Machine Learning and the Image of the City

This project-based seminar explores the potential for machine learning to enhance our creative process as we re-imagine the image of the contemporary city. This work is occasioned by the recent acquisition of Ed Ruscha’s Streets of Los Angeles archive by the Getty Research Institute. In partnership with the Getty, the project-based seminar will utilize the vast digital record of tens of thousands of Ruscha’s photographs of Sunset Boulevard taken between the 1960s and 2010s. This vast digital image archive will inform the development of machine learning processes that will allow students to extrapolate potential alternative images for the contemporary city.

Ruscha’s deadpan photographs of Los Angeles’s iconic streetscapes and automobile-based architectural typologies were appropriated by Denise Scott Brown as a graphic language applicable to the analysis of the Las Vegas strip as published in Learning from Las Vegas. Ruscha’s photographs were equally influential to Reyner Banham’s conception of the city’s Four Ecologies. In both cases, the postwar American city was seen through the lenses of limitless solar plentitude, extreme illumination, and the legibility of information at speed.

The GSD Office for Urbanization has worked in collaboration with Jose Luis Garcia del Castillo Lopez and GSD Laboratory for Design Technologies to develop specific workflows based on the Ruscha archive. Spread across several decades, these images present endless opportunities for surveying change throughout the city. Students will be invited to track, monitor and project forward these changes by cross referencing each image’s geospatial data with object detection and depth map generation algorithms. The workflow will also deploy generative adversarial neural networks, such as StyleGAN, to project a limitless number of imaginary cities extrapolated from the Getty’s archival Ruscha images. These “machine hallucinations” will inform the curation of new time-based images of the contemporary city.

Each member of the seminar will be invited to develop facility with machine learning processes to curate new architectural imaginaries for the future of the city through animated images of the public realm. The seminar will also convene a series of conversations with leading voices across a range of topics including the role of Ruscha’s image of the postwar American city and the potential for machine learning as a generative process. The seminar welcomes candidates from all departments and programs across the GSD. It welcomes students with little or no experience with computation, as well as those with more experience. The seminar forms part of the GSD’s Future of the American City Initiative sponsored by the Knight Foundation and supported by the GSD Office for Urbanization.

Up to four seats will be held for MDes students.

This course will be taught online through Friday, February 4th.