An Algoliterary Journey

The following pages describe the experiments selected by Algolit to share with an audience during the Algoliterary Encounter from 9 till 12 November 2017 in Maison du Livre in Brussels. By automatically exporting these pages from the Algolit wiki, this publication is an experiment in itself.

Algolit is a workgroup around algorithms and literature, a project of Constant, an organisation for arts and media in Brussels. In the footsteps of Oulipo, the Algolit members share their experiences with code and text during monthly meetings, and conduct new experiments together.

Over the last year Algolit explored the narrative perspective of machine learning algorithms as playful and legible stories. Taking implemented models as a starting point, they tried to get a grip on some of the materials that are used in applications around the web. During the Algolitery Encounter, the members of Algolit share a part of this research. They present lectures, workshops and a small exhibition.

Selflearning algorithms are based on algebra and statistics. They often function as opaque ’blackbox’ algorithms, while they shape applications that are daily used on a worldwide scale, like search engines on the web, translation machines, advertising profiling, facial recognition etc. Because machine learning models are so present, the members of Algolit felt the need to retell them in different ways, distillating reading and writing experiments from it. By executing parts of their construction process in a literary context, they become more legible. It is way to experience a few moments that are usually hidden in the making of the model, and that co-design contemporary stories in the way they influence the organisation of information.

The exhibition of Algoliterary Encounter is composed of two parts. Algoliterary works are creations by members of Algolit that have been showed and performed in different places before. Algoliterary Explorations is a collection of experiments that show how the group has been discovering machine learning algorithms. It reads as a methodology of how to understand the code of existing models, running scripts and play with their inputs and outputs, dissecting the process of machine learning models by exploring datasets, looking at methods to transform words into numbers, creating different views on the data and finally changing the scripts to our needs, customizing our tools.

As an addendum to the exhibition you will find the Sources, a bibliography of the books and papers that kept us company and a link to the repository where you can download all scripts and materials.

Organised by: Manetta Berends, Cristina Cochior, Gijs de Heij, Hans Lammerant, An Mertens, Translation: Emma Kraak, With the support of: Constant vzw/asbl, Vlaamse Gemeenschapscommissie, La Maison du Livre, Saison Numérique, Algolit, Brussels, October 2017., Copyleft: This is a free work, you can copy, distribute, and modify it under the terms of the Free Art License http://artlibre.org/licence/lal/en/