eugénie desmedt, wholesale! assembly-line fiction — review
desmedt correctly criticizes the politics behind artificial generation, but doesnt seem to understand how perception and interpretation come to influence the future evaluation and generation of art
even though the text begins with lovelace and the way in which periodization and automaticity are understood as complacent in reconfiguring ai art and writing as ontologically flaccid, i figured id return to her grammatical concern only after tackling the text's treatment of novelty and originality in art first, because of the way this intersects with the critique of novelty in art in the first place, essential in figuring out what aspect of deliberativity and determination ai lacks in order for us to comprehend how we should situate ourselves politically around the question of automation to begin with.
the text sets up the argument about novelty through german thinkers bense and moles, arguing five things, four of which are correctly situated as seeing art as defined by decomposability into elementary signs, viewing aesthetic value as modeled as a ratio of complexity to order, interpreting aesthetic information as improbability and surprise, and seeing that art produces order against entropy or what bense calls aesthetic negentropy. i do have a small comment on its use of improbability and surprise, mainly that it kind of syphons lovelace into the point about surprise, when bense would argue something much closer to the idea that technical objects cant possibly be derivative if they are unsurprising, but that they are transformative when they question existential ontology but only seen through the lens of meaning rather than unexpectedness, when they cant be situated as technical objects. this is because when technical evaluative standards cease and categorization isnt possible, thats when novelty becomes an ontological confrontation for bense.
moving onto the key issue i have with this section, and the only issue i have really with this entire text, desmedt situates moles and bense by giving a sort of academically steryotypical view of objectivity by setting them up as arguing against subjectivism in artistic interpretation and orienting them around objectivity, measurement and scientism, tying them to shannons cyberneticism. this on its own is correctly framed, but it implicitly misframes benson - maybe not intentionally but maybe due to lack of argumentative density - as arguing that information quantity, density or categoricity is related to aesthetic judgement, when in fact for him, experience cant be situated through positivism but depends strictly on relationality as a deeper semiotic judgement, something not entirely reducable to cybernetics or mathematics, but very much in fact dependent on human experience that isnt subjective but is very much relatively justifiable in its interpretative volume and excess.
bense's argument about experience can be seen in works like technical existence, which argue in a technorealist fashion that technics itself can no longer be interpreted through theology or mythology, because of the way the technical artwork is meta-surreal in the way that it doesnt portray technicality but orchestrates it, requiring a more semiotic understanding of relationality within art, even albeit an objectively schematicized configuration.
this is also where i think the whole problem with desmedt's undertaking lies, i believe she fails to consider the way in which human mediation actually enables the relationality that can arise as a result of automated generation. this is maybe even partially because bense himself is very ready to discard of the impact and influence of art way too early if the piece isnt immediately transgressive, which is because he largely believes that the techncial domain is uniquely hyperreal in its consequences to our own lives. however, desmedt doesnt have her own counter-thesis on mediation currently, and, as you will see now, she doesnt even tackle bense's own conclusions on this issue.
desmedt's text argues that we would find it difficult to figure out what bense would have made of the conditions of ordering that has structurally spawned the way we understand and work towards improving ai output. but the text rushes too quickly and doesnt stop to comprehend how the answer already sits within the text, or the ways in which an answer can be derived from it with a slight amount more effort. decomposability as a term is only mentioned twice in the text, and the passage that mentions it doesnt cover ordering in as much as it does aesthetic analysis, which bense argues that is primarily semiotic and interpretative in nature.
this means that for bense, intensities, relations, topological coincidences and configurations are driven through our experience of a piece as a subjective inward analysis of its meaning but an objective outward perception of its contents, whereas categories within an analysis are every other form of self-conceived ontic analysis that attempts to withdraw meaning from what it thinks is significatory but really is a total essential marker that serves no purpose to be considered outside of the very subtle layer of meaning imposed on by our experience of the latent artwork, or in simpler terms, that categorization is a form of descriptive and literal (realistic) signification that never reaches the aesthetic territory.
as such, analysis of ai artworks shouldnt be literalized as self-surreal due to their ontic composition as the text mentions the famous ai artwork of a jesus shrimp, but for bense, through the contextual apparatus that flags the composition itself as an establishment of forms that are subtle and provocative when considered philosophically and not strictly technically. even though the politics behind ai generation is self-limited as the text correctly and succintly argues, it entirely throws away the idea of ai's impact and influence on our own supposedly accidental and deliberate, novelty-producing creations.
in this sense, it doesnt see it as an extension but a regression of our thought, which, fair on its own, doesnt realize that functionally this is already happening even if polemically the way it influences our own novelty in the long run may and likely does have negative influence over our deliberation. therefore, the answer to how bense would view ai art isnt through derivativity or surprise strictly speaking, but through its inability to propose existentially risky works specifically because of the fact that their pre-built hierarchization resists technical transgression if it is influenced and built entirely by the interior world of technical regulativity.
this really is my whole point of contest with desmedt's work, it isnt so much that any of her framing is wrong, but that it categorically lacks developed lines of thinking for any of the philosophers, moving in a very narrow path of thinking that disables the nuance that actually confronts the ways in which human-generative orientations also would produce ontologically meaningful works, a problem i will deal with later.
returning to lovelace now, the text uses her to argue roughly that machines cannot cause origination or surprise as meaningful variables, that originality is not just output novelty but ontological status of thought, and in this sense, the text correctly identifies bense through lovelace, but fails to account for why this is the case exactly, and quite literally states that it isnt sure of how to situate this aspect of things.
desmedt's text uses knipe's grammatization most accurately as a metaphor, arguing that industrial-scale automata in the form of ai produce endless artifacts as a side effect of the digital economy and the way in which it influences consumer consensus. even though aptly this is true on its own, bense would argue that most human-developed artifacts also fall below this line, albeit he would concede that ai-generated artworks due to their objectively weak developmental rng simply cant have a transgressive nature which does work in desmedt's favour. she also further uses him correctly to argue her actual main point, which is that the structural exclusion of ambuigity in objectivist standards does in fact lead to an inability to create structurally innovative artworks.
lovelace does literally argue that representation has no ontological novelty, and desmedt does even pose the big problem with this in her own text, stating that it isnt exactly clear how lovelace's mathematically representational perfectionism currently aligns with the representatively derivative ai creative direction. her use of tyche also blends all of this together, correctly conceiving of an automaton as a structurally impotent form of creative deliniation under the boundaries that she herself puts forward. however, and this finally is my critique, it isnt clear in this text the way in which originality, surprise, representation, evaluation and decomposability actually serve to co-exist within human-generative contexts outside of its general criticism of the de-ontologization of ai, which is banally true and correctly argued for.
the problem with the general line of thinking is the following: a machine through its constituve developments can surprise us ontologically in so far as our own subjective evaluation itself can be thought of as objectively connecting to technicalizaed schematics that arent artistically experiential in nature but under the line of the influence of derivative influence, allow us to produce deliberate and decomposable pieces.
the text ends on the argument that the paradox of controllable creation is the real libidinal charge and simultaneously frustration behind ai-generated content rather than some fantasy of computational originality. even if it is trivially true that sorting chaos into order can be defined as complexity, and even if desmedt's critique is about the limits of machines rather than the way machine-mediated regimes influence our own undertaking, the critique she launches is still a staunchly framed political struggle against a perceived automaton's possibility to exist and create supposedly derivative works, all the while the non-derivative ontology of beings themselves is never posited in the text, nor is the overlap of influence (positive or negative) between automation and ordered human transgressive accidental work of art, which is why i give the text a 6/10. in my opinion this article is a torsion, check out what that means right here. if you enjoyed this video, please check out the rest of my reviews right over here, and also, i will be sure to read all of your frustrated comments about contingency and novelty down in the comments below. and for now, go get some rest.
commentary:
leading onto some of my more secondary concerns produced in desmondt's polemical paper about ai art, i'm a little concerned around some of the more referential and politicized points of argument. in another text with a similar example of categorical collapse in understading bense, frieder nake interprets bense as arguing that works are produced by artistic generation but only come to be works of art through artistic criticism or perception. then he jumps from this interpretation into the argument that, since artists technically always produce the work of art according to the value they deem critics will impose on it after the fact, that this is already constitutively moulded into the creation of it, and that therefore, only an objective post-evaluator and its presence can ever allow for objective evaluation. this claim also sees objectivity as observer dependant rather than co-relational, which is what bense originally argued, conflating relational ontology with epistemic dependence.
it is in fact moles who argued that the dependancy towards perceivers mental states shift the framing of the piece, whereas bense's objectivity is precisely in the non-requirement for art to allow interpretors. artistic pieces are finished, but remain uncompleted epistemologically, and only produce non-actualized signification until a relational observer appears. the taste of the objective observer has nothing to do with influencing the previous trajectory of the artwork precisely because observation is objective but what goes into the creation of the artwork isnt, only its end result. the mediation exists not on the level of pre-creation or post-interpretation but exactly between the way the art is already post-configured and the way this configuration can then be interpreted correctly (and you can only interpret it correctly if you analyze it informationally or in other words relationally, even if your conclusion of it remains subjective in its contents)
this point of contestion i have with desmedt's text can also be passed through its use of zeilinger, who is mentioned as arguing against the imperative that fixed outputs always lead to unfixed meaning, where he apparently states roughly that dissolution of imitation and its production of unexpected outputs itself is a form of originality. on its own this statement is interesting and has a nice location within this otherwise staunchly politically realist text about the hard limits of current discourse (which in and of itself make sense and are argued for well outside of its use of some of the references).
zeilinger himself in papers like the "politics of visual indeterminacy in abstract ai art", interestingly argues that ai-based image generation itself stands as indeterminate for humans and successively replicates abstractivity, as well as the politics behind it being used simultaneously to govern sexually offensivity amidst its current use as a widespread tool for the creation of nsfw illegal deepfakes, all whilst counter-arguing that automatic ai systems in and of themselves flag, filter and censor human-discernible offensive content, which on the one hand is a censorship of indeterminacy by way of forced hierachization of politicized filtering.
and on another hand, proves to show how zeilinger himself understands the political dimension of ai better than desmedt, because this text implicitly recognizes that even human filtering can be thought of as automatic in the ways in which it paradoxically already serves as an incongrous filter of its own, with the only caveat being our self-perceived ideal of self-complexity not fully politically ruiling out the notion that we also have set standards of evaluation in the judgement of taste (which now also due to individuation is scaling into predictability at higher rates), but in the way in which we also simultaneously provoke indeterminacy in art in the way which zeilinger shows, in a manner that questions the limits of analysis as objective when it comes to categorical deliniation and subjective when it comes to our scope of interpretativity when it comes to semiotic and relational signification, something the text previously failed to mention or correctly situate.