About this Project
Can a computer be programmed to make and appreciate great art? This is the question at the heart of the Critical Computation project (2016–2018), which investigated the creative potential and theoretical implications of machine-generated visual art and design. The project was among the first significant scholarly efforts to explore the use of machine learning for the creation and evaluation of cultural artifacts.
Computer scientists are now able to harness sophisticated algorithmic methods to identify common features across huge databases and then apply that learning in new contexts. Machines with the capacity to “learn” are reshaping our society in fundamental ways. New applications have improved medical diagnosis, demographic targeting, fraud detection, and financial analysis. Artists and designers are also experimenting with machine learning, though the humanistic questions posed remain largely unaddressed. What aesthetic possibilities are created through this new technology? How is machine learning itself shaped by the software engineers’ value judgments? Crucially, can machine-learning methods be adapted to incorporate traditional notions of quality?
Working with a team of undergraduate and graduate students, postdoctoral fellows, faculty, and staff, Jason Salavon transformed his studio into a laboratory environment combining art production, computer science, applied mathematics, and philosophical inquiry. In its first year, the group conducted more than 500 experiments to test artificial intelligence capabilities with regard to still images and video. Early in their experimentation they focused on “deep learning”: machine learning that uses large-scale neural networks to solve a wide range of otherwise intractable problems. Interest in deep learning is now spreading throughout the world, and the Critical Computation project placed the University of Chicago at the forefront of exploring deep learning’s capabilities in creative imaging and architecture. The team also developed ties to top researchers and developers working on computational art-making. And they hosted a private weekly seminar to share and brainstorm ideas, address technical challenges, and present formal papers.
Presentations that were open to the public gave the research team the chance to showcase their work and exchange ideas with audiences interested in artificial intelligence and machine learning. Salavon presented a series of “generative” paintings created by a program that had learned how to reproduce Abstract Expressionist style at the University’s 2016 Innovation Fest. He presented more recent experiments at the 2017 Eyeo Festival, an annual conference for professionals working at the intersection of art, data, and creative technology, and also exhibited ongoing work from the project at the 2017 Conference on Neural Information Processing Systems.
In May 2017 the Neubauer Collegium welcomed Zoë Prillinger and Luke Ogrydziak, principals at the pioneering architecture firm OPA, who discussed three projects that applied generative computational methods to the design of residential homes. The University subsequently commissioned OPA to create a temporary architectural installation for the seventy-fifth anniversary of the first controlled nuclear chain reaction. Partially encircling Henry Moore’s Nuclear Energy sculpture with black rubber cord, OPA’s Nuclear Thresholds used computational modeling of unstable processes to provoke questions about the science, history, and existential realities of the nuclear age.
The Critical Computation project enabled Salavon and a team of researchers and developers to create and launch Genmo, a neural-network-driven visual effects application that re-creates any photo or video using an entirely separate set of images. Genmo replaces standard social media filters with generative effects, bringing AI-powered creativity to mobile phone users around the globe.
“The proliferation of user-generated content and the creative limitations of existing technologies have paved the way for artificial intelligence to rethink the social photo/video creation and sharing experience, allowing for content creators to leverage their idiosyncratic behaviors and augment their visual production,” Salavon said. Genmo won the Winter 2018 UChicago App Challenge and launched in 2018. Post-launch, the technology will learn about users’ content interests and behaviors, and the app’s visual effects will evolve accordingly.
March 21, 2018
Jason Salavon (DOVA, Computation Institute) has won the Winter 2018 UChicago App Challenge for Genmo, a photo filter that uses generative AI effects. The app was developed as part of the Critical Computation project at the Neubauer Collegium.
April 14, 2017
Faculty Fellow Jason Salavan (Critical Computation) will be the first artist exhibiting in a new installation space in downtown Chicago, reports Chicagoist.
April 1, 2016
There are no events associated with this project yet.