Research Project

Collaborative Artistic Production with Generative Adversarial Networks

At various points in history, technologies such as written language, the printing press, optical lenses, and semiconductors have dramatically changed human creative and artistic processes. Machine learning and the automation of generative processes might represent a similar paradigm shift. Given the rapid acceleration in the capacity and variety of machine learning systems, humanities scholars and artists must engage from both a practical and a theoretical level. Our project, entitled ‘Collaborative Artistic Production with Generative Adversarial Networks’, examines these new processes available to creative practice, and how they intersect with the assumed authorship of the artist. This project specifically uses Generative Adversarial Networks (GANs) to see how basic artistic principles such as form, function and aesthetics might change due to the introduction of a semi-autonomous system of generation.


We take advantage of the interdisciplinary skill set of researchers within the Augmented Creativity Lab and use a ‘rolling PI’ model to transfer the focus of our project over four separate research stages, each involving their own sets of sub-questions. First, we examine how GAN systems can function as artistic tools for the creation of 3D forms, and how reversible encodings of these forms can be used to experiment with traditional semiotic systems. For example, if a computer can encode specific information in the form of trees or mountains, how might this change classical systems of image-symbolism? Second, we examine GANs’ ability to discern patterns and embedded information in a diverse data set of hand tools spanning human history. The result will be 3D printed objects that are reflections of the analyzed hand tools and the processes of the GAN system. This has the potential to illuminate unforeseen connections across eras of human-centered design. Third, we apply a GAN system to a music ecosystem in an attempt to blur the boundary between the human creator and the generative system in a live performance context. Fourth, we take advantage of cultural studies and ethnographic methods to examine how our practice-led approach and our use of these technologies can be resolved more broadly within historical models of the artist and the artwork. This phase of the project also explores ways in which GAN's - as socio-cultural objects - are recursively defined in artistic processes.


Figure 1

Figure 1.


Figure 2

Figure 2.