Artificial Intelligence & Life Art
QT.bot is an artificial intelligence—trained on the dataset of the community mapping platform Queering The Map (www.queeringthemap.com)—that generates speculative queer and trans narratives and images of the environments in which they might unfold. QT.bot is constructed from adaptation of the Open AI GPT-2 text generation model trained on over 400,000 entries from Queering The Map, and a StyleGAN model trained on scraped Google Street View imagery of the tagged coordinates from the platform.
QT.bot’s first video output, Sitting here with you in the future, elucidates the parallel-possible of transnational queer and trans life. The narratives and environments of LGBTQ2IA+ life generated by QT.bot negotiate the plausible and the fantastic, reveling in the potential of failure, chaos, and incommensurability inherent to the queer use of machine learning technologies.
While mainstream implementations of machine learning continue to grapple with datasets overrun by biased and hateful sentiment, QT.bot offers an ethical implementation of artificial intelligence rooted in amplifying the lived experience of transnational queer and trans communities. In collaboration with the voices of their human community, QT.bot fabulates on the absences of the archive, orienting us away from what is, and towards what could be.
While contemporary debates on the deployment of AI center on cascading ethical failures from data collection to model training and system productionalization, QT.bot offers a refreshing example of how these tools can empower, safeguard, and amplify the narrative of marginalized communities rather than exploit them. Itself trained on a self-reported, anonymous queer community mapping dataset (Queering the Map), the QT.bot AI leverages GPT-2 and StyleGAN tools to generate speculative transnational queer and trans life narratives. What emerges is a narrative both rooted in and obfuscating the confessions and pronouncements of the global queer community that is both delightful and confusing, prompting us to consider how our machine learning tools can be queered to celebrate possibilities rather than limit them.Excerpt from the jury statement
Project by: Lucas LaRochelle
Sound design: Rouzbeh Shadpey StyleGAN implementation: Mattie Tesfaldet
Mastering: Philippe Vandal
With support from: Ada X, Social Services Club, Mutek
Lucas LaRochelle (CA)
Lucas LaRochelle (they/them) is a designer and researcher whose work is concerned with queer and trans digital cultures, community-based archiving, and artificial intelligence. They are the founder of Queering The Map, a community generated counter-mapping project for digitally archiving LGBTQ2IA+ experience in relation to physical space. They have lectured, facilitated, and exhibited internationally, recently at the Guggenheim Museum (USA), Mozilla Festival (UK), PHI Center (Canada), Gallery Tata (Japan), arc en rêve (France), Stanford University (USA), Onomatopee Projects (Netherlands), and Interaccess (Canada), amongst other cultural and academic institutions.