NeurIPS - Workshop on Machine Learning for Creativity and Design

A Graphic Novel Generated by Artificial Intelligence (2020). While speculations about the growing role of machines in artistic production have been a consistent trope in modern and contemporary art debates throughout the 20th century, comics from their early beginnings, have been symbiotically expanding with the development of printing, distribution, communication and media technologies. These industrial processes of completion based on generalized automation, standardization practices and an orchestrated division of labour are so embedded in the ways we understand and consume comics that have become an essential feature for the conceptualization of artistic practices in the medium. A typical production line of manga comics for example involves dozens of people handling specialized roles in a quasi-taylorist production belt, often in ways that have been criticized for resembling a sweat shop, while distribution has been increasingly involving massively digitized operations of logistics and global supply chains.
Comics is an industrial form of artistic expression. At the same time, the comics industry has been quite reticent in embracing the complex nature of technological developments in artificial intelligence. But this situation is about to change. The online abundance of digitised media content, available through third-party groups of comics fans, the increasing convenience of programming language frameworks and machine learning libraries, the secularisation of knowledge through e-learning and the plummeting prices in specialised hardware might contribute to reach a critical point where artificial intelligence will be gradually integrated in the comics pipeline. Synthetic and generative processes might soon reshape the ways we produce, consume, archive and distribute comics artefacts.

Applied Memetic is a transdisciplinary team of artists, researchers and computer scientists with the goal to synthesize the first comic book narrative entirely using machine learning. It’s an opportunity to explore processes that don’t conventionally account for the production of comic books, by weakening the aesthetic predispositions and received knowledge that are reproduced through specific (human) evolutionary interpretations of artistic production. Instead, we are interested in harnessing the machinic understanding of comics through recurrent patterns, probability distributions and outliers in comics language that have been lurking in our pre-attentive reader’s cognition and that we haven’t been able to articulate in words.