Yesterday, news dropped that digital art collector and funder Seedphrase (aka, Daniel Maegaard) Price Per Stemthe first commercially available dataset of British artist Anna Ridler, released by Avant Artecompletely. If you’re not steeped in the concept of dataset-as-art, you probably haven’t been following what’s happening in the wider world of technology-driven art (taking a spin at the Lumen Award finalists and winners can help you catch up). For several years, the rank and file associated with technology-driven technology with green NFTs; more recently, the term may bring to mind Refik Anadol’s ever-expanding AI-powered data diagrams. But in reality, it’s much broader than that, including everything from the early production work of artists like Michael Noll, Béla Julesz and Frieder Nake to Miriam Simun capturing and automatically generating the scent of the nearly extinct flower Agalinis acuta using chemical data.
Ridler, like many other artists working at the intersection of art and technology, has chosen data as his medium. He is acclaimed for his use of artificial intelligence and machine learning to create rich visual and conceptual works that explore the biases, histories and ethical implications embedded in these technologies—all built on carefully created datasets with which he explores human intelligence, technological failure, ontology. , environmental processes and biases. His work has been featured in prestigious exhibitions around the world, including the Barbican Center and the V&A Museum in London, the Center Pompidou in Paris and Ars Electronica in Austria.
Myriad (Tulips)installation work from 2018, consists of a dataset (or training set) of 10,000 tulips captured and labeled manually by Ridler. That was the foundation Moses’ germAI-generated video work using his dataset, and 2019 NFT work Bloemenveilinga tech marketplace where people can buy GAN-generated short videos with self-destructing tokens. Together, these three works explored the work that goes into AI training programs, the relationship between human intelligence and machine output and modern cryptocurrency speculation, linking it to the Tulip Mania of the 17th century.
Seedphrase’s latest acquisition, Price Per Stemit is actually a series of twenty-five works of art with both physical and digital elements: printed photographs of real peonies taken by Ridler and animated videos of the flowers dynamically described using a complex model of peony price fluctuations. During the acquisition, the Observer caught up with the artist to learn more about her exploration of how data input affects aesthetic outcomes and her thoughts on artistic integrity in a cultural landscape shaped by increasingly independent creative forces.
Your practice includes manual data sets—are you particularly interested in how human choice shapes outcomes or is there another reason you chose laborious maintenance over automation?
Part of every project is always creating a dataset—a collection of information that can be read by a machine—as it is an essential part of machine learning. A big part of my process is building things from start to finish. Without working with the dataset in some way, I won’t be able to work for everything. The images and results I create from the data are my own and come from the way I saw and saw the world; mistakes and choices are mine and mine alone. Each photo and caption has been painstakingly crafted to reflect this. That’s why for me, a data set is a creative work and becomes a work of art in itself. I’ve been working this way since I started making art with machine learning almost ten years ago.
There is also something a little absurd about doing this long process. Making this piece feel very strange, especially buying hundreds of pounds worth of peonies just to take one photo of them every few days—that’s the exact opposite of how data sets are made for commercial reasons (or connecting to the internet without even looking at the medium—just relying on any glimpses were nearby—or harvesting data) by considering each flower a little and allowing it to die naturally and decay. The cost of everything is indirectly mentioned in the title of the piece.
The peonies were a deliberate choice; they have an incredibly short season (this project started in April to capture it) and they have a stock in the flower market. However, their price is changing. They require very specific conditions to grow (including some time in the frozen ground in winter), which due to climate change has changed where they can be planted (Alaska, for example, has become a major producer of peonies) and this year the season was very short and just started which affected the amount and the kind of flowers I could get.
Because my data sets are relatively small, it also means that there will be a lot of slip-ups and mistakes, which I welcome—most AI production has a shiny, smooth quality to it, which I’m less interested in when it fails.
When it comes to intelligence, do you see any hard stops where automation or AI is concerned? Like, if an artist does X or Y, has he crossed the line where he’s no longer an artistic agent?
It is different for every artist and work of art considering what their purpose is and what they are trying to explore. A lot of the debate about creativity and AI goes back to the debate and concerns about conceptual art in the ’60s (and earlier), and I think a lot of my favorite things come from this list.
Does your work comment/address the wider implications of technology on art and perhaps society, particularly ethical questions around data collection, ownership and consent?
Although this piece has features of a dataset to it, it is not just a dataset. A piece of animation, partially made with AI generating using a combination of many custom models I made for myself (over 50), drawing a little on the financial model I built to estimate the price of a peony using various factors (also using machine learning), partially covering the hand-drawn parts of those generated images (so that the piece reverses from analog to digital to analog before settling into this form hybrid) more obviously interested in trade, globalization and its inherent stupidity.
READ MORE: Cecilia Vicuña Receives Beauty of the Universal Order
By taking something as simple as a cut flower-which is sometimes seen as just a good thing or maybe kitsch-and telescoping to see how they are bought and sold and how they are sold and the dynamics of the cut flower market was something that I always had. you are interested in doing this piece; creating my own dataset was the way to get there. Because it is done using AI, there are echoes of how machine learning is increasingly used in market trading. It also questions the value of these fake eternal flowers that are sold and stored long after their real counterparts who were there to make them in the dataset have withered and decayed.
Is there a difference between a collector receiving a dataset-as a work of art versus other output produced using that dataset? What is the difference between a digital dataset versus a visual representation of a dataset?
For me, machine-readable datasets (raw images and/or text files or CSVs) are not displayed; which is always displayed as a representation of the dataset. This presentation wraps some of the decisions I made (eg, image, label) into one visual output. In the case of Price Per Steman artifact includes both a subset of a dataset and the output trained on that dataset. For me, a data set can be a work of art—though not always, depending on the artist’s intention and the way he chooses to work—as much as the output can be a work of art. Although again, not always, it depends on the purpose, so there is not much difference in my head: both are different ways of doing the work.
By getting the Seedphrase Price Per Stemdoes it mean that he could use the data set himself as a generation tool?
Having received the input of images, he will have to do a lot of work to make it into a dataset to use, scan each image, remove the handwriting, build his labeling system, adjust his model, get the idea he wants to work with this data. Of course he could do this, but there would be enough work involved that it would stop being my job and maybe his. I’m never interested in using AI to create a picture, I’m always interested in using it to think about different ideas.
What’s next for you—more data sets or something else?
I’m always working with datasets! I’m particularly interested right now in exploring the very large datasets that power LLMs, etc., and seeing if I can sort them out, moving a little bit from me teaching a machine what the world is to me trying to understand what teaching a machine sees.
