As published in Art in America, November 27, 2019

Clouds have long posed representational quandaries to painters, photographers, and programmers alike. Artist and architect Lucy Siyao Liu (ACT Lecturer, 2017-18) met with network engineer-cum-English professor Tung-Hui Hu to discuss their respective work on these elusive wisps. The exchange took place on a summer Sunday in Liu’s Brooklyn studio over buckwheat tea. While Liu’s art has grappled with how to represent clouds, Hu’s 2015 book A Prehistory of the Cloud asks what “the cloud” represents, showing how the digital term, so often used to imply placelessness, actually refers to a hefty infrastructure. Liu’s installation A Curriculum on the Fabrication of Clouds (2017)—currently on view in a group show at the Moderna Museet in Stockholm—mixes the artist’s drawings and animations with works loaned from archives and museums. It’s modeled after art historian Aby Warburg’s Mnemosyne Atlas (1924–29), an arrangement of images (originally displayed on wooden panels) that reveal visual commonalities across time and multiple cultures. Liu samples attempts to record the ephemeral, including drawings by nineteenth-century British critic and artist John Ruskin, works by eighteenth-century British landscape painter Alexander Cozens, and 1920s photographs by Japanese physicist Masano Abe.

LUCY SIYAO LIU There’s not much in my studio now, because I shipped all of my drawings to Stockholm. I do have a brainstorming board for my installation there, A Curriculum on the Fabrication of Clouds. The piece includes archival drawings and diagrams by various artists and scientists related to the depiction of clouds. I made an animation that I’m projecting at a forty-five-degree angle, reinterpreting the rectilinear and curvilinear grids Ruskin used to draw clouds in perspective. I’m also showing some drawings following the rules and schemas of Alexander Cozens. And there’s my scroll with samples of etching techniques commonly used to depict clouds in reproductions of paintings.

TUNG-HUI HU I’m struck by Cozens’s procedural approach to landscapes: he treated it as something that could be executed by computers, whether a human computer or a machine. I love the way your connections map the representations of clouds through systems, whether those of Cozens, Ruskin, or John Constable.

LIU Cozens’s book A New Method of Assisting the Invention in Drawing Original Compositions of Landscape [c. 1785] is mostly about his blotting techniques. He would create blots, then draw a landscape from those forms when everyone else was painting en plein air. The book has twenty or so plates of cloud etchings with descriptions like “streaky clouds at the top of the sky.” Cozens wasn’t necessarily drawing in relation to a real landscape, but generating landscapes. I borrowed his approach to draw clouds computationally, then borrowed his captions for titles. Of course, my methods are very different: I’m programming and controlling a pen plotter, not etching.

HU There are brief moments of interruption in the plotting, not just smooth lines. I’m curious about the ways you produce something with fixed rules, but also interrupt it or introduce contingency.

LIU I like working with the plotter because it takes time. It’s not like a printer that prints all in one go: the plotter reads vectors, and vectors have end points. This duration introduces a lot of opportunities for interjecting my hand back into the work, or letting circumstances or materiality have some say in the matter.

HU Your attention to each individual line makes me think about labor. I’ve been reading about the people whose job it is to render clouds in video games: they archived all their failed clouds, which got me thinking about the labor of producing a background. It’s these things in the background that make things in the foreground, like characters, seem livelier. What is it like to do all this painstaking work to make the background?

LIU All those nineteenth-century Englishmen were leisurely strolling in the pastoral landscape, exclaiming how beautiful it was, and then trying to capture it. I relate to their pleasure and enjoyment. It’s interesting that today, when we see line drawings, we think of labor.

To me, the ones doing the labor are people like the physicists I interviewed at the International Meteorological Institute in Stockholm. They told me they have to manually check the numbers that their cloud model outputs, then synthesize them and analyze them. They’re responsible for giving an accurate forecast to people out at sea, but I also think about the labor of being a human in this computer process.

HU As an artist, you’re doing programming work but also historical work. I’d love to know more about how you see your role, bringing together and even combining different representational systems.

LIU It’s a question I ask myself. Artist Mariam Ghani, who works with Afghani archives, once noted how hard it is to deal with archives, because you have to contend with the social structures and ideologies that determine what’s worth archiving. I really feel that kind of responsibility in acknowledging that the references I bring to A Curriculum reflect a colonial past and a deeply Western tradition of attempting to understand or control weather. That’s why I’m mixing my own work with historical work, but not in chronological order. Like Aby Warburg’s Mnemosyne Atlas, my installation is a brute-force formalist comparison. I’m trying to assimilate these precedents, not put them on a pedestal.

HU I’m reminded that the etymology of the word “cloud” follows your method of formal comparison: it comes from “clod,” because the clumps in the sky look like the ones on the ground. There’s something beautiful about the fact that language sometimes comes from formal intuitions. Your achronological model of history makes new insights possible through these moments of formal comparison.

LIU People are always asking artists about their responsibility to history, more than I think they ask other people. There’s always the assumption that the artist is here to mess things up and put things out of order; that by misrepresenting things, one might get the audience to reconsider them. But misconstruing history can be really dangerous.

HU The model of history that Silicon Valley wants us to have—and even the model of history that was sort of in the air when I was writing my book [A Prehistory of the Cloud]—is one of breaks or disruptions, often a kind of technological determinism. It was important that I thought of my book as “a” prehistory, not “the” prehistory of the cloud. I was trying to connect strands but not in a directly causal way.

LIU I was intrigued by the discussion of “virtualization” in your book. You wrote about the history of data centers and networks, and how the word “cloud” now symbolizes that infrastructure. I’m thinking about virtualization in my work a lot, because I use physical materials (archives, drawings) to get at the physical qualities of natural clouds, which is quite different than today’s dematerialized sense of the word “cloud.”

HU Both the material and the virtual play important roles in my book: the data centers, the fiber-optic cables along railroad tracks, but also virtualization, which is what computer scientists are referring to when they talk about cloud computing. I was trying to make the claim that virtualization is at the heart of our economy now. Amazon Mechanical Turk treats freelancers as tools, just as it does any software package or storage service. It doesn’t really matter whether it’s a human or a computer producing the result. Cultural studies increasingly focus on infrastructure and the material layer of computing, but I think we should also be looking at the systems by which things become abstracted and virtual, and what happens to history and the archive when they become progressively virtualized.

LIU The cloud metaphor constrains our ability to see its infrastructure. But we like that, because it also makes it easier to accept.

HU Yes, the metaphor shapes how we think. When scientists were trying to figure out what to call computer viruses, one proposal was the term “weeds.” They went with “virus,” which spawned a whole line of thinking about contagion and other biological metaphors, but what if we talked about gardening computer weeds instead?

And, as you say, we desire that mystification—not because we’re stupid and we’re all being exploited by evil people, but because we have a really strong desire to trust that all our memories are being protected or saved, even if we kind of know that’s not totally true. I wonder what the cloud is in the popular imagination. I’d guess that it stands for something not here, but somewhere out there: in the network, in the ether. I wonder if children who are growing up now really see the difference between “cloud” and “not cloud.”

To counter that mystification, artist statements frequently invoke the trope of “making the invisible visible.” But your work doesn’t quite do that. Instead, it makes visible techniques for visibility, like talking about an alphabet before talking about concepts written in that alphabet. You’re pointing to systems of drawing before getting to drawings themselves. When I was in grad school, computer art mostly involved data visualization: like, hypothetically, a red, green, and blue display that visualizes the sound of water, or a more political work that might visualize the number of deaths in Iraq. A lot of these projects would be examples of “making the invisible visible.” Though there is some of that inherently in the idea of drawing a cloud, or trying to capture a constantly moving object.

LIU In my title A Curriculum on the Fabrication of Clouds, “curriculum” implies a technique or methodology, as well as a survey. I also say “on the fabrication of,” fabrica in Latin, meaning skill, craft, or production: not necessarily a replication of a real thing. I am making visible the techniques, but also questioning them by putting them all together. Alone, you might think Cozens’s method, or Ruskin’s method, is “the” method, and that one of these people is “the” authority on how to represent clouds. But put those techniques together, and you might ask which one is more accurate, or conclude that they all produce fabrications.

I also have some prints by George Sand [pen name of Amantine Lucile Aurore Dupin] in my installation. Before the Rorschach test, putting ink down and folding the page was called “klecksography”: Victor Hugo also made some. I included some prints from Sand, who would peel the paper away; if she would see veins that look like a tree, she’d make a landscape. I want to muddy the waters, asking how we have chosen to standardize representations, so I included magical methods alongside masterful ones. There’s also August Strindberg’s celestographs: he rubbed photosensitive chemicals on paper and exposed the sheets to the stars to make photographs without a lens. I’m working toward a pluralist understanding of what method is, because even “method” is a pretty colonial idea: having to methodize something as knowledge production, rather than allowing for idiosyncrasies.

HU It’s fascinating to recall that early photography couldn’t capture clouds. Newspaper editors would always ask photographers to dodge and burn those clouds back in. Later critics such as John Szarkowski claimed those attempts were a holdover from a painterly conception of landscape that people weren’t ready to let go of. Photographers who left blank skies in the backgrounds of their images, such as Timothy O’Sullivan, were really embracing the technology of photography: they were modern, in Szarkowski’s narrative. And though we now know that’s a simplistic view, the difficulty of representing clouds is still a hinge point for thinking about new systems of representation and new ways
of looking.

LIU Today, machine learning models have difficulty interpreting foregrounds versus backgrounds. Our images are getting reduced and compressed: I wonder if clouds will always be part of them.

—Moderated by Emily Watlington


This article appears under the title “Computing Clouds” in the November 2019 issue, pp. 30–32.