( OS prefers reducing animation! )

Research Blog

Home > Research > CI Research Blog > Pelican Stairs: an AI-Assisted Pandemic Memoir

Pelican Stairs: an AI-Assisted Pandemic Memoir

Creative Informatics Researcher Caitlin McDonald discusses her project Pelican Stairs: a Wapping Great Pandemic Memoir. Pelican Stairs is a multimedia art project, merging diary entries with GAN-generated images to represent the blurred, overlapping, confusing nature of the first six months of the Covid-19 lockdown. A full paper on the project will be published in the forthcoming proceedings of Generative Art XXIV. Here, she introduces the project and discusses how concepts explored in Pelican Stairs connect with avenues of research for Creative Informatics. 

Introduction: what and why

Pelican Stairs is a multimedia art project started during the depths of Covid-19 lockdown in 2020. I took photos of my local neighbourhood, Wapping in London, on my daily walks between March and September. I often visited the Thames shore area by climbing a set of steps known as Pelican Stairs. Using those photos as a training set, I generated new images with a Generative Adversarial Network (GAN), which I paired with excerpts from my diary during the same period. The unsettling, almost-real images contrasted with the mundane reality of the diary entries allow the viewer to viscerally (re)experience the inherent tension between an increasingly uncertain external reality and internal attempts at control or sense-making through normal, everyday habits. The full project is hosted at https://pelicanstairs.art.

Figure 1: 27 March 2020 diary entry including GAN-generated imagery from Pelican Stairs.

About a year later I used the photos I took during that time to create new images using StyleGAN2, which I trained using the commercially available tools in RunwayML. The distortions and surreality introduced by feeding the images through a GAN represent the blurred, overlapping, confusing nature of that time. Pairing them with real extracts from my diary entries during that period captures the tension between the attempts at everyday normalcy and an ever more uncertain outside reality. (The diary entries are real, but I’ve changed people’s names to protect their privacy.)

One of the reasons I chose to use the images in this way is to explore our increasingly digitally-mediated lives: more and more of our interactions with others take place through digital technologies. For me, this became particularly noticeable during the pandemic when those were often the only means available for contacting loved ones or simply conducting day-to-day business. In my experience closeness is both facilitated and occluded through these digitized windows into each other’s worlds. The GAN images, based on reality but not real in themselves, amplify the distortions and communicative frictions that accompany the digital communication tools we all came to rely on more than ever over the past few years.

Figure 2: 2 April 2020 diary entry including GAN-generated imagery from Pelican Stairs.

The GAN model can generate a theoretically infinite number of new images based on my original collection of photos. I chose to bound this project within the six months between the announcement of lockdown and my move from London to Edinburgh, but for as long as life remains disrupted, I can continue to generate new images that smear out the initial confusion and blurriness of that time to infinite strange new futures. Here, in this space between the river and the shore, in this time adrift, is Pelican Stairs.

Figure 3: 27 September 2020 diary entry including GAN-generated imagery from Pelican Stairs.

Video Presentation

I recently presented the project at DigiFest, an arts and design festival hosted by the Faculty of Arts and Design at Durban University of Technology, where Pelican Stairs won the Best Individual Presentation award. You can view the presentation in the embedded video below.

For a deeper dive into how I practically created the generated images and the Twitter bot accompanying the project, keep an eye out for the forthcoming Generative Art XXIV publication.

Discussion and future research areas

The arts have long been a means for driving innovation in the field of technology as well as for critically reflecting on potential technological impacts. I use the arts to enhance my research: to better understand novel technologies, their uses and misuses, I explore them through art projects that are illuminating, aesthetic, or simply whimsical. Pelican Stairs offered the opportunity to continue my exploration of the reflexive interplay of my practices and capabilities as a technologist, an artist, and a researcher. It fits within a growing body of work in sociotechnical studies examining the critical and productive possibilities at the intersection between the arts, design and cultural sectors and the tools, techniques and data of the technology sector. This fits squarely within the Creative Informatics role of exploring data-driven innovation in the creative industries. There are several concepts from Pelican Stairs that could be further explored through the CI programme of work.

First, how does adoption or ubiquity of technological innovation create opportunities for non-specialists? For example, digital innovations such as desktop word processing, photo and video editing opened up new technology-assisted creative possibilities for those without specialist training in design, publishing, photography and film. Spreadsheets created new informational management capabilities reaching far beyond accountancy. Similarly, tools like RunwayML, which I used to generate the images for the project, open up machine learning techniques to those without data science skills.

Second, I believe that inexpensive machine learning tools ready for use by non-specialists creates new opportunities for reengaging with questions about the meaning, role and definition of art, much as paint-by-numbers kits offer a vehicle for critical reflection on the distinction between commercial and fine arts, or Warhol’s Brillo boxes inspired Danto’s reconsideration of what distinguishes art from other objects or modes of expression. What is the root of the creativity in a project like Pelican Stairs? Is it in the original photographs I took? In the sophisticated processes created by the authors of StyleGAN2? Somewhere between the work and the viewer, who must apply their critical reflective skills to create meaning from the images and text? Or is it imbued throughout all these processes?

Reclaiming technology as a means for realizing imaginative possibilities encompassing the irrational, intuitive and whimsical is another avenue for exploration within the Creative Informatics mission.

Finally, using digital tools to create the surreal specifically for the purpose of an artwork upends frequent assumptions about the value of digital technologies, particularly AI. Among technologists, AI is often presented as an uncritical inherent good for all; an inevitable march of progress. There is a growing body of critique against this assumption, including among the work of my peers in Creative Informatics and in the broader Edinburgh University research community. Using a GAN for a deliberately non-functional creation can be viewed as an act of rebellion. Reclaiming technology as a means for realizing imaginative possibilities encompassing the irrational, intuitive and whimsical is another avenue for exploration within the Creative Informatics mission.