Generative Anesthetics

A eulogy for generative art

You spawn in from your bedroom closet. You poke your head out and look at the calendar. August 24th, 2021, it says. That’s not right… You run to the sink, splash cold water on your face. You look up at the mirror with a cool grin. You like what you see—no, now’s not the time! You glance down at your Rare Pepes x Hermes wristwatch: 4:69am. Go. Down the stairs. You stumble past the Fidenza hanging in your foyer—"foyay", you've recently started pronouncing it—and rush outside into the somehow brisk August night, down your porch steps. Your wife, kids, and dog bark at you in unison from the window. Keep running. Step after step, you hurtle forward into the dark night. A sudden whoosh; the world drops out from under you. You’re falling, falling, into the deep, dark abyss. The black is looming. Impact is imminent. You wake up.

Generative art has had a moment. Some might call it a fever dream. From certain vantage points within the cryptoart community over the last few years, it has looked like it’s just about all that’s out there. For the better part of that period, Art Blocks dominated collector discourse. Even as cryptoart chatter diversified, it predominantly spread to new platforms that also center around generative art: fxhash, Alba, Prohibition, Verse, Bright Moments, Highlight. What deviations there now are from this norm are almost exclusively focused on AI art. Art on the internet affords a vast array of weird and wild possibilities, most of which remains untapped in terms of commercial support. Instead it has remained locked in on a single genre of art and its concomitant commercial paradigm. By most measures, it’s been a great success. Cryptoart has developed its own vernacular. It has a set of stylistic norms, production techniques, and commercial processes all its own. It's an art economy and culture unto itself. A brand new collector base has been brought into the fold. Institutions at the pinnacle of the traditional art world have begun to take notice of cryptoart. This doesn’t happen without commercial success, and there’s an argument to be made that this commercial success has been engendered by the generative art paradigm. A lot of people love generative art. At times throughout the past few years it felt like everyone in crypto loves generative art. I am not one of those people.

A crucial thing to note is that I’m using “generative art” here and for the remainder of this essay as shorthand for “longform generative cryptoart,” i.e. algorithmically-generated art that uses randomness to produce large collections of work, usually in the hundreds or thousands, and minted onto a blockchain. Roughly speaking. I know that there’s a rich history of generative art that exists beyond this scope. For the purposes of this essay ever seeing the light of day, we’re going to have to mostly skip over that. If you have a terminological quibble, you can just mentally replace “generative art” with the above longhand anywhere else you encounter it below. Meanwhile I'm using the term "cryptoart" here to refer broadly to the art movement around NFTs, not art about or distinctly in the flavor of crypto. So basically any artwork that is minted.

Whatever you choose to call the genre just described, for a long time I felt alone in my antipathy toward it, which I mostly harbored silently. I didn’t have a well-articulated critique, and in any case I prefer not to critique publicly without a compelling reason to do so. Despite its ubiquity, I allowed for the possibility that I simply hadn’t engaged with generative art sufficiently to get it. If I would just read a little more about the super cool use of shaders in this project, or of the computer art titans it owes its origins to, maybe it would click. Alas. A couple years on, I’ve engaged with a whole lot more generative art, and now have more conviction than ever in my critique. 

Interspersing this essay with generative art I love to balance out the critique. Here, a grid of outputs from Maya Man’s Fake It Till You Make It.

Now is a good time to hammer home a point that’s at risk of being lost in my critique: I’m not flatout against generative art. Rather, I think that the tendencies it provokes, combined with the cultural and commercial climate surrounding it, make it as a medium tend toward being a locus of weak work right now. It’s a probabilistic thing. As I wrote in my recent piece on manglecore and the aesthetics of AI art:

“The medium isn’t quite the message, but it is a possibility space, and thus a probability cloud, of all that the message might be. Gasses take the shape of their container, and the message will conform to the affordances of its medium.”

Generative art comes with its own such set of affordances. Some of these promote good art. Others do not. I want to talk about a little of both. However, plenty has been trumpeted about the virtues of generative art. I decided to spill my thoughts because I’ve seen precious little serious critique of the genre.¹ I believe that good art matters, that hegemony threatens it, and that dissenting voices are instrumental to fighting that threat. I also hope that people like myself, who might feel dumb or tasteless for not “getting” the artistic vogue of the moment, will find some vindication in the critique. 

Some might protest that the gen art wave has already begun to flag significantly, that I’m beating a dead horse or kicking a movement while it’s down. Yet the movement is only flagging relative to the unsustainable pace of 2021 bull market frenzy buying, in which collections released from the same source on a weekly cadence sold out to the tune of hundreds of thousands of dollars from early career artists in mere minutes. That was insanity. Generative works are still selling, making debuts at art fairs of international repute, fetching millions at top tier auction houses, and bringing collectors new and old into the mix. It still dominates the cryptoart discussion relative to the scope of what’s out there, with AI work of course seeing a recent notable uptick in the conversation. Things are alright. To the extent that generative art’s dominance has dwindled, this marks the present moment as a meaningful juncture to speak on it. Ardent supporters of generative art may be open to new possibilities for cryptoart as a medium, or technology, or social movement, or myth. Detractors might recognize that the time for their thing has finally come. Though I’ve been taking notes for and thinking on this piece for over a year now, that reason—the opening up of these new horizons—makes now the right time to examine the shortcomings, strengths, and directions forward for generative art, and cryptoart as a whole.

The wailing laments from the generative art world over the last few months are telling. The market is down bad they say. It need some milk. Billowing negative sentiment within cryptoart at present largely revolves around work not selling or reselling. My critique then, like all things crypto, begins with the market.² Art is always a product of the social and economic milieu within which it’s situated. Likewise the art market, a corner of this milieu, has always played a hand in shaping the way art is produced. Cryptoart, however, takes this coupling to new heights. Crypto is a fundamentally financial thing.³ The tokens, wallets, and transactions on which this world was built are all outgrowths of a vision for a radically reconstructed financial system. That’s really cool, actually. But what happens when you build an art movement on top of financial building blocks? Unsurprisingly, it becomes deeply imbued with the financial character of its DNA. So too, then, with generative art (i.e. longform generative cryptoart, as a reminder), the most successful movement within cryptoart to date. Generative art, probably more so than any art that has come before it, must be understood as a product of the market context within which it’s situated. 

That market context is defined by the predominance of software. In any era, art is drawn toward the powers that be, as it seeks variously to harness them, to serve them, or to critique them. Over the span of history these powers have ranged from the church and the state to the academy and the aristocracy. Of course none of these have died out. Some are doing quite alright. But we now live in an age marked indelibly by the flourishing of technocapital (i.e. the generation and movement of capital through technological means), software being one of its primary modes of expression. For a while there, software art was almost entirely divorced from the software economy, the “techno” and “capital” kept apart. Digital art, for most of its history, was a playground for exploration of technology largely unencumbered by the weight of capital. Yet those two have a way of finding each other, and NFTs bridged that gap in a meaningful manner, bringing a native financialization mechanism to digital art. If NFTs wed digital art with digital economies, then generative art consummates the marriage. Art meets capital at conception in generative art—the transaction creates the artwork, per the standard method of random generation upon mint. If you’ll pardon the extended and momentarily lewd metaphor, generative art is then born with a kind of original sin, its genesis seeded from the Edenic fruit of capital penetration. One might argue that the transaction hash is merely a handy source of randomness, that I’m taking the symbolism too literally. But it’s not just symbolism—it’s a literal monetary transaction that creates the artwork. This is interesting! It’s incredibly elegant and native to the blockchain as a medium. It also means that the specter of finance is an inescapable element of the generative model as currently designed. Even in the case of a free mint, a transaction occurs and gas fees are paid, and your art piece doesn’t exist until this takes place. That means you haven’t even seen the work until after it has been bought. The payment is primary. There’s some precedent for this in the traditional art market, where collectors will buy a piece sight unseen because their dealer recommends it. But this practice is typically referenced to point out the egregious financialization taking place in the contemporary scene, a hyperbolic exception to the rule. This is inverted in generative art, the exception of buying the work blind now instead the rule. 

After purchase, the work is sent to your wallet, which constitutes your identity in this realm, standing in for you as a person.⁴ The wallet is the new identity primitive in the onchain economy. Again, this is pretty neat and makes great sense as a technological and even political solution—a unified, cross-platform, low-friction unit of identity articulated in modern society’s lingua franca, the language of the market. But the implications for the conjoinment of art and finance remain.

A piece from Emily Edelman’s Reverie

However I haven’t actually said anything about why this flavor of entanglement between capital and cryptoart is bad, just that it exists. For the anti-capitalist, the critique is implicit in the connection—anything touched by capital is tainted, guilty by association. I’m not of that persuasion nor do I expect my reader to be. But even across ideological dispositions there tends to be a distaste for money infiltrating art to too great an extent, a hope in art as one of our last pure things, tantamount to sacred—valued for the encounter with it per se and not for any extrinsic motivations like the accumulation of wealth. Such hopes held too preciously veer into naivete, but they can be viewed on a spectrum, and the above example of buying work sight unseen falls well into the territory of the perverse for most.⁵

But even excusing this, the economic incentives that inform the generative model far too often pressure a sacrifice in the quality of the art. This trajectory developed in a two-part process. The first step was moving art natively into the digital realm, where its reproducibility as software gives it the propensity to mirror SaaS and digital entertainment businesses with their near-zero marginal cost curve. The digital artwork can be distributed ad infinitum basically for free. However, instead of being hyperfinancialized, early digital art was radically de-financialized because while reproducible, it lacked the means to capitalize on this reproducibility in any economic form. Who wants to buy a JPEG that anyone can right click save, amirite? Enter crypto. NFTs not only reified digital art, conferring unique identity to works by means of tokenization and giving them a thingness; they also embedded digital art in a market structure that is consumed with fast-paced, transparent transactionality. This put the mimetic psychology of the art world, the logic of the social animal, on steroids and an Adderall XR. Digital art not only could be bought, it would be bought in a multiyear frenzy.

These dual processes, digitization and tokenization, mark all NFTs, but a third factor sets generative art in a market category of its own. Where the means of distribution of an artwork become digitally native via the reproducibility of digital files, the means of production become digitally native thanks to the algorithmic production inherent to generative art, where new artworks are created at the mere click of a button. The former changes how work is shared. The latter changes how work is made. What follows, then, is a mass consumer artform. By enabling fast, large-scale production, generative art and the market around it incentivize the creation of cheap, diluted work—conceptually barren and aesthetically milquetoast, propped up by technical jargon that conjures a mirage of depth. It is assembly line art. I mean, the infrastructure upon which so much of it is produced is even called “The Engine.” But the assembly line of generative art is one finely tuned to an individualist world. Each piece that comes off of it has the minimum viable difference to make it feel uniquely yours.⁶ This appearance of an individual relationship to the piece, however, is a hollow construction. A meaningful relationship with a work is predicated on personal identification with its form/content or the artist/story behind it. Generative art denies the prospect of this kind of relationship through its arbitrariness (the work you own is randomly generated) attempting to recreate the character of such personal relationships through structure alone. The “unique” nature of the work is ascertained by a token number and a random hash used to generate its appearance. Its personal value is instead signified by the exchange of coin. The formal husk of a relationship is left in place while the color is drained out. It’s merely structural, a form emptied of meaning. A common response to this experience for the collector is, after minting, to convince themselves of how much they love the piece they specifically minted. They find all sorts of reasons to identify with it—“wow the flow fields are beautiful” or “no way, green’s my favorite color!” Identifying with the work would be a lot easier if we could just choose the ones we want to own in the first place, but that would require real discernment.

This flavor of collecting is a symptom of what I call mass individualism. Ostensibly oxymoronic, the term represents the parallel movements toward a globally systematized world and a world in which the primacy of the individual triumphs over all—in which the person is more independent than ever yet paradoxically relies on unprecedented levels of human cooperation for this independence. We can live in the solitary isolation of our homes, cut off from society with everything we want and need delivered to our screen and doorstep, but only thanks to interpersonal labor coordination at scales never before seen. In this mass individualism we try to have it all—the assurance of our own idiosyncrasy and agency on the one hand, with the perks of society and community on the other. The logo on your shoes signifies membership in the tribe of Nike, but that specific colorway is so you. This is quite effective as a commercial model for consumer culture, where industrial systems can produce goods at scale with the minimum viable difference needed to offer a veneer of personalization. Yet the individuality this offers is an impersonal one, and its communalism a shallow one. The system is doing the least it can get away with doing, naturally—it operates in a spirit of optimization. The problem is that we’re left with the mere facsimile of what we as humans truly crave—agency and distinction in the individual condition, deep interpersonal connection in the communal one—and that that facsimile is sufficient to keep those cravings at bay. Those cravings aren’t gone, mind you. Such means will never satisfy the underlying longings. They’re just pushed down as we convince ourselves that the next round of surface-level satiation will finally do the trick, keeping us on a merry-go-round of existential dopamine hits. This illusory promise of fulfillment is the foundation of consumer culture.

The generative model recreates these dynamics of mass individualism in the domain of art. Automated systems produce work at scale. Generative randomness renders each piece a unique “one of one of x.” Collectors become members of a community of other collectors, yet maintain notional individuality through “token identity.” And as the structural dynamics of consumer culture are recreated, so too are its problems. These dynamics encourage artists to overproduce. If you increase the supply you can increase profits. If you increase the trait variation you can justify the supply increase. It’s a slippery slope. Suddenly the art is changed, but you have a story to tell yourself about why these changes were fitting and totally justified—nay necessary!—for the work. This is almost always cope. It’s fluffed up financial padding that dilutes the punch of the artwork. The financial opportunity created by speculative mayhem and a relatively inexperienced collector base further pushes artists in the direction of these incentives. The design of this ecosystem nudges us toward indulgence of profit-making motivations at the expense of the art. These pressures are difficult to resist, and equally difficult to see at work in oneself. Generative artists may be working with algorithms, but they’re only human after all.

You might ask what’s the harm in producing more work if the art lies in the algorithm itself, as many argue, and the algorithm has already been set? Producing more work increases accessibility and democratizes art, right? There’s a subtle circularity here, lying in the fact that the artfulness of the algorithm rests in its ability to produce a diversity of outputs at a specific collection size. Fidenza is acclaimed because it produces varied and interesting outputs over the span of 1,000 works. At 10,000 it would grow redundant. I doubt Tyler Hobbs would object to this claim. In fact, judging from his writing on these matters, I’m guessing he’d emphatically support it. And yet given the effortlessness of producing 10,000 works instead of 1,000 (merely typing an extra zero), the temptation to create work ill-suited to its own algorithm can be irresistible when met with forceful enough market pressure. Who among us hasn’t seen a collection pushed well beyond the size it supports? We can only understand generative art collections by looking at them in their entirety, and the diversity or redundancy of the works in light of collection size is critical to our appraisal. They cannot be understood as algorithm alone, but need to be considered as a system of algorithm and outputs.

Screenshot of a work from Jan Robert Leegte’s interactive Web collection

A related feature of generative art is the fact that to produce within these structures you have to create the type of thing that can be reproduced en masse. Powerful and idiosyncratic things tend to elude such reproduction. Generative art thus traffics in what’s replicable. One result is that it’s all too often trying to model via computer systems that which we’ve already produced by human hands, amounting to a rehash of things that have come before, different only in method. Perusing generative art collections and their abundant geometric abstractions can often feel like deja vu for various periods of art history from the last century or so. Insofar as art should convey new and moving visions of the world, this is a failure. Generative art is not incapable of this. Maya Man has done it. Deafbeef has done it. Siebren Versteeg, 0xFFF, and Mathcastles have done it. Emily Edelman, Stevie P, Jan Robert Leegte, and Lars Wander have all done it. But the bulk of generative work fails to meet this standard. 

The reproducibility that makes it so banal is by equal measure what has granted generative art such great success over the last few years. It’s immediately parseable both as code (by the likes of the Art Blocks Engine) and cognitively (by humans). Collections fit a mold. We are thus instantly familiar with each. We know the parameters, we can evaluate, compare. It’s formulaic. We can become a very limited kind of expert in a very limited time, which of course is all we can spare these days. You look at the preview image. You read the collection description. Or maybe not. You scroll through a dozen or two of the sample outputs. You get the idea. You yawn. You buy a piece. It’s selling fast, after all. Fuck it, you buy six. You smile mildly. You share your outputs in the Discord, some Discord. You tell your wife about it. She doesn’t get it. She’s happy for you, or worried about where all the family money is going. It’s good babe. Things are good. 

Contrast this relationship to the artwork with something like Heather Dewey-Hagborg’s Stranger Visions, a generative collection of a quite different kind which can’t be compared along set axes to really any other work I know of, much less dozens of collections. What this formulaic quality means is that the generative art projects of this ilk all blend together. Produced by the same Engine, they are themselves all outputs of a single algorithm in a way. And maybe that’s how this era of generative art (c. 2020-2024) is best remembered. The constraints of the genre are set, and the generativity of the individual imagination works within them to produce a new idea for a collection. The differences that would usually exist between pieces from a single art collection are now of a similar size to those between entire collections. Perhaps then, the Art Blocks Engine, that infrastructural behemoth, is the great work of longform generative cryptoart.⁷⁸

Generative art’s problems go beyond these structural, market-oriented critiques, but they all have their origins here. From this broader umbrella we might funnel them down into distinct but inevitably intertwined shortcomings along three key axes: the technical, the conceptual, and the aesthetic. The latter two find their roots in the former—the technical—around which generative art’s identity is based. The medium (as scoped for the purposes of this essay) is inherently technical, defined by the processes it employs to produce its artworks. This is true to some extent of any medium (painting and sculpting involve much technique), but generative artists seem especially insistent on bringing it to the fore, highlighting details about GLSL, Perlin noise, Bezier curves, and voronoi tessellation (I don’t know what I just said either). Process can inform the meaning of a work, but it shouldn’t replace it, and generative artists have a bad habit of over-indexing on these technical features to the exclusion of pretty much anything else of substance. Doing something complex and creative with code can be quite difficult, and that difficulty can sneakily enamor one to the output in a way that the output itself doesn’t merit. The generative artist can lay a turd and be rather impressed with himself, because he did it with code. Yet the code isn’t immediately visible to the viewer. All they see is a pile of shit. In painting, the brushstrokes stay behind as a visible artifact of the process. The gesture and the mark are, in semiotic terminology, indexical. An intentionality is revealed. We see the artist’s hand in the work. With generative art, the gesture and the mark are separated. The artist’s hand is obscured, and so its process becomes more challenging to communicate. The process is decoupled from the art object. Pixels on a screen don’t reveal keystrokes in the same way that brushstrokes reveal the movements behind them. Because we can’t infer the process behind the pixels the way we can with brushstrokes, generative artists take it upon themselves to explain it to us. And we follow along, dazzled by the esoteric technical details into thinking that something of import is being shared. The complexity leaves us with the impression of meaning and obscures the work’s shortcomings both to the artist and the viewer. In doing so, this technical infatuation contributes to those same shortcomings by propping the work up and giving it space to get away with them. Consequently, a hallmark of generative art is the artist falling in love with the tech at the expense of the art. Per Goethe, “Technology in alliance with bad taste is the enemy of art most to be feared.” Tyler Hobbs has tried to impress this fact on fellow generative artists:

“As generative artists, we think we have such a cool story to tell about how we make our work! It makes us sound smart! We can do weird things with computers that go right over most folks' heads! Maybe we use an esoteric tool set, or sophisticated math or programming techniques. The viewer does not care…Our technical prowess may be good for impressing other programmers, but even these folks, in the moment they view the artwork, don't care. They only know if the image speaks to them. Our technical abilities are only important so far as they allow us to create images that speak to us. They are enablers, not a source of value on their own.”

Again, process can and should contribute to the meaning of a work, but esoteric technical details aren’t a stand-in for a real concept. This problem is pervasive in the genre, and conceptually, generative art collections tend to say little of substance as a result. The typical fare involves an algorithm born of a marginal technical novelty (generative audiovisual something or other! shaders!) and passable visual appeal that then has a tiny dab of concept slapped on post hoc to construct a facade of meaning. Usually this involves something about the “dynamic interplay between order and chaos” or “the emergence of complexity from simple algorithmic rules.” The works are algorithm-first. Meaning, the real human impact of the artwork, is with frustrating frequency only an afterthought. In place of meaning come technical explainers that fill the void where concept should be. The result is a deadening effect, more art for computers than computer art. Taken up as a self-conscious gambit, a project that creates “art for computers” would actually make for a potent Kafka-esque critique of a society bogged down by systems made nominally for humans yet which end up leaving those same humans behind. Yet this is not the thrust of the generative oeuvre. 

From Lars Wander’s “how you see me”

A valid line of protest might go that art doesn’t need to say something, it can just make you feel something, be beautiful, move you. Pero por qué no los dos? Why shouldn’t we demand that the art we champion and spend great heaps of time, money, energy, and infrastructure on be moving/beautiful and meaningful? This line of reasoning also elides the fact that rarely does generative art truly move or shake us with its beauty. Instead, we far more often get momentarily entertaining pretty pixel arrangements that make for fantastic feed fodder, easy to look at as we scroll infinitely through the thousands of variations that their algorithmic nature yields. We are served up a conveyor belt of endlessly stimulating inanity. We’re superficially entertained but bored in a much deeper way—spiritually bored, existentially bored, bored in the way that only infinite entertainment can yield. The repetition is soul-crushing. 

A recurring visual motif across this repetition is geometric abstraction, and the genre finds itself constantly retreading territory already well-explored by its forebears in Modernist abstraction and the significant history of generative art that preceded the advent of the NFT. Most everything conceptually and aesthetically interesting done in computer art was done well before the NFT wave. The bulk of what we’ve had of late is a monetized rehash. Comparisons to the “zombie formalism” era of the 2010s are apt, the term perhaps even more fitting here than in its original usage. It describes a repetition of 20th century process-based abstraction that sleepwalks in the footsteps of its pioneers, failing to contribute ideas of its own beyond “do the same, but with code.”

On these grounds generative art fails us aesthetically as well. The norm is work that’s visually fanciful if we’re lucky, yet in the end uncompelling. Work that’s dangerous or difficult are rare finds. Generative art is often pretty, but pretty is cheap. As fellow artist Sten put it, the genre is full of “decorative art, challenging no one.” The infamous feud between Jerry Saltz and Refik Anadol over the massive MoMA installation of Refik’s work Unsupervised was waged upon these grounds. In quite memorable if somewhat cruel fashion, Saltz’s critique variously deemed the piece “screensaver art,” “a glorified lava lamp,” and “narcotic pudding.” “It’s comforting, really,” he went on, “Unsupervised has the virtue of not disturbing anything inside you; it triggers no mystery.” Though I never got to see the work in person I think Saltz was probably a bit harsh—the thing looked, at worst, awe-inducing—but the substance of his criticism could equally be applied to the majority of generative art of this age. Echoing poet Cesar A. Cruz’s quote that “Art should comfort the disturbed and disturb the comfortable,” Saltz points to an endemic lack of aesthetic or emotive depth or impact in generative work.⁹¹⁰ Rarely does it disturb or challenge us. What comforts it offers are shallow, like a topical anesthetic. And an analogy to anesthesia isn’t too far off. For the last several years, generative art has lulled us into a kind of anesthetic stupor. It has served as a palliative, a numbing agent that clouds our vision to all that exists beyond its veil. Little else matters. Yet the numbing deadens our spirit. The aesthetic and emotional responses it educes are blunted. Its conceptual provocation is minimal and trivial. The elements of art that are most human are left behind. The title of this essay takes after this thought (an-aesthetics, get it??), and the two generative collections I released recently were pointed in this direction. Catalog, consisting of “generative Duchampian readymades” was an indictment of the current status quo in generative art as a cheap mass consumer artform.¹¹ DVD Screensaver was “a meditation on the place of art in a culture of infinite entertainment—screensaver art for the screensaver art era.” Both critique generative art by parroting certain elements of its character and drawing them out to comic hyperbole. Pointed as they are, the intent in each case was constructive discourse rather than all-out condemnation, because] despite all I’ve said over the last several paragraphs, I do still feel there’s much worth preserving about generative art. 

Generative art offers the prospect of digital objecthood and ownership of art. It democratizes participation and ownership in art through digital scalability and the global reach of the internet. It introduces randomness, seriality, and systematicity, which can be used to reflect conditions of contemporary life, as formal principles in its artworks. It creates inherently networked assets. It gives voice in fine art to the digital culture and technology of the present. Its enablement of capital, though challenging in plenty of ways, opens a path to commercial and institutional legitimacy if we so desire. If done right, it can inform us on the relationship between human and machine. Some of these it inherits from cryptoart or digital art more broadly, but some are uniquely generative art’s own. Maya Man’s Fake It Till You Make It uses the generative paradigm’s randomness and repetition to conjure a nonsensical simulacrum of online self-care culture. It’s what we might call “hyperpop art.” Stevie P’s Dopamine Machines similarly harnesses a hyperpop aesthetic with a maximalist bent to capture the neurotransmitter-warping chaos of online existence. Emily Edelman’s Reverie employs generativity to echo the fragmented and meandering yet somehow still meaningful speech patterns brought on by a loved-one’s dementia. Andreas Gysin, Kim Asendorf, and DEAFBEEF, across their respective bodies of work, all feel like they speak the language of the machine and compellingly translate it to something still strange yet human-interpretable. Generative art can be good, human art. And digital art can go well beyond the bounds of the (longform algorithmic) generative paradigm. Look at OONA. Look at SHL0MS. Look at Ana Maria Caballero, Margaret Murphy, Mia Forrest, Operator. More of that please. Create more of it. Collect more of it. Talk about more of it. Champion more of it. The slouching cryptoart market gives us a moment to step back and look in the direction of this kind of work. I’m hopeful that we’re heading there. Until then, generative art’s greatest importance might lie not in the work itself, but in the fact that wide swathes of our cultural niche were enraptured by it for a moment, and what that says about our age—about this market, about this culture, about us. Yet we’re more than that. Now we have an opportunity to look in the mirror and decide what to say in response.

1. The whole cryptoart scene is lacking in serious critique, serious bag shilling taking its place instead. On the one hand, critique is important for constructive development. On the other, I do value the spirit of positivity in the space and would like to find a way to put it in service to this constructive development.

2. I'm saying this partially with tongue in cheek.

3. "Thing" in italics because it's hard to pin down what kind of thing exactly crypto is. It's more than a technology. It's a new financial system, but more than that too. It's a sociocultural project and network, yes, but those are rather fuzzy titles. A "digital sovereignty" gets at something meaningful. It's a new kind of reality, albeit one of limited dimensions, or at least offering limited space we four (three?) dimensional creatures can operate over. Yet there's something undeniably spatial feeling about it. Temporospatial, even. Spatiotemporal? No, temporospatial. It's most easily compared to the internet, which I'm still not sure we have an apt word to categorize, perhaps because the internet was the first of its kind in this regard. Now with blockchains, which enable us to see them + the internet as a class of things rather than either a lone anomaly, we could probably use a word for that class of things.

4. Of course the wallet bit is true of all cryptoart, not just generative art. Many of generative art’s challenges are those it inherits as a subset of the broader medium.

5. From a certain viewpoint, this seeming perversity can be flipped on its head. The engaged gen art collector, having perused sample outputs, is submitting to the generative spirit of the work, placing faith in the artist to have competently designed their algorithm to produce a strong, diverse array of coherent yet meaningfully unique pieces. It’s in some way actually a culmination of the generative paradigm, where not only is the artist’s hand removed from the act of creation, but the collector’s is taken out of the act of selection. The generative randomness pervades the entire lifecycle of the work, bringing the genre to the full fruition of its ethos.

And yet from another angle, this ethos is the epitome of modernity’s effort to remove humanity from the literal picture by reducing the world to atoms and bits. Generative art advocates for systems that, though created by humanity, continue to thrive in our absence. It is the artform of an anti-humanist project. We don’t need more big systems that remove the human hand. We need more life. Maybe the above paragraph is then an accurate yet damning account of generative art more broadly, beyond just its crypto confines—that the structure of traditional generative art implicitly houses certain ideological commitments; that longform generative cryptoart is their natural conclusion; and that their conclusion is not good. I’m steelmanning an argument here that I don’t know I fully believe. I’m laying it out just to share the thought and start the conversation.

6. I'm playing off of the startup-speak term "minimum viable product" (MVP) here to co-opt the spirit of optimization that generative art economics share with it. By minimum viable difference I mean the least amount a product or art piece can differ from another and still feel like its own distinct thing.

7. I don't think the medium is actually as narrow as it has presented itself, but we've entered a creative "mode collapse" that has functionally confined it within these bounds. Mode collapse is a term used to describe GANs (a kind of AI image generator) getting stuck in a rut producing a single type of output which is non-optimal or incomplete yet sufficient to meet the criteria it was given.

For a little more detail, GANs (generative adversarial networks) consist of two different neural networks, a generator and a discriminator. The generator tries to convince the discriminator that its outputs are a certain class of thing that it has been trained on, while the discriminator tries to sniff out fakes. The two then develop in tandem in a sort of competition. Mode collapse occurs when the generator finds a single type of solution that consistently bypasses the discriminator yet which doesn't represent the full diversity of the target class' data distribution (e.g. only producing quokkas when the target class is all animals).

8. The multiple references might make it seem like I’m shitting on Art Blocks. I’m definitely not. I think they’ve done incredible things for the space. Everyone I know or have met from their team, Snowfro included, is kind, and I consider them many of them to be friends.

9. This quote is often misattributed to Bansky, who did in fact use it, but after Cruz. It appears to have its origins in a similar quotation spoken a century prior about newspapers.

10. Refik's work involves the use of AI, so Saltz wasn't speaking strictly to generative algorithmic work, but the point remains.

11. The Duchampian gambit was two-fold. With the digitally-native readymades, of course, but also with the bare sterility of the concept and aesthetic output serving as a counterpoint to the precedence in generative art of what Duchamp would call “the retinal”: pretty pictures that lack conceptual depth. Though remembered as dryly conceptual (which he frequently was) Duchamp’s real (or at least stated) intention wasn’t to make art devoid of visual aesthetics but to take an art world that he felt had become dominated by the retinal and return it to a balance with the conceptual. Most would say he overshot a bit, though he probably couldn’t have anticipated the magnitude and endurance of the effect of his work. With Catalog and DVD Screensaver, my intention was to make a dryly conceptual point as a counterbalance of this same kind, and then return to a more “consummate” approach that brings the conceptual and the retinal into harmony with my work going forward. In Utero I think struck this balance well. I may go back to that place again, but only to the extent that it exists in balance in my body of work.