and images, and how to use AI as a critical tool: turning frames into text, sorting an archive by meaning instead of keywords, and treating its mistakes and biases as material to work with, not errors to fix. We can also turn the machine against its own logic by asking it for the frames that best match a word like "ice cream" in footage that has nothing to do with it, and building a narrative out of what it comes up with. Each week we read texts and look at artworks from different perspectives. We take a fun and humorous approach to AI while keeping a critical position toward technological systems. Note: This is part one of a two-semester seminar. The first semester is about practical tools and readings. The second semester shifts toward project-based practice and presentation, preparing a shared public exhibition. The seminar is funded by Stifterverband / eTeach-Netzwerk Thüringen. |
