Gans And Nfts

When Christie’s hosted its first Art + Tech Summit in 2018, the topic was the blockchain. The second edition, in June 2019, focused on artificial intelligence. Blockchain and AI are two big, buzzy topics, and they have intersected in unexpected ways, especially during this year’s crypto art boom. Artists whose work uses generative adversarial networks (GANs)— algorithms that pit computers against each other to produce original machine-made output approximating the human-made training data—have turned to crypto platforms not only to sell their work, but also to explore ways of critically and creatively engaging the blockchain.

People who make creative work with AI tend to be self-taught, as artists or engineers or both. They’re drawn to new technologies and ideas taking shape at the margins of culture. There’s a provocative friction between the figure of the tinkering outsider and the reputations of AI and blockchains, in the popular imagination, as rapidly growing forms of technological infrastructure with massive resources invested in them, behemoths that are transforming the shape of everyday life by digitizing more and more of it. Artists who sell their work as NFTs have been criticized for contributing to an ecologically destructive, toxically libertarian culture; artists who make work with AI have drawn fire for normalizing the technologies that enable corporate surveillance and predictive policing. The artists who take up these tools despite the problems associated with them aren’t utopians. However, they see firsthand the reality that new technologies are not monoliths but evolving systems, rife with flaws and potentials.

The Nude as Data

A warped abstract computer-generated composition hints at the 19th century nude portraits on which it is based

Robbie Barrat: AI Generated Nude Portrait #1, 2018, digital image made with GAN.Photo : Courtesy the artist

The first work posted to SuperRare when the platform launched in April 2018 was the output of a GAN: Robbie Barrat’s AI Generated Nude Portrait #1. At the time, Barrat was making work with neural networks he trained on collections of nineteenth-century portraits. By selecting him as its first artist, SuperRare declared a particular orientation: if the early crypto art scene was dominated so far by cartoons and memes, by CryptoPunks and Rare Pepes, this new platform would support work engaged with advanced technologies and art history. The palettes and brushwork of nineteenth-century painting are almost recognizable in Barrat’s warped portraits, but his GAN’s interpretation of them yields grotesque results. The computer sees patterns of hue gradations and paint lying on the surface, and in its imitations of them the model’s painted flesh—arranged in configurations that only remotely resemble a human body—ends up looking mottled, partially rotted. It’s as if the computer wriggled through digital color like a worm through an apple. Portraiture begins to look like still life: the nude body approximates a fruit or flower that shows signs of decay as a harbinger of death. The forms of Barrat’s GAN works are alien and a little sinister, a reminder that machines often distinguish people from things as a mere matter of form.

Barrat subsequently minted more works in the same mold, including a set of warped nudes taken as stills from an AI generated animation. The swag bags at Christie’s 2018 Art + Tech Summit included cards with QR codes, each linked to one of three hundred unique frames from the animation he named AI Generated Nude Portrait #7. There was little interest in NFTs then, and most of the summit’s attendees tossed their QR codes in the trash. Only a fourteen were claimed at the time, and some others were identified and claimed later; these pieces now appear on Barrat’s SuperRare page. The rest are known as “Lost Robbies.”  Maybe the loss is for the best: it moved the works from objects of speculation into the realm of anecdote and legend. AI Generated Nude Portrait #1 has been resold twice according to the public ledger on SuperRare, most recently for an amount one hundred times greater than the nominal sum writer collector Jason Bailey paid Barrat in the platform’s first transaction. Disillusioned with the rapid commercialization of crypto art, Barrat has since occupied himself with making performances in the shooter game Counter-Strike and other projects that are difficult, if not impossible, to sell.

Tulip Auction

A series of screenshots of an online auction house for selling AI-generated tulipsd

Screenshots documenting Anna Ridler and David Pfau’s Bloemenveiling, 2019.Photo : Courtesy the artist

Bloemenveiling (2019), a project Anna Ridler made in collaboration with AI researcher David Pfau, invoked the seventeenth-century Dutch tulip mania to comment on the contemporary Bitcoin gold rush. In hindsight, it looks like a lucid prediction of this year’s NFT boom. Bloemenveiling, Dutch for “flower market,” was an online exchange for auctioning AI-generated digital tulips that were programmed to wilt a few days after purchase, like real flowers would. Ridler and Pfau built their own decentralized application on the Ethereum blockchain—a DIY Nifty Gateway—to make the sales. They also programmed bots to take part in the bidding and drive up prices, reasoning that today’s financial markets are influenced by trading algorithms, and an update of the tulip auctions of yore should account for this innovation. Bloemenveiling’s auction house is no longer functional. The digital tulips are dead, having deleted themselves. The project exists only in sceenshots. Like the tulip market of the seventeenth century, it’s history.

London-based Ridler is self-taught as an artist and programmer. Her education in English literature might explain the essayistic bent of her projects, which weave historical references with observations about the impact of machine learning on contemporary culture. Selling disappearing flowers for cryptocurrency is an exercise in the virtual and ephemeral, but Ridler involves herself in the materiality of digital tech. She builds the data sets that she uses to train AI; she has exhibited the thousands of hand-labeled images of tulips used to generated the flowers sold in Bloemenveiling as Myriad (Tulips), 2018, to draw attention to the labor involved in collecting this data. She programs GANs and codes smart contracts. For Ridler, making the tools of her art rather than using other people’s code and data sets is a moral stance, declaring her independence from big tech, and also an artistic one, a way of creating her own worlds to reflect on this one.

A Blockchain Mythos

A group of figures sit by a lake at dusk, framed by gilded polyhedrons

Still from Holly Herndon and Mat Dryhurst’s Crossing the Interface (DAO) I, AI-generated animation with text by Reza Negarestani, 15 seconds.Photo : Courtesy the artists and Foundation

For Holly Herndon and Mat Dryhurst, building community around the use of new technologies is just as important as experimenting with them. “AI is human labor obfuscated through a terminology called AI, and our goal is to use technology to allow us to be more human together,” Herndon said in a 2020 interview for A.i.A.  “Forget about the AI for a minute and think about how to make the laptop an organizational brain for upward of ten people to perform around—that’s a challenge but it’s never about the laptop. The laptop has great capabilities, but it’s always about the communication between the people.” This week Herndon and Dryhurst minted their first series of NFTs on Foundation. Titled “Crossing the Interface (DAO),” it adapts a 2014 performance at the Guggenheim Museum in New York. Each piece in the series is a short, AI-generated video that draws on documentation of the earlier work. The animations project the worried, meandering line characteristic of GAN art into the temporal dimension. The visual field is organized by shapes that move in repetitive, quasi-recursive patterns, creating a feeling of both movement and stasis. The first NFT is a tableau of figures seated by a lake at dusk. Gilded polyhedrons hover and quiver over them, and the scene seems to constantly withdraw from view, even as the figures remain still.

In all the videos of the series, a warm but somber voice reads a text Herndon commissioned in 2013 from philosopher Reza Negarestani, known for blending theory and fiction. His text is a fantastic, oblique parable about humans who deal with a predatory threat by finding new ways to share thoughts and feelings. “The people of the den have since developed tools for proactive search of sounds,” the voice intones in Crossing the Interface (DAO) VIII. “They have diversified the cache of intimations into an endless chain of associations.” The parenthetical DAO of the series’ title stands for “decentralized autonomous organization,” an administrative tool that uses a blockchain to distribute authority and responsibility across a group. Herndon and Dryhurst celebrate this new form of collectivity not through a reasoned argument about the blockchain’s advantages, but by mythmaking, by creating an intuitive emotional way of understanding it.

Crypto Art as Service

Three computer-generated faces, with warped, pixelated featured look out toward the viewer

Mike Tyka: from the series “Portraits of Imaginary People,” 2017, from left to right: hamidmansoor123, JoshuaSpence88, and chalizzTrt, AI generated images.Photo : Courtesy the artist

This month Hic et Nunc surpassed OpenSea as the biggest platform for selling NFTs, in terms of daily active users. It’s built on the Tezos blockchain, which verifies transactions with a proof-of-stake algorithm that is far less costly and energy-intensive than the proof-of-work protocols used on the Ethereum blockchain. This means it costs pennies to mint a work on Hic et Nunc, whereas an artist must spend hundreds of dollars to put a work on an Ethereum-based platform. What’s more, anyone can mint on Hic et Nunc—no invitation required. When Hic et Nunc launched in March, its popularity was driven by artists prominent in the AI scene. Memo Akten, known for his work with GANs, initiated the debate about crypto art’s ecological impact, and his peers took it seriously, possibly because they are already acutely aware of the energy expenditures their work requires. Rendering art with a GAN can take hours, even days, of processing by a high-powered computer. And so they flocked to Hic et Nunc.

One such artist was Mike Tyka, a researcher at Google and a co-founder of the corporation’s Artists and Machine Intelligence program. After joining Hic et Nunc, he minted images from his 2017 series “Portraits of Imaginary People,” which was inspired by the proliferation of social media bots that spread misinformation during the 2016 US presidential election. He used a data set of photos scraped from Flickr and borrowed titles for the portraits from Twitter handles of fake accounts, thus duplicating the image-gathering methods and naming practices of bot factories. Tyka’s imaginary people have bulbous eyes, cheeks that twist into lumpy planes, and yawning pores. The edges of their faces blend with the speckled voids they emerge from. By now “creepy GAN faces” is a commonplace of AI art; apps like Artbreeder let you make your own at the push of a button. But in 2017 Tyka took advantage of new technological developments, and exhibited “Portraits of Imaginary People” as the first hi-res GAN faces at that year’s edition of the new media festival Ars Electronica. Minting is a way of commemorating that, and making the works available for sale without transforming them into prints, as Tyka has done in the past.

Since joining Hic et Nunc, Tyka took up new projects responding to the blockchain. Terms and Conditions, as the text of the token declares, is an artwork that provides a service. Collect it, and it performs a backup for you on the distributed encrypted server known as the Interplanetary File by “pinning” files so they aren’t deleted by the automated cleaning system. Evolving Complexity is a fractal drawing designed to grow and change over time; the collection of an edition pings a script that makes the work bigger and more complex. It visualizes collecting as a collective activity.

How AI Sees Nature

A shiny 3D model of a fictional insect

Sofia Crespo with Dark Fractures Studio: {Specimen no. 2, Amylococcus}, 2020, AI-generated 3D model.Photo : Courtesy the artist

When Sofia Crespo first encountered AI as an artistic tool, she was attracted to the possibility of teaching computers to experience emotions and learning what trauma might mean to a machine. But she found more fertile material by turning outward to the natural world, and thinking about how the computer perceives and processes information. The images created by GANs tend not to parse objects the way human-made pictures do. They bind and blend elements that we think of as discrete, even though they may be more interconnected than we’re able to apprehend. In Crespo’s series of prints based on nineteenth-century naturalist illustrations, fish, fowl, and plants melt together as appendages of unwieldly wholes. The Argentinian, Berlin-based artist is curious about using AI not to represent the natural world so much as to raise questions about how humans represent it. Her “Specimens,” made in collaboration with Feileacán McCormick (aka Dark Fractures Studio) and minted on SuperRare, are fictional insects made with a GAN trained on real ones. Insects are easy to isolate and scan, though they operate collectively. Textured with another GAN, Crespo’s creations have shiny exoskeleton, patterned like snakeskin or a marsh, warping fluidly over their ridged thoraxes and horns. Each of the “Specimens” rotates in a blank void, eliding the isolating taxonomies of entomology and SuperRare.

An AI Dream of a Crypto Museum

White text on black announces a 'Doodlecoins Exhibit'; to the right, reddish curving shapes and black calliigraphic lines, all AI-generated, align in the form of Egyptian hieroglyphs

A spread from Helena Sarin’s CryptoGANComix, 2021.Photo : Courtesy the artist

Helena Sarin says she’s interested in “the edge of the goofy and the beautiful.” She started working with GANs in 2017, after encountering them through her work as a tech consultant, and as she experimented with them she drew on her background in clothing design and food styling. She feeds her own drawings to a GAN and alters the results, sometimes offering them back as further fodder for the AI. The feedback loop between Sarin’s hand and the computer yields beveled meanders, recursive asymmetries, and other forms within forms. She tempers the wild, weird AI abstraction by using it as ornament for drawings of flowers and fruit, or for portraiture that recalls fashion illustration.

Sarin joined SuperRare in 2020, but switched her minting activity to Hic et Nunc this year. She appreciates the peer-to-peer atmosphere of artists collecting artists; she compares it to Etsy. To flag her support for the community, she made TezosPunk portraits: pixelated faces, like CryptoPunks, wreathed with her AI embellishments. She released a series of editioned works under the umbrella title CryptoGANComix, and sent collectors a 14-page PDF gathering all the pieces in the series. The first spread is a self-portrait; the next is a striated black-and-white tableau of viewers contemplating murky canvases at a museum. The following pages depict other areas of the museum. A “Doodlecoins Exhibit” distributes curving reddish shapes and calligraphic lines across a papyrus-colored background. They seem ancient like hieroglyphs but also futuristic and alien. The “performance center” features “Massive Steal: a cryptofarce in bids and withdrawals,” and though the text evokes an auction, the imagery consists of organic folding shapes stuffed in gridded squares. At the end, the museum store promises “a fine selection of the perfect amulets for your crypto wallets,” and a large illustration of one such bauble. CryptoGANComix is a surreal blend of two environments for art, the museum and the crypto space, spliced together as if by AI. Sarin gestures toward the emergent, yet indeterminate context of her own work, while suggesting a way to collect and experience it.

Surviving Planned Obsolescence

Two images show the same hazy blue-green landscape stamped with different black text, stringing letters and numbers

“The strange thing about this letter was that even its nonsense had a meaning of its own; even its want of coherence made a kind of sense,” reads a passage of “The Art Critic.” Mario Klingemann authored this short story in collaboration with GPT-3, a sophisticated neural network that responds to text prompts by producing reams of prose hard to distinguish from human-written text. Klingemann put the story on Hic et Nunc shortly after the platform began accepting PDFs as a mintable file type, adding text as a medium for NFTs alongside still and moving images. Klingemann, an early adopter of Hic et Nunc who helped proselytize it among AI artists, also minted several of his older GAN portraits, eerie digital faces in a gothic style he pioneered. But he also litters his account with jokey ephemera of the crypto art scene. There’s a web comic summarizing how the Bitcoin, Ethereum, and Tezos blockchains have responded to concerns about the environmental impact of crypto, and a screenshot of a notification that Beeple followed Klingemann back on Twitter. Not for sale, this NFT is tagged “trophy.”

Based in Munich, Klingemann has consulted major museums and libraries on digital archiving practices, and he has run DIY makerspaces and project spaces. He moves fluidly among contexts, bringing a tinkering trickster spirit to all of them. Perhaps the work most emblematic of his approach to the crypto space is Planned Obsolescence (2021). It began with a gauzy, blue-green GAN-made landscape of a forest’s edge, released as an edition. As multiples were collected, a script Klingemann wrote stamped the remaining pieces with the addresses of the collectors’ crypto wallets. The more the work is collected, the harder it is to see the image that is purportedly for sale. The blockchain is a touchy system to work with, and scripts break easily. A few weeks after its launch, Planned Obsolescence needed to be patched, and now it’s out of order, in limbo. But the work is generative in other ways: Klingemann’s gambit inspired Tyka’s aforementioned Evolving Complexity as an opposite endeavor. Instead disappearing as it’s collected, Tyka’s drawing gets bigger. Both treat blockchain as an open field, highlighting possible exploits in a changing system.

Published ByArt In America


Share and Enjoy !


Leave a Reply

Your email address will not be published.