An AI artwork needs its pace preserved, too

In Electronic Media Review’s 2019–2020 volume, conservators of Mario Klingemann’s Memories of Passersby I reported that their test graphics card was almost 45% faster than the original. Rendering and transitions still needed the artwork’s original pace.[1]

The 2018 work’s Companion version places two screens above a chestnut cabinet containing the computer. Its neural networks generate a continuing stream of portraits. They were trained on historical portraiture, and their outputs feed the changing sequence. A photograph records one appearance of this work; it cannot keep that process unfolding.[2]

Describe what must survive

The Smithsonian American Art Museum’s identity report asks conservators and curators to specify significant properties, exhibition history, the artist’s information and conservation risks. A separate iteration report records each installation. Those records can refine the identity report as people learn what changes between presentations.[3][4]

The distinction gives preservation two different questions. What makes this the same work? What happened when it was installed here? Keeping both answers lets a later conservator judge a change instead of merely discovering that somebody replaced a component.[3][4]

Save a working environment

Docker packages an application with supporting software in an image, a template used to create a running container. The image and the running instance are different objects. That distinction matters for an archive: retaining the environment gives someone a way to start the program again, while documenting a running instance records a particular use.[5]

The team created a container while retaining dependencies on NVIDIA hardware and drivers. The package preserved a software environment that still needed a compatible host.[1]

The next visitor can encounter a new portrait. Preserving this work means keeping its image-making process available at the right pace.[2][1]

Report an error or suggest a correction