DALL-E 2 and Objective Art
AI-made art has never inspired anything more than a shrugging response in me, which creates some difficulty for understanding the explosive appeal of the recently publicly released DALL-E 2 technology. To describe its function and productions as novelty-based would, I think, be accurate, but to leave it at that — as an explaining-away dismissal — doesn’t do much intellectual lifting either, since novelty is predominantly what our cultural productions run on now. So, let me just state: I am distinctly unimpressed by DALL-E 2′s productions (even if they represent a technological leap (they do)) and the novelty of a thing very rarely moves me to action outside of an attempt at critical exploration. So, I’m going to do a bit of that critical exploration.
On one level, it is mildly sociologically interesting how DALL-E 2′s capabilities tie into the wider social media phenomenon wherein the first impulse is to make a sort of joke. The aim of these graphic combinations is distinctly not artistic sophistication but, rather, a brute-childish “lmao i squished the two things together” act of play. In that sense, it also ties into how the idea of a “meme” is now implicitly comedic. The things that spread the fastest online are charged by humor or anger, or a mixture of the two, and anything that can facilitate that process in a sort of automatic way is likely to be adopted on a large-scale by users. In a sense, DALL-E 2 has acquired a functional near-equivalence to something like Meme Generator, although its element of chance — Is it able to create the visual combination I’ve specified? — endows user interaction with a game-like aspect, giving it added entertainment value.
What often seems to be most impressive about these examples is not the images themselves, as standalone artistic works, but the program’s ability to approximate the desired combination with either vague recognizability or convincing fidelity; and it is in that way that I understand the phenomenon as novelty-based. There’s a sort of admiration of raw technical success here, a kind of humored “gee, wouldja look at that” sentimentality over feeding the machine one’s absurd idea and it succeeding, on some level, representationally.
I am tempted to form a connection with the multiverse/universe-colliding trend in superhero media and other comparable media. With those projects, I get the sense that some wall of (corporate) imagination has been reached, and the only way to simulate/stimulate creative development is to push everything beforehand together for a sense of cohesion and narrative elongation. Individuation is lost as the conglomeration of All Content takes precedence. Similarly, with DALL-E 2′s images, what is driving a lot of the content’s production is a “what-if” desire to form mounds of history-/chronology-jumbling ersatz that may be amusing in the moment, because of that frisson of combinatory recognizability, but which also seems historically dislocated, bound to the attention of a fleeting Now.
When someone in a server I attend (half-jokingly) wrote, “think of all the porn this can generate”, their comment seemed unintentionally spot on to me, comparatively speaking. Pornography — or the (masculine) pornographying impulse within sexual play — and technically advanced image-generation such as that afforded by DALL-E 2 offer a similar, frisson-like pleasure in putting things together which “shouldn’t” go to together and suddenly seeing their absurd, sometimes grotesque, overlap. Prompts such as “Bugs Bunny as a Greek statue in Athens” or “still of Shrek in Spirited Away (2001)″ are really not that far off from “asian bbw inflating a balloon.”
Another person elsewhere wrote, “Don’t like the way AI art interacts with a physical multiverse which contains all that can be imagined. There’s something distasteful about the way this reaches into the realm of… platonic realism, idk.” This reminded me of the disturbing trend of grotesque, algorithm-generated “kid-oriented content” on YouTube, excellently described by author James Bridle in a 2017 essay. The machine cycles through an internet-derived bank of content, producing countless combinations of material, with a kind of simultaneous resourcefulness and somehow-wrong ugliness.
One last observation before I conclude: a hindsight-obvious yet crucially important distinction of AI-made images, like those of DALL-E 2, is that the machine can never refuse a request; it can only acquiesce and then approximate. The curatorial aspect of production solely involves what’s to be found in the bank of images. In a strange and unprecedented way, it represents “objective art.” But the ability to say “no” is just as important as the ability to say “yes” within the artistic process. AI-made art reduces this process to the variables of content relevance and graphical fidelity.
I’ll end with a quote from a friend during a discussion about this topic.
I think this sort of thing could ultimately be good for the arts. It could cause a reckoning that leads us to understand that the value of made things-as-products falls away at a certain point, and we create in order to practice our humanity, our humanness, and to find whatever we find, as individuals, in artistic creation.
I came to this conclusion when I read a lot of praise for a poetry generator that turned out some genuinely interesting material, and then paired that with the tech-minded praise & enthusiasm over having “automated poetry.” […] There’s this notion that Heidegger explores in his essay on technology called Bestand, which can be translated as “standing in reserve.” He used frozen chicken in warehouses as an example. We store and store resources with the intent to use them, but they do not always actually get used, so our priority isn’t to use later but to store. And the units become meaningful in themselves, not because we have a relationship to them, but because they are stored.
And I believe this is very close to our attitudes on information, now, with the internet. I think it was a governing (if unintentional) concept behind the mass surveillance push after 9–11, which certainly wasn’t optimal surveillance. And this extends to computer generated poetry & art. What does it exist for? It is not a response. It is a thing. And the things are entertaining. But valuing art as things rather than as responses, and that is exclusively a property of human creation, turns them into Bestand. Things as things to exist.
