THE LENSA DEBACLE
Sweeping the internet over the last several days… your friends as vaguely airbrushed, digitized astronauts, fairies, psychonauts, anime hotties etc etc. This glut of new profile pics was shortly followed by sweeping condemnations.
If you are reading this, you have likely seen some of the posts and comments, outlining specifically why Lensa is problematic, so I will not wade into the murky depths… Long story short, Lensa draws from the Stable Diffusion model, built as a non-profit, and is using the work of copyrighted artists, sans compensation, to generate these renderings.
Before I knew any of this copyright kerfuffle, I noticed a few things. The first handful of Lensa sets I saw were neat, after I saw a few more, they were plagued with a kind of sameness. Classically masculine features are distilled into astronauts, psychonauts, and concept art for videogames. Classically feminine and more androgynous features are distilled into faeries and anime babes. All faces also produce traditional portraits, mostly voiced in digital imitation of conte charcoal or guache paint textures.
The relatively limited aesthetic range of this Lensa output quickly reduced the whole thing to another viral social media moment, and it seemed like by the end of day 2, people were issuing sheepish, apologetic disclaimers for participating; but sure, who doesn’t want to see their own digital psychonaut self smoldering back at them?
My first real criticism of the Lensa fad, particularly after I realized people were paying for these images, was ‘oh great, we are training past AI art’s last great tell.’ I’m sure you’ve noticed, but AI art is pretty bad at faces.
With the exception of thispersondoesnotexist.com (which only renders photographs, which might be even creepier) most of the popular AI image production engines kind of don’t know how to make a human face yet. They are weirdly melty, somewhat too smeared to be a face, or weirdly pinched and horrific. Since the ‘arrival’ of AI art with the DALL-E trend this summer, I have found the approximations of human faces to be the one great skill gap that will giveaway an AI made it almost every time.
I was bemused to realize I knew so many people who had paid $4 or $8 to help train AI art past this skill gap, and I would not be surprised if in 6 months or so, their digital likenesses are trying to sell me shit. I didn’t purchase the Lensa slate, so I assume I’ve bought myself a little bit of time before my digital self is shaking my supple digital ass in my flesh face saying “you would deserve love if you bought yourself some of these fucking MeUndies.”
So many of these neat social media gimmicks are rarely as benign as they seem, for example, see Facebook’s controversial facial recognition system, or some of the ‘which character are you?’ quizzes harvesting data for Cambridge Analytica to influence the 2016 election.
And that brings me some somewhat clumsily into the much larger debate that people are really talking about in their disavowal of the Lensa moment and AI art at large, what are these things going to do to the creative economy?
ART IN THE INTERNET ERA
On the internet and social media (it has become impossible to talk about one without inherently speaking about the other) there is sort of a pandora’s box quality, if somebody made it and published it, somebody else will be able to recreate it, even if the original maker determines it should never have existed. This is generally true with any given thought object in cyberspace, too much resurrection. Now that we have seen the nascent arrival of AI Art, with output that primarily seems to parrot extant art, many feel that the writing is on the wall.
The internet and particularly social media have irrevocably changed the arts economy. In this moment, its hard not to feel that the change is grim, and has been for the worse, but this is shortsighted. The internet has lowered and removed, physical, monetary and temporal barriers for artists and viewers appreciating work in a way previously unimaginable. This means that anyone with an internet connection can sincerely engage with arts movements, arts moments, and the work of individual artists.
As a personal example, discovering the work of the Dadaists and Surrealists were crucial junctures for my own arts journey. I grew up in Longmont, Colorado, I graduated high school in 2007. It is staggering to consider how limited or delayed my exposure to these ideas might’ve been, or the limitations my local library would’ve presented, were I born even 15-20 years earlier, let alone 50.
The ubiquity and availability of art and creative work on the internet has also harmed the arts economy. As someone who has been involved in the arts for 15+ years, I can’t help but observe how I have both benefitted and suffered from this. I have seen, enjoyed, and been influenced by art both historical and contemporary, largely for free. This has also meant that the vast majority of people who enjoy my artistic output do not pay me any money to do so. Social media has made art essentially a free public utility, which most people take for granted.
Current social and public models have reiterated this in the Flesh World. In my cozy ~170,000 pop. city, free public art manifests itself in the guise of large murals, painted electrical boxes, painted pianos, and free concerts in public spaces. I am not against these programs, they have provided meaningful paychecks for many hardworking, talented individuals. However, they only fund certain kinds of public art, both in terms of medium, and also in terms of a relatively limited aesthetic scope, while reinforcing the impression that art is free. As rent and real estate costs increase sharply every year, working artists ask themselves ‘do I change my art to improve my chances of capturing some of this piece of the pie? Do I exert my focus on a more dispersed aesthetic-specific market I’m going to capture on the internet? Do I relinquish my expectations for what % of my livelihood will come from my creative endeavors?’
These questions set the stage for the looming spectre of AI art to be pure late capitalism menace. And yet, AI art is far more than the DALL-E grid or Lensa avatar. AI art is a giant umbrella, but given the share of attentive capital these viral trends capture, they eclipse a more meaningful conversation.
Before I go any further, I need to lay my cards on the table. I do not make AI Art. I do not make digital art. I earned a Bachelor’s Degree in theatre 10 years ago. I make money from art, but I do not make my living from art. My creative life has been an indelicate (at times turbulent) balance of performing arts practices (theatre, short film, stand up, sketch, clown) and visual arts (relief prints, zines and comics making). As a point of pride, I consider myself to be a prolific arts maker.
I am not really a technical guy. I am not really a hardware guy. Exposure and contact with many artists across many mediums has provided me with some resources and insights in these dimensions. I think collaboration is one of the most important artistic tools any individual has.
I think I am a pretty good art critic, though I have little formal arts criticism training. I consume a lot of art, and bad art makes me angry, and I think that’s enough. I am distrustful of art critics who are not engaged in some kind of art making.
AI AS AUTOMATISM AND INNOVATION
On Wikipedia, AI Art is rather nebulously defined as ‘artwork created through the use of artificial intelligence’. As it stands, AI Art involves the human participant, whom we could consider the artist/commissioner/patron to offer input to an engine/model/program which than produces many iterations of images to parse through, to determine which are ‘successful.’ In some cases the human does the parsing and in some cases the machine does.
Thus far, our viral internet fads have been composed of relatively simple versions of this relationship. DALL-E takes simple text prompts, and returns simple, small images. Lensa takes 10 of your photographs and iterates among a clearly finite number of visual styles (my very first complaint!)
Many of the programs being used to facilitate AI art are much more complex. As any producer of digital art (visual, music, film or animation) will tell you, outsiders (or analog purists) severely underestimate the learning curve of using these interfaces to do anything at all, let alone fine tune one’s creative vision. Talk to your local digital painters and electronic musicians and you will be staggered by the hours of input it takes to hone and refine their work.
Marvel and Star Wars movies employ hundreds of VFX artists, and there have been recent news stories of those VFX professionals worked to the point of burnout. The Oscar nominated animation shorts show animation teams of dozens of people. The barriers to entry, both in terms of technological aptitude, but also financial resource, become more daunting for independent makers every single year. Enter automation, where Artificial Intelligence and Machine Learning technologies can enable one person to do the work of dozens.
My prime example of this is the youtuber Joel Haver. He makes live action sketches that are charming enough in the rambling post-Apatow, we-know-that-you-know-that-we-don’t-know Justin Roiland template. But where Joel really shines is in his animations.
His animations transport him and his costars into fantasy worlds, videogames and high concept sci fi settings, with the bodies of monsters, if they want them. They are rendered in a charming and crude marriage of Ralph Bakshi’s rotoscope and Newsgrounds flash animations from the very beginning of the popular internet. Haver has been remarkably transparent about his process, he uses a program called EbSynth (available free) that helps to automate the rotoscoping process.
Talking with some of my animator friends and looking at their EbSynth experiments, it is by no means perfect, showing some of AI’s trademark smeariness, it requires much editing and negotiation. But it is an incredible resource for animation teams that are not composed of dozens or hundreds of people, and its existence will doubtlessly aid in the work of many independent creators. I worked on a 8 minute rotoscoped short film with a skeleton crew in 2018, I cannot even quantify the number of hours I spent on it, we did it with no budget, I would not do it again, except maybe with the assistance of a program like EbSynth.
In the realm of visual AI art, I must reference the work of an artist named Lanny Quarles, who I follow on Facebook. He produces stunning visual work with the program Midjourney, another nascent and continuously developing Artificial Intelligence program, that similar to DALL-E, uses text prompts. However, the depth and gravity of the text prompts, in addition to the images the program produces are a whole other animal. Quarles posts selections and batches of his work, and occasionally will share the text prompts he used to produce them. Looking both at his text input, as well as the variations from image to image, it is clear that Midjourney prompts are essentially their own programming language, one that demands a keen sense of detail and a knack for poetic associations.
Quarles’ work will at times parrot the visual textures of 1960s sci fi films, or William Blake’s paintings. He is creating an uncanny dreamscape, historic work from a world that was nearly ours. If you don’t think Lanny Quarles is making art, come to my house and I will fight you in the street. Lanny Quarles is making art that is a romatic futurist in conversation with chaos. He is continuing a tradition that defines my own arts making practices, and one that predates computers. Before we could’ve blamed it on the boogeyman of AI, we understood it as the mystic.
The British painter Francis Bacon believed that painting was a violent and sporadic process. He would make ambiguous strokes, at times throw paint at the canvas from the other side of the room, and then attempt to wrestle the image back into his control, back and forth, back and forth, until the work was done. If not for this sense of volatile negotiation, his artwork would not have been his.
William S. Burroughs and Brion Gysin developed ‘the cut up method’ where the writings of Burroughs, as well as authors he had particular esteem or disdain for, and totally random innocuous writing was typed up uniformly. The header, footer, and margins would be cut off, and the remaining block of text cut into quartered rectangles. The pages were than randomly rearranged, and the new sentences invented in the juxtaposition were copied down, to be later wielded and arranged by the author, acting as a sort of compositor-alchemist. Of this process, Burroughs said “When you cut into the present the future leaks out.” He would compose three novels, called The Nova Trilogy, out of the cut up process, and they are cornerstone texts amongst the genre of Experimental Writing.
Bacon and Burroughs were engaging in analog automatism, and one can only assume they would’ve been absolutely tickled by the possibilities opened by Artificial Intelligence’s impact on the work and the process. Neither individual invented the principles of automatism, instead these practices were formally established in the 1920s.
HISTORY LESSON
During and after World War 1, soldiers and veterans, primarily in Germany and France, contemplated why they fought in World War 1. The complex lattice work of political relationships that led to 20 million deaths over 4 years left them cold. Many of the veterans appreciated that they had more in common with the enemy combatants than they had with the political leaders who had sent them to battle.
This produced a great upheaval against institutions, primarily the state and the church, though these makers did not have a lot of love for capital either. They sought out to reclaim image and meaning. They called themselves the The Dadaists. I have always felt vaguely ambivalent about Dada’s treatment in formal arts history courses. It is basically most well known for a shit post.
In almost every text book and slideshow the first (often only) work you see is “Fountain by R. Mutt” (1917) which is a picture of an upside down urinal that has been signed by the artist R. Mutt (actually Marcel Duchamp). It is a perfect spiritual representation of the attitude and modus operandi of Dada, in its aggressive reclamation of image and meaning. But its also short shrift for the arts movement that invented collage, automatism, and basically every transgressive arts making approach in the western world for the last 100 years.
Dada preached chaos and automatism, doing before thinking, and sorting out the pieces later. As more and more of the Dadaists familiarized themselves with the works of Sigmund Freud and Carl Jung, a significant number of them became Surrealists instead, who identified a psychological and spiritual resonance and meaning in the chaos they produced.
The Dadaists and Surrealists would’ve felt right at home in the 2020s, and one has to imagine they would be thrilled with the magnitude of new tools at their disposal. Instead of writing off the new technology as a grift, they would be wondering how they could take control of the grift.
100 years later, we find ourselves in a similarly uncertain and vitriolic time, with more people rejecting societal norms and losing faith in institutions. It is an outrage culture, with an abundance of reactionary and extremist world views bidding for your time. And the artistic output that speaks to people now is split into two lanes. One has grown increasingly bizarre and convention challenging, the other has melted into brain soothing spectacle.
Who could’ve have predicted this, I mean, except for all of the people that did? Namely: Margaret Atwood, J.G. Ballard, Jorge Luis Borges, William S. Burroughs, Octavia Butler, David Cronenberg, Philip K. Dick, William Gibson, Thomas Pynchon, and Paul Verhoeven to name a few.
SO WHAT IS TO BE DONE?
A piece of wisdom I think I first heard in Standupland can be simplified to “find your audience.” Some people will quote you an explicit number, like 150 people or 300 people or 500 people, and say you have to build an audience of x size that’s going to buy whatever you put out. I haven’t found any explicit number to be particularly realistic yet, but the general notion is very true and I find it to be a more pragmatic, less cynical version of “everyone is a brand.”
If you are an artist, produce opportunities for people to give you money, and don’t let grind culture destroy your love for the art, because it will if you let it. All of these environments are wars of attrition. Whenever anybody has me talk about art to young people, or when people starting out ask me for advice, I tell them its really important to develop a good relationship with the process. If your measurement of success is money or prizes or getting laid or any other kind of external validation, this art shit will just destroy you.
Maybe not if you are that good. But in my experience, it turns out, that almost nobody is that good and that all of the arts will offer you plenty of incentives to quit. Blaming the economic tension in art purely on emerging technologies is escapist old person shit, and its also too late. But do we have to hope for and engage in conscientious consumption.
Bringing us back to the Lensa grift/ AI art debacle. Assume that whatever the next viral phenomena is, it’s also a grift. Nobody will bat a .1000 on this, unless you just don’t participate in any viral phenomena, but thinking of yourself as both a consumer and a product may encourage some wariness.
Don’t assume that AI art is pure evil. It is such an umbrella catch-all for so many emerging technologies, and invariably at this point, Artificial intelligence is going to be part of our collective cultural future, at least until things go full bloom Mad Max. Some of these technologies will have the ability to elevate unique voices and makers that would otherwise be lost in the shuffle, and yes, some of them are here only to eat your soul.
My last piece of advice to the civilians will be my most radical and offensive. If you like the work of an artist, support them financially. Between Patreon and OnlyFans and Substack there is an increasing market of companies that exist specifically to facilitate this relationship.
A more organic version of this is to spend money when creators you like provide opportunities for you to do so. Buy the book, buy the shirt, buy the digital download. As a general rule of thumb, they understand that a lot of people are not going to pay them, so you don’t have to leap on every single opportunity, but doing it occasionally is a lot more than never doing it.
You could just send them some money too. Both if they explicitly post a request for financial help, or if they gripe about car repairs or dentistry etc. I know way more artists who are too proud to put up this kind of post than I do artists who would be deeply offended if you venmo’d them $5 out of the clear blue sky. Sending $5 or $10 or $15 is way more than $0. If you’ve made it this far, and have enjoyed reading this you can venmo me @kickholland, but I don’t expect you too, that’s not why I’m writing this tonight.
There are specific mediums and formats of art that have become especially difficult to steadily monetize in the modern era, particularly digital artists, producers of memes, makers of short film/sketch/animations, musicians who don’t play out. The list goes on and on, if you can think of them, they probably wouldn’t mind a little bit of your money.
In closing, yeah, the Lensa grift was a grift. Based on my feed alone, those guys made about as many $8 dollar installments as Elon Musk did. But this momentary bitterness and fear does not warrant writing off nascent technologies that are invariably here to stay. The value automation brings to independent makers and the capacity to expand the artistic conversation and ensuing achievements are too furtive to stop AI Art. This new technology is not what has broken the creative economy, it is merely another signal that the extant structure is no longer tenable.