Background
As innovation with using artificial intelligence in art making continues, museums’ conservation practices must also evolve. Continuing my research on AI art preservation, this article will narrow down the subjects to AI-powered art since it deviates further from what traditional conservation practices apply to. Unlike static digital artworks printed on canvas, AI-powered art always has temporal attributes, presents constant variation through an algorithm, and relies on digital files and equipment that runs continually. It also shares much in common with other digital art, which makes conservators accustomed to working with tangible materials deal with electronic devices, computer programs, and often, the underlying data as a whole (Lloyd-Baynes 2020). Although current studies have not focused on AI-powered art, it is worthwhile to see how ideas and practices of digital art preservation are transforming in ways that could help preserve AI art in the future. Since manufacturers' innovation will quickly outdate museums’ technology supporting such art and it will be hard to tell when the doomed obsoletion will happen, being proactive is essential (Minneapolis Institute of Art n.d.).
Introduction: Guggenheim Case study
Museums are among the earliest digital art collectors and have had conservation projects for AI-powered art’s parent genres for years. Their research on time-based media art (TBMA) and computer-based art are of great help for understanding AI-powered art's durational property and technological components. Among the museums I looked at, the Guggenheim was the only museum that divided conservation specialties for these two types of art. This article will utilize a single case study on the Guggenheim to explore AI-powered art preservation from different perspectives.
Figure 1. Institutions with Conservation Projects for Media / Digital Art. Source: Author.
Documentation for Time-based Media Art Conservation
Similar to many institutions that conserve TBMA, the Guggenheim is concerned about the failure and obsolescence of technology within an artwork. When component replacement is inevitable, the museum will determine the acceptable degree of changes made to a piece in the short, middle, and long term (Guggenheim n.d.). The museum also considers TBMA's allographic nature, a characteristic that means that the composition and the execution of artwork are two separate stages (Marçal n.d.). Apart from protecting an art piece, the museum needs to limit iterations during the second stage to represent a work in an authentic way (Guggenheim n.d.). Going through the decision-making process as well as the installation, comprehensive documentation, and deep understanding of the relationships between the components and the artist’s intention are the keys to developing best practices at the Guggenheim.
Figure 2. Example Iteration Report for TBMA at the Guggenheim. Source: Guggenheim Museum.
The documentation efforts begin with artist interviews as soon as a piece of TBMA is acquired. From the interviews, the museum typically records the artist’s interpretation of the work, the technological component analysis, and the results of the quality check for both the original work and any other copies. Since museums are currently conserving contemporary artworks whose creators are typically alive, not only should a museum take action with artists' approval, but it should also track how artists want to reframe a work even after the acquisition (Marçal n.d.).
As for the technological elements, the Guggenheim categorizes the digital files as well as the equipment, after which they differentiate in handbooks. Given that AI-powered art runs algorithms and codes, its future collectors can customize their categorization standards based on project file formats, sizes, platforms, or computer languages apart from the quality of video and audio, such as HD or SD. In terms of devices that display AI-powered art, the Guggenheim's classification criteria are particularly useful considering both the obsolescence degree and specific equipment’s uniqueness. The criteria directly tell when an item needs a substitute and whether it is exchangeable with products currently circulated in the market.
Figure 3. Guggenheim’s Classification for Equipment for TBMA. Source: Author. Reference: Guggenheim Museum.
In addition to pre-examination, on-site documentation is another step in the Guggenheim's conservation work for TBMA because, as mentioned before, subtle changes may happen during the installation procedure under different exhibition environments. By updating records about an artwork's vulnerability, obsoletion level, etc., the museum gains opportunities to form institutional knowledge of their TBMA collections (Guggenheim n.d.). The museum, however, does not have permission to document every artwork, either thoroughly or partially. Some artists may disapprove of, for instance, video recording of the installation process since that is a part of a work; in this case, the museum can reinforce human memories instead through repeated hands-on experience of restoring and dismantling (Roemich and Frohnert 2019).
Cross-disciplinary Collaborations for Computer-Based Art Conservation
Computer-based art—which includes integrating computer hardware, software, and codes into an artwork’s exhibition—resembles TBMA in terms of the inherent instability related to technology obsolescence and physical deterioration (Guggenheim n.d.). It intersects TBMA when a final artwork has a dimension of time. AI-powered art falls into the intersection of these two types of art due to its dynamics presented over time. When the conservation team faces either computer-based art or TBMA, the core task is exploring the interdependence between technology and an artistic object. Apart from Iteration Reports, Identity Reports about an artwork's anatomy, intended experience, historical contexts, installation parameters, etc., are vital tools utilized by the Guggenheim.
Figure 4. Example Identity Report for Computer-based Art at Guggenheim. Source: Guggenheim Museum.
While conservators undertake some hands-on work for migrating data or preparing copies of files or accessories, most of them, trained to treat physical objects, need sufficient technical instruction and support. What makes the Conserving Computer-Based Art Initiative (CCBA) unique is that there is a robust collaborative relationship among conservators, computer scientists, and engineers since analyzing artist-created source code is often unavoidable.
“The Art of Code: Join Dylan Beattie - programmer, musician, and creator of the Rockstar programming language - for an entertaining look at the art of code.” Source: YouTube.
The Guggenheim built a well-rounded conservation team by developing an academic-museum partnership with New York University. The university's computer science faculty and students are responsible for probing into an art piece and providing documentation, such as annotation added in a copy of source code. While conservators will keep refining these written records, including the reports mentioned above, they also adopt screen recordings and narrations of an artwork's virtual environment, software interface, and operation procedures. Such vividness ensures a more profound institutional memory of artworks regardless of any personnel changes within a conservation team, despite that the documentation may still be subject to artists' permission.
Figure 5. The Model of Conservators and Computer Science Experts’ Collaboration for CCBA. Source: Author. Reference: Guggenheim Museum.
When it comes to AI-powered art preservation in the future, museums can modify and strengthen such joint efforts by bringing in data scientists because of the algorithms, machine learning, or artificial intelligence involved, which are at the core of AI-powered art. For an artwork that requires AI training, conservators need to interview artists and data analysts about how they proceed with each step. Since AI-powered art results from an actively running AI program, conservators may also want to record any self-evolution of the AI after it receives live data feed, such as real-time recording transmitted to the computer. Rather than being auxiliary components, algorithms and data for AI-powered art are not only co-creators with artists throughout the entire production process, but they also determine the varying effects during the final exhibition. New documentation forms and classification standards will likely be tailored specifically based on types of data and algorithms, such as Generative Adversarial Networks (GANS), to recognize an artwork's identity.
Conclusion
Given that the Guggenheim's CCBA Initiative has already conserved a programmed robotic installation created in 2016, collecting AI-powered art or other works co-created by humans and machines may not be a distant future. Although technology-based art challenges conservators' current skillsets, discovering effective external support will be a helpful start and stimulate the internal transformation of conservation ideas and procedures. As long as museums have developed a flexible conservation system that welcomes alterations and additions of practices, it is time for them to unfold their existing tech-based art collections and prepare for acquiring and conserving more non-traditional artworks, including AI-powered art.
Resources
Guggenheim Museum. “Time-based Media.” Accessed May 6, 2020. https://www.guggenheim.org/conservation/time-based-media.
Guggenheim Museum. “The Conserving Computer-Based Art Initiative.” Accessed May 6, 2020. https://www.guggenheim.org/conservation/the-conserving-computer-based-art-initiative#dissemination.
Guggenheim Museum. “Identity Report for ‘Can’t Help Myself’ (2016) by Sun Yuan and Peng Yu (PDF).” Accessed May 6, 2020. https://www.guggenheim.org/wp-content/uploads/2019/12/guggenheim-identity-report-cant-help-myself-sun-yuan-peng-yu.pdf.
Lloyd-Baynes, Frances. "Preserving Digital Art: the Innovation Adoption Lifecycle." Museum-iD Magazine, February 7, 2020. https://museum-id.com/preserving-digital-art-the-innovation-adoption-lifecycle/.
Marçal, Hélia. “Contemporary Art Conservation.” TATE. Accessed May 5, 2020. https://www.tate.org.uk/research/reshaping-the-collectible/research-approach-conservation.
Minneapolis Institute of Art home. “Preserving Mia’s Time-Based and Digital Media Art.” Accessed May 6, 2020. https://new.artsmia.org/art-artists/research/case-studies/preserving-mias-time-based-and-digital-media-art/.
Moxley, Mitch. “The AI Art Boom.” Barrons, September 24, 2019. https://www.barrons.com/articles/the-ai-art-boom-51569348001.
Roemich, Hannelore, and Frohnert, Christine. “The Art of Conservation.” Interview by Gabriel Barcia-Colombo. State of the Art, Spotify, December 13, 2019. Audio, 34:50. https://open.spotify.com/show/26Qh9Oe7yQ3dAJ5VGUf9jl.
Web Desk. “How Artificial Intelligence Is Changing the Art Industry (Infographic).” Digital Information World, February 9, 2019. https://www.digitalinformationworld.com/2019/02/infographic-how-artificial-intelligence-is-changing-visual-arts.html.