Magnetic Fields Reimagined - Where Surrealism Meets A.I. (2022)

For this assignment we were asked to work as a team to develop group communication skills and amplify the dimension
of the final product. The project revolves around the key words “Missing Stories”. Devised as a reinterpretation of the first literary piece of surrealist work, this book converges the concepts of surrealism and artificial intelligence (or A.I.). It does so by taking as the object of both study and exploration of content Breton and Soupault's “The Magnetic Fields” (1920), deconstructing it and further re-organising it through Machine Learning mechanisms. Namely, these are RNN - which generates the bulk of the text with Breton's texts as an input - and VQGAN+ CLIP in turn generate the visual contributions
to the work, culminating, therefore, in a book mostly made up of generated text and visual content.
         Magnetic Fields Reimagined explores two crucial and reverse ideas. Firstly, it explores the relationship between
man and machine through the automatism processes associated with the surrealist movement. Secondly, and conversely,
it explores the relationship between machine and man as the first attempts to close the gap with the latter, that is,
the machine strives, through datasets and inputs, to analyse all information so as to become closer - or further - to surrealist writing, which is in Breton’s book characterised for being nonsensical.
This project and Breton’s work are similar in the way the final product is presented, becoming, gradually, with each page, more blemished. They differ, however, as in the original piece of work Breton accesses the subconscious through a method of automatic writing, resulting in many loose sentences that when read together are often incoherent. Magnetic Fields Reimagined, contrariwise, makes use of the fact that the writing processes in machinery are already automated to achieve those almost identical results.


Marta Martingo
Beatriz Caetano
Leonor Mendes
Steph Ramos (Developer)


João Castro
João Martino
João Tiago Santos
José Bartolo
Margarida Azevedo
Maria João Baltazar
Miguel Salazar
Rafael Gonçalves
Susana Fernando


KYRT 2022

Academic Project