Last Night A Dall-E Saved My Life?*
"Mash Communication," Artificial Intelligence, and the Future of Art
*(apologies to Indeep for appropriating the song title).
Buried somewhere among the half-comprehensible scrawls of my dream journal, you can find a dream recall in which I found myself attending some kind of “rave” with a macabre twist: all visible phenomena were seen through a filter that made them look like subjects of a Francis Bacon painting. Against a static black backdrop, spotlit dancers contorted themselves in spasms of simultaneous ecstasy and agony, with Gaussian-blurred facial features only occasionally coming into focus, and then receding again into somber, pallid blotches of purple, white and orange that I identified with that artist (disappointingly, I have zero memory of the audio accompaniment to all this). Like most of my more notable dreamed experiences, it was a one-of-a-kind performance that could not be “willed” into a recurring form, and wasn’t something that my fumbling untrained hands could sketch out with any degree of accuracy. That is, until I was introduced to Dall-E Mini.
This undeniably strange text-to-image generation model, which entices woith promises to “[generate] images from any prompt you give!”, not only rendered something very similar to the visual content of this dream, but seemed to render nearly everything in the style of Francis Bacon’s stentorian nightmare portraits when pressed into service again. Searches for human subjects, be they instantly recognizable public figures or simple first-name queries like “Andreea,” would invariably return images whose defining features were cruel or comical distortions: facial melting, bizarre pretzel-knot reconfigurations of human flesh, black hole voids or ectoplasmic glows in the place of eyes. It was really only a matter of time before something this far removed from customary experience would be pushed to its limits by an “always online” culture slavering after constant novelty, and as of this writing, it is difficult to even use Dall-E Mini on some devices: its host site is just too heavily trafficked since its rapid popularization. Entire Twitter accounts, e.g. Weird Dall-E Mini Generations, exist to advertise the model’s propensity for visualizing Surrealist chance meetings of sewing machines and umbrellas upon dissection tables: and even though the generations chosen for sharing typically draw from the well of pop culture, such search results as “courtroom sketch of the xenomorph from Alien” look no less weird for that fact. The connection to pop culture will be touched on more in a moment, but for now, suffice it to say that Dall-E’s creators had to build in the “Bacon”-esque elements as a kind of stunting feature to prevent the arising of legally actionable “deep fakes” featuring users’ favorite or least favorite public figures getting dragged into every compromising scenario imaginable.
There is something about this kaleidoscope of mutant distorted weirdness that, for me, embodies some of the less popular, or at least less lustrous prophecies coming from the realm of A.I. studies. To be sure, the current Dall-E aesthetic is a more believable approximation of the actual human evolutionary process – insofar as we acknowledged that our evolution has not optimized for intelligence - than any number of slick, seamless and sterile attempts to Apple-ify the future. It does not yet inspire rapturous Kurzwellian dreams of von Neumann probes colonizing the observable universe, nor nightmares of an oppressive singleton which methodically eliminates any sentient life that could intelligently oppose its plans. If anything, the artistic signatures I have so far noticed in Dall-E imagery - jarring incongruities, inappropriately swollen or emaciated bodies, and implausible textural combinations - all bring to mind the possibility that a successful superintelligence might choose to act “perversely” and in a chaotic fashion completely alien to the teleological thinking seemingly held in common by most of the human species. In Bostrom’s musings on the actual motivations that an advanced synthetic intelligence might have when becoming indifferent to the dictates of its creators, he does forward the possibility of “an intelligent self-modifying mind with an urgent desire to be stupid,”[i] and, of equal interest, reminds us that
The orthogonality thesis suggests that we cannot blithely assume a superintelligence will necessarily share any of the final values stereotypically associated with wisdom and intellectual development in humans – scientific curiosity, benevolent concern for others, spiritual enlightenment and contemplation, renunciation of material acquisitiveness, a taste for refined culture or for the simple pleasures in life, humility and selflessness, and so forth.[ii]
That does beg the question of just how many humans can tick a single box on this checklist of ethical and intellectual development (for what it’s worth, many would also not pass the Turing Test meant to distinguish synthetic and organic intelligences). A more important discussion, though, relates to how representative Dall-E Mini is of the time and place in which it is presently embedded. Speculating on what it will evolve into, how it will render artists obsolete etc., should be accompanied by plentiful reflections on the present. For the time being, I suggest that we view Dall-E as the latest (and not necessarily culminating) phase in a societal drive towards the normalization of “mash-up” art, namely a form that relies heavily on the quotation of existing and previously completed artworks rather than on the more traditional practice of deriving meaningful information from “raw” meaningless, non-communicative data – this being anything from sine waves to un-hewn stone.
In certain ways, the “mash-up” or “culture jam” is now at the center of our aesthetics and our communication habits rather than being the peripheral, avant-garde form of cultural critique that it may have been in the heyday of Dadaist collage, earlier iterations of clearly activist sampling-based music (e.g. Negativland, or much of the “classic” hip-hop canon). The possibility of being simultaneously “consumer and producer”, while engaging in this act with little expectation of monetary compensation, is now a very quotidian aspect of interpersonal communication. As is the cultural dismantling of “binary oppositions” in physical form or disposition. Other supposed radical re-orderings of the creative order have not gone quite as planned once this normalization process took hold: “the death of the author,” with the abolishing of the hierarchies that came along with the assumption of authorship, has been replaced in mash-up culture by a new set of hierarchical relationships based more upon technical and persuasive virtuosity than upon claims to originality: the works of someone like Banksy are for many the authoritative statement on “street art”, and now the works shared by accounts like Weird Dall-E Generations may come to be seen as the signature examples of creative mischief using that particular tool.
Meanwhile, the prevalence of something like the animated .gif on chat and messaging platforms is, I think, proof enough that millions of people feel the need to make even the simplest communications more authoritative and exclamatory by quoting the words and actions of cultural icons more recognizable than themselves: a statement as basic as “I agree” is somehow given a semantic quality above and beyond this affirmation when accompanied by an animated loop of Taylor Swift or Patrick Bateman signaling their approval. Essentially the same holds true for large swathes of the marketplace for static meme imagery: superimposing emotionally charged text onto images of Kermit the Frog or Drake allows users to “appeal from authority” on any issue from gaming platforms to libertarian economic policy, while also playing on the goodwill of the spectator, who expects (and sometimes even receives) a comic jolt from these incongruous pairings of famous spokesperson and obscure / petty grievance. That so much of Dall-E Mini’s output is steered towards similar content as the aforementioned .gifs and memes is hardly a surprise, given how much of that is already floating through the infosphere. In fact, with things as they now stand, it would be a surprise if people were not compelled to create imagery from queries like “the Demogorgon from Stranger Things holding a basketball” (the “pinned tweet” on the Weird Dall-E mini Generations account as of now – surely likely to change).
With these tendencies practically saturating modern culture, it is really not advisable to aim for “challenging subversive” status just by situating pop cultural references in implausible scenarios. As Michael Serazio insists, “bricolage for its own sake […] hesitates to espouse anything more than detached, wry commentary,”[iii] and I would argue also that such is analogous to robbing banks in order to give a black eye to “the system,” when such an act tends to flatter that system’s logic more than anything. Critic Mickey Vallee correctly, I think, elaborated on the “paradox” of the mash-up by noting how it “allows us to deconstruct binary oppositions, but is silent regarding the inner mechanics of the system it deconstructs […] even though users actively create multiple lines of flight, they are unidirectional: towards the social imaginary of pop cosmopolitanism.”[iv] Furthermore, when considering the sort of utopian communalism that some critics still expect to blossom from the fertile soil of pop cultural sampling, the on-the-ground reality is less romantic: streaks of non-cooperative rivalry, one-upmanship and other disharmonious behaviors have not melted into the ether simply because information that “wants to be free” has been freed, with various intellectual property rights disregarded en route to that goal. In one notable instance, the mash-up producer Freelance Hellraiser complained that others were making derivative works based on his own derivations, comically (but apparently without irony) complaining that these works made their rounds “without giving credit to the original idea.”[v]
Whether respectful or insulting to its subjects, whether collaborative or antagonistic to others, I feel that present-day, self-referential mash-up art rarely finds a way to achieve anything an ability for the artists / collagists to reflect their understanding of the current zeitgeist back to themselves; to help calibrate themselves to the social expectations of the present (such as alleviating the “fear of missing out”). Again, this is often highly fun, and occasionally worth its weight in belly laughs, but I believe it is a completely different experience from that of drawing on the relatively unknown, rather than the easily recognizable, as one’s creative raw material. My sympathies tend to lie with those who can accurately relate an emotional state or a resonant, trans-generational truth via something as bewildering as, say, projecting sequences of non-informative strobing visual data or rendering compact discs’ non-“musical” error-detection data as a form of organized sound. There is something noble and Promethean about the quest to use the data of scientific, non-aesthetic inquiry to aesthetic ends and to possibly fuse the goals of these disciplines. This, I feel, is where self-referential mash-up, calling mainly or exclusively on pre-packaged information, falls flat: it can make aesthetic reference points bounce around wildly and entertainingly, but I have increasing doubts about its ability to illuminate anything we don’t presently understand. The meaningful history of the artistic or poetic impulse seems, to me, defined by successive acts of forging analogous (and not immediately comprehensible) connections between opaque data and established information, rather than subsiding solely upon juxtapositions (no matter how ingenious) of the familiar.
This is where the relation of this cultural output to A.I. comes in to the picture again. Presumably, the goal of much A.I. and transhumanist experimentation is to put us in touch with this very unknowable material that we have historically attempted to reach with artistic communication. Ideally, this could be accomplished via an intelligence that can process information much more rapidly than our comparatively lagging rate of “internal communication speed” (a frightening point of comparison comes from Bostrom, who notes that our “axons carry action potentials at speeds of 120 m/s or less, whereas electronic processing cores can communicate optically at the speed of light [300,000,000 m/s]).”[vi] As such, I am under no illusion that we would be feeding exclusively Dall-E image search results into something that we eventually hope to become the general intelligence that can match human intellect in any area.
I wonder, though, what such an entity would think if this were a significant part of its “training”: being blasted daily with hundreds of thousands of exemplars of this sort of creativity, could it maybe come to view humans as hopelessly caught in a loop of commenting upon their present realities, and, upon reaching the stage of advanced (read: superhuman) intelligence, come to find its hosts not worth serving nor even co-inhabiting physical space with? Given, a superintelligence that arises under any circumstances “[might not] necessarily share any of the final values stereotypically associated with wisdom and intellectual development in humans”, and might ruthlessly direct all of its intellectual resources (and all available planetary resources) towards self-propagation. Would it be wise to further goad it along this path by convincing it that the human species is locked in some strange self-referential loop, unwilling itself to outwardly expand? To propose yet another scenario, could flooding the A.I. with a sugary diet of quirkily mashed pop culture cause it to simply resist more nuanced forms of innovation; to hasten the time at which it “goes rogue” and refuses a final goal of utopian evolution in favor of its own focus on amusement and ultimately meaningless combinations?
To really answer those questions requires, naturally, far more insight into the precise operations of A.I. than can be brought to the table right now. However, as we invest ourselves more and more in the creation of a general intelligence that will match and then surpass our abilities, we should not lose sight of just how “unknowable by nature” (to borrow James Barrat’s term)[vii] this entity will be. While the mash-up aesthetic serves a limited role in commenting on various aspects of our known present, it is not that well-equipped to reveal any truths about an extremely uncertain – perhaps terminal – phase of human development. I would argue that, if we want to fully explore the coordinates of a world where A.I. overcome us, and if we want to determine whether such a reality would even be worth aspiring to, then we will need more creative forms focusing on the unknown and unknowable. We may need to once again commune with the storehouses of mystery we can find in nature, particularly when considering what George Dyson has suspected: that nature is “on the side of the machines.”
[i] Bostrom, N. (2014). Superintelligence: Paths, Dangers, Strategies. Oxford: Oxford University Press.
[ii] Ibid.
[iii] Serazio, M. (2008). “The Apolitical irony of Generation Mashup: A Cultural Case Study in Popular Music.” Popular Music and Society, 31(1): 79.
[iv] Valle, M. (2012). “The Media Contingencies of Generation Mashup: A Žižekian Critique.” Popular Music and Society, 36(1): 1-22.
[v] McLeod, K. (2005).”Confessions of an Intellectual (Property): Danger Mouse, Mickey Mouse, Sonny Bono, and my Long and Winding Path as a Copyright Activist-Academic.” Popular Music and Society, 28(1): 79-93.
[vi] Bostrom (2014).
[vii] Barrat, J. (2013). Our Final Invention: Artificial Intelligence and the End of the Human Era. New York: St. Martin’s Press.