TheHmm-Preserving Practices

AI tools promise to save us time by taking care of the tasks we keep postponing, and archiving is usually the first to sink to the bottom of the to-do list. But what actually happens when we let automated algorithms touch our archives? Do they fill in the gaps and make our collections more complete? Or do our archives become training data, letting the algorithms decide for us?

Automated Archiving is the fourth and last session of How to Archive Better: Preserving Practices, a four-part workshop series on digital archiving that gives creators and makers the tools they need to archive their own practice.

In this fourth session, we dive into how algorithmic tools can support knowledge archiving, unpack how these systems work, and examine their pitfalls. Through these conversations, we want to raise questions about when algorithmic tools are actually useful, and what we should be conscious of.

The afternoon starts with a presentation by artist Linda Dounia Rebeiz and her studio partner Delali Vorgbe. Together, they have been working on a model that acts as an indigenous knowledge system repository and living archive of the Serer cosmology—the worldview of the Serer people of Senegal, the Gambia, and Mauritania. Linda and Delali will share insights into the art of building archives for model training.

Next, early internet artist Martine Neddam will briefly explain how the archive of her iconic artwork Mouchette (1996) formed the basis for a new project in which one can chat with Mouchette. Get a sneak peek into the work-in-progress.

The day will end with a working session led by artist and designer Leo Scarin, who will teach the practice of scraping. Learn how to extract data from online archives or even your own website. What does this reveal about how machine learning models are made, and which tactics can be used to keep work from being crawled by large-scale models?