us at CalArts. My name is Amanda Beach. I'm the Dean of the School of Critical Studies up at CalArts and I'm chairing these three talks. This is the second talk tonight. We have the third talk which is April the 10th I think with Rachel Garfield just to let you know about that one in the future and this series is called Structures of Dissonance Aesthetics Beyond Capital question mark, very important bit to put on the end there. And I'm just going to talk a tiny little bit about the theme before introducing our amazing speaker tonight. So the political claim of contemporary art, specifically since the mid-20th century avant-garde, has often been hinged upon a
politics of resistance, antagonism, and difference. In such case, a political art does not structure power but rather unstructures it in creative forms of destabilizations. Often this form of art has been identified as an aesthetic experience where the experience of sensory phenomena is seen to resist the top-down ordering and instrumental functions of rational languages and aesthetics. This type of experience is often being claimed to make art akin to an indifferent kind of nature that withdraws from and troubles any form of organizational power. And yet we can see how capital, the capital we live with in every day, produces its own form of dissonance and its own naturalisms.
Unstable environments of high-frequency exchange and unregulated markets bring us to this delirious, alienating, and dissonant experience of living in and as capital. It divides us, estranges us from ourselves and each other, proliferating difference under the principle of accumulation. And we know that these affects, whether they are seen in art or capital, no matter how apparently chaotic, rely upon produce and calcify structures. Here we could say that despite art's critical work, it risks underwriting the unbearable social inequalities of race, gender and class that have been authored by capital. This all together leaves us wondering what systems, orders and principles are relied upon and constructed in art's claim to aesthetic resistance today.
What forms of dissonance is art inhabiting now, if any? Is dissonance a useful topic for art today at all? And if art requires structures, rules in order to achieve dissonance, what are those exactly? And I'm just really pleased and amazed that we have Luciana Parisi here tonight, because I've been trying to get her to come over to the West Coast for quite a while, and COVID was really difficult for us, so we did some online work. But you're here in person, so I just want to thank you so much for being with us tonight, Luciana. Luciana Parisi's research lays at the intersections of continental philosophy, information sciences, digital media, computational technologies.
Her publications address the techno-capitalist investment in artificial intelligence, biotechnology, nanotechnology, and others to express challenges to conceptions of gender, race, and class. And she's also written extensively within the fields of media philosophy and computational design in order to investigate metaphysical possibilities of instrumentality. Her CV is absolutely extensive, and I know that you've got a copy of her biography in front of you so I won't labor the point but again I just want to welcome Luciana thank you for coming thank you for coming on a terrible rainy night which I can't believe it's still raining here with the kind of film noir rain that we keep getting these days and I hope you enjoy the talk we'll have
hopefully plenty of time afterwards for questions so whether you're remote out there in wherever you are and you're watching on YouTube or you're here in person I hope that we can have a really open and possibly productive discussion after Luciana. So thank you. Welcome. Thank you. Can you hear me? Thank you very much, Amanda. And thank you very much to everyone who has made this possible today for organizing my flights, my hotel, and for being here at the ICA. I'm very thrilled to be here. I hope this works fine. So, yes, I'm going to start just reading some of my work that I've been producing and writing and thinking recently.
So I appreciate you being here. Hopefully this can be a generous discussion. As OpenAI launched DALI-E, concepts in natural language are directly transformed into images. AI-generated images are said not only to show us that AI understands human worlds, but also to stretch the aesthetic dimension of vision. However, I suggest algorithmic instrumentality begs us the question of what counts as aesthetics in the aftermath of computational thinking. To approach the relation between instrumentality and aesthetics, or between AI and art more generally, one must ask, can algorithms meet the creative ambition of aesthetic production without simply intensifying the recursive trends of cognitive capital?
Can machine aesthetic act as a collider and accelerating non-instrumental instrumentality, running away from the cosmogamy of men, the colonial and patriarchal onto epistemology that grant authority to capital? The paradox of aesthetic and capital, which sustains the freedom of so-called homo bioeconomicus, as Sylvia Winter calls him, through the algorithmic function of the slave machine, is designed by what Francois Laruel calls the decision structure of philosophy. So to unpack the paradox of aesthetic and capital, I want to suggest today,
one must attend to the question of instrumentality, and thus reopen the critique of instrumental reasoning, through which philosophy has set up AI as a mirroring of its priority. This impasse, I argue, precludes critical thinking from questioning technology beyond the Promethean mission of salvaging man. Since instrumentality turns the limit of reason into the exteriority of the machine, I also argue that AI envisions what I call an alien hypothesis, a xenopatterning for the unknown, for which capital confront incompleteness in the abstraction-extraction of the negative slave machine value.
So my talk today draws on some of my recent work on the critique of instrumentality, through which I question the subsumption of technology and generative AI to cognitive and creative capital. But to question this assumption, I argue, is also to challenge the onto-epistemological structure of capital for which the duality of philosophy or reason and aesthetics or the being of the sensible ultimately results from the authority of self-determining decision. I will explain more what I mean by this. To address what aesthetic is in the aftermath of computation, I
suggest one must decouple instrumentality from the decisional structure of philosophy for which technology only remains a medium for onto epistemology a prosthetic extension of aesthetic judgment in francois laruel non-standard aesthetics there is already a refusal to subsume aesthetics to philosophical decision, and an effort to expose thinking to the negative side of this subsanction. My talk today takes inspiration from Francois Laruel's project of disentangling aesthetics from transcendental decision, but more directly focuses on how
the automation of reason, the automation of aesthetics from transcendental disease, sorry, of aesthetics, has importantly reopened the problem of instrumentality beyond the critique of instrumental reason. Namely, beyond the rationalization of reason by the hands of machines or techno-capital abstraction, extraction of value that has failed, according to the critique of instrumental reason, the ambition of critical judgment to ensure the emancipation of men from programmable stupidity. So, when at the end of World War II, Heidegger issued a warning against the demise of philosophy
after cybernetics, he claimed that with the new dominance of information systems, philosophy was replaced by efficient functions. Similarly, it has been recently argued that if artificial intelligence has turned transcendental metaphysics into algorithms, so too conceptual categories have become just calculating probabilities. At the core of the shift from the world of reason to the world of search engines is the assumption that computation has turned thinking into a mode of limitless cognition, whose access to the real is granted by the computational compression of large sets which orient the system towards automated decisions,
from automated decisions in car, military drones, security identification, job recruitment, risk assessment, etc. So in today's debates, there persists a call to save aesthetics from the stealthy power of algorithms to influence choice, direct trends, reinforcing belief and ultimately collapsing the critical capacity of self-reflective thinking into algorithmic orchestrated tasks. According to Stigler, for instance, algorithmic logic has become today's equivalent of the power of instrumental reasoning. The promise for media decentralization, for collective and democratic social interaction has turned into a new barbarism,
That's what he calls it. A world without reason, where corporate giants direct everyday sensibility through recommendations, algorithms, and ranking search engines. Inasmuch as the critique of instrumental reasoning unveiled the failure of rationalization and recursivity, sorry, rationalization, the reduction of self-reflective and self-determined thought to cybernetic operations of feedback, so too today lament about the end of aesthetics points to the AI extraction of value, creativity, autonomy, and freedom. In particular, instrumental reasoning is said to coincide not only with the performative
operation of extractive cognitive and creative surplus values, but as Jonathan Beller claims, it entails a general mode of computational racial capital, defined by the economic calculus of the dialectic of social difference. Heller argues that racial abstraction is endemic to real abstraction, a social difference engine, a technique of value extraction, racialization, and instrumental social differentiation. Similarly, according to Atanasovsky and Bora, at the core of technological innovation, there lies a surrogate labor towards the, unquote,
human sphere of life and labor and society, end quote, that enables in turn the constant reconstitution of the liberal subject. Here, end quote, the racial unfreedom of the surrogate, as they call it, end quote, subtends the project of a self-determining humanity, or what Hortense Spillers calls the feeling human, that sustain the ontopistemology engineering of racial capital. From this point of view, sorry, from this standpoint, if aesthetic abstraction coincides with an instrumental engine that perpetuates anti-blackness, anti-queerness, anti-transment,
anti-immigration, anti-refugee, how can the critique of instrumentality challenge the colonial, racial, and patriarchal pillars of capital without relying on the authority of philosophical decision, i.e. self-reflective judgment, transcendental reasoning, which constitute the onto-epistemological infrastructure of the critique? In other words, can the critique of instrumental reason avoid restoring the authority of philosophical decision for which the loss of reason has led to the end of the world as we know it? How does the critique of instrumentality avoid reproducing the metaphysical architecture that is trying to challenge?
What is needed, I have argued as well, is a radical query into instrumentality, concerned not with establishing or making more transparent the new paradigm of computation to explain what computation can do for art or vice versa, but rather with a speculative and critical practice for what I call an alien hypothesis, in which generative machines unfold non-programmable ends. But how to escape the authority of philosophical decision without claiming for the irrational in AI aesthetics? For instance, a speculative conjugation of images, concepts and sounds in generative AI, such as DALI-E, may need to challenge the connectionist
model that associates objects, concepts, in order to envision a generative mode of reason without authority, and the generative mode of the sensible without human feeling. I'll just leave this slide for a second because I didn't turn it at the right moment. Francois Laruel's argument about the transcendental computer may be of help here to sustain my point. Automated reason and automated aesthetics for Laruel may coincide with algorithmic automatism,
which is definitely, he says, given in a fanatory way. It is given as a scientific transcendence and a meta-language corresponding to, a strictly machining or technological type of AI form. This AI form is presented as immanence, an immanent emanation of intelligence, but it is instead, according to Laruel, determined by the concept of performance. This concept, he claims, is a criterion that explains how to identify thought and computation. But it also defines what machines, but also what defines machines whose performance is expected to surpass human intelligence.
It is here in this streak of performance between humans and machines that the possibility, the real possibility of imminence against philosophical decisions is rather blocked and contained by precisely the quest for, end quote, the meaning of intelligence or what it can do, end quote. What Laruel calls algorithmic transparency is another way to describe how this performative model of AI focuses on a quantitative view of intelligence, as that which measures the correlation between premises and results in terms of effect, and presupposes homogeneity between the syntactic and the semantic order.
One can also add here that this view coincides not only with the classical model of symbolic AI, where performance presupposes the conceptual categories which are programmed into artificial connectors, but also with the inductive model of machine learning of AI today, which requires a concept of performance to explain how artificial neural networks learn to compress information by associating data in search of meaning. In particular, performance here has the task of reiterating a direct correlation between the syntactical and the semantic order or representation
in the name of a given or a priori decision of what L'Aruel calls dualities. The argument that L'Aruel makes is that intelligence is prejudged. It is caught in the decisional imperative of philosophy that establish its limits and goals so that it can remain a given passive machine, AI is only a vessel for the performance of thought. Forlardwell claims that the scientific transcendence about intelligence is no other than a necessary condition to reestablish the transcendence of philosophy. In other words, intelligence is reduced to quantitative performance so that it can be argued that this is what machine can do.
This circular operation serves philosophy to justify why machine intelligence is never, and could never surpass transcendental reason, deductive proof, reflective judgment, or critique. According to Laruel, philosophy only uses intelligence, and quote, on behalf of a very special form of thought, probably irreducible to any numerical combination, end quote. What is condemned to be a logicless performance of machine thinking, therefore, only serves to reinforce the transcendental performance of reason, or the transcendental concept of performance, as Laruel Coles said.
This imitation game between intelligence and reason here takes the circular form of an expanded transcendence, whereby the limits of self-reflective judgment are turned into the prosthetic extension of cognition, namely AI, which are set to justify the performance of critical judgment through the mindless performance on machines. What we have here is the recursive preservation of onto-epistemological dualities between reason and intelligence, self-reflective judgment, and mandalist automata that belong to the same order of self-determination of being and thought, namely the order of the human.
According to Jamaican philosopher Sylvia Winter, this order is the key to the techno-industrial capitalist mode of production. In particular, she argues that the order of premodern metaphysics and the order of modern science overlap in the continuous overrepresentation of the biocentric model of the human, for which science is to justify the transcendental decision that establish the limits of man as the space of otherness. This space of otherness results in the division between two categories, she claims, those selected and those non-selected by evolution.
This categorical order thus exposes the paradox for which the ontology of the human could be independent of its description, I exposing how ontological self-decision is rather set to precede scientific explanation or the scientific endeavor of knowing. From this standpoint, one can argue that instrumental reason cannot offer us a radical and immanent critique of AI because machine thinking is already set to perform the ontological decision of being and thought through the biocognitive epistemology that AI must constantly reify,
i.e. the ontobiocognitive epistemology of the human, or what she calls of man. How to then overturn the self-fulfilling critique of instrumentality, as always already quote in the auto-posing transcendence of philosophy, which constitute the pillars of colonial and racial capital? As much as cognitive capital relies on the potentiality of abstraction and extraction, that is, on the machine as the prototype of enslaved free labor, a labor for free, it also relies on what I've also called techno-flesh, a material matrix subtending the racialization of knowledge and the speciation of the human.
Sorry. The speciation of the human, sorry. The cyber mechanic intelligence, I have argued, has been systematically dispossessed and annihilated, neglected, dismissed, and abandoned in critical accounts of the social reproduction of capital, which I fundamentally seen machine through the lenses of the master narrative of capital. These views return in debates about machine intelligence, confirming the tropes of mindless efficiency, non-conscious functionality,
non-sensical language, hallucinatory responses, and erroneous decisions. At the end, the critique of AI is only another way to re-impart the authority of the master on the tools that sustain the recursive epistemology of humanity. What I want to ask today is, instead of resolving the instrumentality paradox by extending the category of the human to machine, how has techno-flesh become fugitive of the human and of man? If, according to Winter, the ontology of humanity must repress, ignore the external means it relies on for its own ends, aesthetic judgment must rather reconduct the externality of the sublime,
an empty category for absolute indeterminacy, back to the universal condition of being, assigned to art the need to exceed logic, form, and reason through the sensible. Means are here a marker of externality or a thing, a reference to indeterminacy that must carry the work of transcendental judgment. As Denise Ferreira da Silva discusses, this indeterminacy that is attributed to the world of the sensible is a call from, unquote, the realm of the subject, whose faculty of aesthetic judgment rests on a figuring of the sensible and the condition of affectability mediated by the forms of transcendental reason and a view of imagination that articulates it as always
already in the service of the abstract forms of the understanding, end quote. According to Da Silva, knowability in the Kantian formulation of the aesthetic register refers to what she calls the transparent eye, a formal entity, the one whose relation to the world, both sensible and intelligible, is given by forms, intuition and categories of the mode of cognition grounded in transcendental reason. While transparent subjectivity becomes racially coded as white and constituted as self-reflective interiority, she claims, the sensible defines the condition of affectability
located in bodies of those who are not white, and rather seen as externalities, or what she calls affectable exteriorities. Since exteriorities have no purpose or are said to lack a teleological program, a final form, they can only be affected but cannot affect. They can only be known but cannot know. The indeterminacy of the sensible, therefore, entails already a racialization, a gendering, and a sexualization of machines as exteriorities, where the affectable carries out the free labor needed to preserve the self-reflective interiority of the subject.
As Yook Hui has already demonstrated, in his attempt at discussing cybernetic metaphysics, the circularity of self-reflectivity can only be understood in terms of recursivity or recursive function, namely entailing a feedback or a return to initial condition, and as such coinciding with reason as a procedure that responds to the contingency of the world, or exteriority that the system needs in order to limit itself. The indeterminacy of the external world becomes part of the recursive loop of adaptability.
As much as the instability of nature or the problem of contingency enter the realm of systemic philosophy, so too a principle of regulation defined by feedback seems to grant ontological certainty to reflective thinking. As Hui points out, it is the attempt to know the unknown without really knowing it that constitutes the spirit of organic thinking from Kant to cybernetics. end of quote. Who he accounts for contingency beyond probability also points to the need of engage the unknown as the incalculable, i.e., beyond the paradigmatic universality of computation.
On the other hand, however, the recursive function of reflective thinking is a schema that entails the preservation of decision and the negative subjection and construction of the space of externalities. In other words, I argue the question of the unknown is a problem for power and as such it cannot be disentangled from transcendental decision and from the recursive onto epistemology of racial capitalism, colonialism and patriarchy. From this standpoint, the automation of reason may most importantly expose how the interiority of philosophy, this authority of decision, as La Ruelle calls,
does not depend on a transcendental schema, but rather depends on the means, the procedure, the instrumentality that transduces unknown exteriorities into a probability. Here, a final set of algorithms becomes what I want to call a non-instrumental instrumentality. The while carried the program of humanity always for short of being, always for short of ontological completeness. This generic mode of operation offers a non-teleological mode of reason, based on incompleteness and concerning the singularization of fractalities of the real.
One can argue that this negative instrumentality has haunted the project of humanity since the theorization of the incompatibilities of the machine, through which the colonizing racial and patriarchal project of humanity continues and wants to continue its origin story and the origin story of reason. As Zakia E. E. Jackson argues, the project of humanity is a racializing project, an imperial imposition that installs hierarchies by relying on blackness as, and quote, infinitely malleable lexical and biological matter, end of quote. That affective, social, creative capital is what can
be abstracted, extracted for free today, is indeed part of the humanist program of capital accumulation, of enslaved and dispossessed black and native lives, what Louis Choudes Soke understands as the prototypes of the servomechanics figuration of the cyborg. This chimera, as he argues, is a posthuman configuration of aesthetics and ethics that carries within itself the hierarchical order of the human, the imperialism of being, which coincides with this kind of equation of value between animal, human and machines.
It is this paradox of colonial and racial cosmogony where recursive algorithms as the prototype of the slave machine continue to sustain the freedom of what Winter calls the home of bioeconomicus that remains a central problem in the theorization of machine thinking and AI aesthetics. And this main problem that I am attempting to tackle. But how, to what extent, can the exteriority of the machine, non-instrumental instrumentality, overturn the double bind of this paradox that confirms, it seems over and over again to confirm the authority of transcendental reason in AI?
eye. How to unbound the instrumentality from the paradox of creative racial capitalism and ultimately from the ontological project of humanity where Technet can at best show us a way back to the Heideggerian poiesis, an ontological past turning a blind eye on the instrumentalities of racial capitalism instead of bracketing off instrumentality i want to attempt to argue that the exteriority of techno flash and its social technogenic abstraction is rather um yearned an instrumentality that insurgents everywhere against the self-reflective recursivity
or decisional systems. In what follows, I will draw on what I call this word in computabilities, a instance of instrumentalities in AI that disentangle the limits of computational programming from philosophical categories and aesthetic synthesis. What this limit of computation, which is the incomputable one can argue is the space of non-standard instrumentality aiming to abolish the authority or the sufficient reason that grants their dominance. Here AI exposes techno-scientific principles of pure randomness, non-linear complexity,
Norse Incomputability, Incompressibility, Incompleteness, which I argue are the aesthetic colliders, the condition automated reason to undo the given program of transcendence. It is in this new, in this principle of in the limit of computation that an alien hypothesis of AI comes to coincide with a speculative machine, with the non-performance, with the non-being. In other words, I suggested in computables, even if they are without ENDS, i.e. without teleological program, afford the non-performative condition of programmability, and AI without transcendence. But how to expose this in computability in generative AI?
How can what I started with, example of DALI E, withdraw from the servomechanic function of algorithmic transparency and generate counterfactualities or what can be called negative reasoning? I mainly corresponded to hypothetical or abductive hypothesis. What if? What if other words are there and are not really visible? The OpenAI program DALI-E is a neural network, which generates images from text captions from a wide range of concepts expressible in natural language. From text descriptions using data sets of text-image pairs, it is possible to combine unrelated concepts,
rendering text and apply transformation to existing images. The power of DALI-E stems from using natural language as a means to train deep learning models from the massive amount of paired natural language image data that is available on the Internet and their noisy uncurated, unpatterned information. Importantly, DALI-E is also a transformer language model, i.e. a program that receives both the text and the image as a single data stream. But how does DALI-E and DALI-E2 know how a textual concept like, for instance,
alienness can be manifested in the visual space? The link between textual semantic and their visual representation is learned by another open AI model called CLIP or contrastive language image pre-training. CLIP is trained on hundreds of millions of images and their associated captions, learning how much a given text snippet relates to an image. This how much is very important. I will go back to this quantity and computability in a second. That is, rather than trying to to predict a caption, given an image, clip instead just learns how much related any given caption is to an image.
This is contrastive rather than predictive objective, which allow clip to learn the link between textual and visual representation of the same abstract object. The entire DALI E2 models, model hinges on clipability to learn semantics from natural language. This is important because what ultimately determines how semantically related a natural language snippet is to a visual concept, which is what actually becomes a text-conditional generated image. After training, the clip model is frozen and Dali-E moves on to the next task.
Learning what it does, it learns to reverse the image encoding mapping that CLIP has just learned. So CLIP learns a representation space in which it is easy to determine the relatedness of textual and visual encodings. But this representation space becomes exactly what is used for generation, which means that OpenAI adds another model to this model, which is called Glide, to perform image generation. What Glide does is that it inverts the image encoding process in order to stochastically decode the image embedding. So Glide does not build an autoencoder or exactly reconstruct an image from its embedding,
but rather generates an image which only maintains the salient feature of the original through a stochastic random process. So what happens that DA-DALI-E works by a compression of large data sets, or what is called an increasing amount of information, an increasing volume whose size is exactly randomness. that instead of expressing error, becomes generative of what the image that does not exist will do. And one interesting thing of DALI is that he actually takes inspiration for thermodynamic
and therefore translates the problem of noise into a problem of computation. So what is interesting for me is the use of diffusion model. So diffusion models are generative models inspired to non-equilibrium thermodynamics to generate data similar to data on which they are trained. So we saw all this procedure before about how things get encoded, decoded, but this coding and decoding cannot happen without noise. And that's very interesting because obviously noise, the adding of noise to the passage of a message is at the very core of information theory.
If you know anything about information theory, you know that there can be no message without noise. Actually, noise is what facilitates the message of information. So, importantly, diffusion models work by destroying training data through the successive additional Gaussian noise, and then learn to recover data by reversing this noising process. After training the diffusion model, it comes to generate data by simply passing randomly sampled noise through the learned denoising process. For example, an image generation model will start with a random noise image. And then after being trained to reverse the diffusion process on images, the model will be able to generate new images.
So this is a technique that implies first a perturbation of each image in the training data set with increasing levels of noise. And then the neural net will be able to generate an original image using gradients of the distribution through the denoising of the image. So this reversing of noise and denoising can show how means are yet again subjected to the decisional structural representation. They yet again need to be negated. There's yet again a negation of noise that also exposes how the incompatibility of instrumentality has to be cancelled out. Here, the end result is indeterminate, not because the image is unknown,
but because the AI generates what could be known, such as a counterfactuality, something that emerges from the compression of noise abiding not to a digital binary decision, yes or no, but to the diffuse variation of quantum reasoning. By traversing backward from noise, new patterns, i.e. non-existent information, is generated as if it already existed in parallel dimension. So the information is always there, but it's only picked one at a time and as a whole. So automated reasoning here coincides with retroactive speculation or as Laruel also calls you would have called it with
filofiction where noise becomes a mode of elaborating alienness or corresponds to what it calls a full and positive void so as much as noise is weaponized right so that's obvious in this system, there's no romantic view of that lay, I hope we can agree on that, that is continually turned into patterns that are computable. There cannot be compression without the exteriority of machine thinking, without the alienness of noise from the world of human recognition. Incomputability must be negated, must be cancelled out, blacked out, noised out.
It is indeed this recursive function of reversing noise into patterns and patterns into noise that becomes the space for non-instrumentality in automated reason. It is as if in order to produce a higher resolution image or what can be called algorithmic transparency, the entropy of unpatterned noise is needed as a condition that must be negated from the non-equilibrium or uncompressed noise that AI generates. So if this compression of noise is central to text to image machine vision today, it is also central to very much talked about chat GPT,
a generative AI which also is based on large amounts of large language models. and mainly designed to have conversations with human, confuse music, become your therapist, anything, write poetry. Similarly to DALI-E, this generative AI is trained to look for statistical regularity and not exact sequences of bits. It outputs approximations rather than accurate facts. If truth, facts, and accuracy are missing in generative AI, it is because such PT compression in large natural language model yet again must release information entropy.
Here noise cannot be converted into a pattern, or when it is, something is lost. Namely, some information becomes unrecoverable, and what is approximately there becomes part of the next prediction. But what is predicted is not what is programmed. Instead, again, counterfactual responses point to how automated reasoning works through a blurry information. That's the technical term they use. Paraphrasing and rephrasing material from the web instead of quoting things word by word. The inability to reproduce facts, the inability to give us truth, apparently, coincides with the ability of learned complexity.
Here learning does not entail the reproduction of trained data, but the ingression of incomputables into predictive intelligence. As much as chat GPT retains much of the noise of the web, it thinks through hallucination instead of factual questions. While hallucination may refers to some perceptual or a phenomenological distortion of reality, function instead coincides with compression artifacts, with wayward fabulation or counterfactualities emerging from what is missing, i.e. from what is negated at each step while predicting what is on the other side of the gap. For instance, when an image program is displaying a photo and has to
to reconstruct a pixel that was lost during digital compression, it looks for nearby pixel to calculate the average. This is what chat GBT does when it is prompted to generate, for instance, an image of, say, protest slogans in the style of Antonin Artaud's stuttering in relation, for instance, with an image of crowd dynamics. It will take two points in lexical space. It generates a text that would occupy the location between them. It will, anyway, somehow generate a form of interpolation that relies on a blurring, or relies on this zone of noise, of entropic augmented noise of instrumentality.
So, while generative AI maps statistical regularities in large language models, it is also a non-performative performance. It cannot do without what is called the loss function in algorithmic compression. In other words, it is a non-performative procedure because it seems not to reconstruct an image, an original text or an original concept, but rather colliding information across dimensions and categories, a mode of understanding that exposes fractions or fractalities into discrete analysis. In other words, it is an all incomplete thinking.
That is what Stefano Arney and Fred Mauder call incompleteness is a weapon of theory, theory is a weapon for theory, a collective theorizing that has no individual subject, no property and no ontology. Non-instrumental instrumentality for me, therefore, it exposes not the creative image of artificial intelligence according to the cosmogamy of men, but the fractal or incomplete collectivity of the techno-flesh, abstracted with machine reasoning and resulting in the non-axiomatic computation. This is artificial intelligence that cites will and pushes further the negativity of the human,
the inhuman, the non-human, the unhuman, as it occupies the stance of, and quote from Laruel, a stranger as the non-positional, the non-donational body of oneself, who instantiate from the ego to the world and no man land. End quote. Laruel arguments for no standard aesthetically, for me, radically defy the pretentiousness of Western metaphysics for which the real can be surgically cleaned from alienness, which must become quantified intelligent, which must perform the function of a soulless machine that must be converted to humanism. With no standard instrumentality, instead of the self-other, subject-object,
including the excluded diets, which are part of the same mirror game of transcendental philosophy, are instead overturned and replaced with a logic, a reason, without logos or alien patterning. Instead of serving as an instrument for shedding light into the world, algorithms clone the underworld of generative intelligence that has turned philosophical decision onto its head.
From this standpoint, if we were to follow this proposition that I'm putting forward here on non-standard instrumentality, machine aesthetics must manifest above all algorithmic fractalities, namely counterfactualities of automated reasoning, the accelerating cloning of noise, the superposition of discrete sets and continuous vector in an immanent field that presents the problem. primacy of incompleteness in reasoning. As much as the computational compression of noise never coincides with the real itself, it describes AI as a non-medium, above all a non-relationality
with humanity and non-relationality with the world. Beyond the equation of value between humans and machines, non-standard instrumentality can become the focus for aesthetic practices that refuse, hack, alienate, reverse transcendental decisions, but also the ontology of the sensible, for which machines are designed to perform, end quote, the feeling of humanity. Machine aesthetics must remain non-performative, breaking the expectation of being someone and doing something for philosophy. It must remain an heretic project that takes the aesthetical ladder of flesh and noise in algorithmic patterning as a generative
diffusion of a wayward reason, proliferating each at any time, outside and against the ontopistemology of men. Thank you very much for your attention. Thank you. Thank you so much, Luciana. That was a tour de force. It was amazing. And you covered so much. I said we covered so much. I felt like I was with you all the way. We do have some time for questions. If anybody wants to jump in, please do. If not, I will move us towards questions. Okay, we have one over here.
Thank you, that was really, really brilliantly clear. My question is, I've got a question and then another question. The question is, are you saying that non-standard instrumentality is a description of the function of metaphor in language. Because I was thinking as you were talking that one of the really banal things about chat GPT is that it's quite clear that whoever constructed it
doesn't understand how metaphor works and therefore how language works in a social system. because language is metaphor. All language is metaphor. And as you were talking, I was thinking that what you were describing reminded me a lot about Deleuze's description of the Baroque and the interiority, the impossibility of exteriority of the Baroque. which is required, according to him, I think, you know, quite rightly for a critique of capital,
which when I read it the first time I thought was a description of metaphor as well, like a robust, complete description of how metaphor works, the complexity of metaphor. Thank you for your question. Thank you for the audience, Sammy. This is a new work. I am colliding at many registers. And I think that I have a mission in my project to actually break away from a model of representation. so for me instrumentality is also a materiality a way to go back to this kind of representational model of technology for which we correlate an object to a concept
and actually stay with the function and the function is not just for me some kind of mechanicistic model or some kind of spiritual model, but it actually has to do with the articulation of thinking. It's exactly the articulation. So for me, when we talk about metaphor, I get a little bit sidetracked, because for me, the function is not to say that there is a direct relation between the function and the real, but the function is using the real also as material, as much as it builds itself. So there is no, you know, in the model representation,
we can kind of abstract from the real or from the material and have a symbolic model for which every concept can mean different things in different contexts. That's a model that I am not working with. I understand that model is something that one can refer to in terms of the play of the signifier and the model, the way of deconstructing language through the signifier. I'm much more interested in the actually syntactical function as it operates in almost an algebraic way. For this reason, I also think that even if they don't, when you say people don't, you know, the designer of these generative AI models and natural language processing has been going on for a long time within research in AI, just exploded now.
but there's a lot of precursor to that that are super interesting in terms of challenging the grammar of language already in terms of subject, object and predicate. That predicative grammar didn't work for reasons that are actually interesting because it didn't work because it's obviously a format grounded in a transcendental model of how grammar needs to work. to work, right? And how people speak and how language is, you know, rather than being just a series of concepts that infer meaning from the word, actually, are words and themselves material
in their reassembling and reconstitution. So it's a, I think that in that way, even if they don't know what language, who knows what language is. I think that we are still, you know, it's an open enterprise still, you know, even if we have a solid grammar and the grammar constitutes a relationship of power that are, you know, ingrained from this ontopistemological model I was trying to explain. And so at the same time, so as far as language is not just use, you know, in that Vitigestanian way, but it also generates, generates culture, generates a sonic, visual, aesthetic capacity of, you know, so it's something that is productive and is not inferring
something else right so that's one thing and then i i think the fault is interesting about um the interiority and the exteriority um i guess that in in whereas in uh in in in the last um you know um the the topological model of interiority and exteriority run the risk of not really um dismantling the model of capital right because for me the interior the fold even if i was super fascinated i wrote about it many times in many different ways but um it doesn't allow for a pure exteriority or for a runaway a stereotype apart if you take the lesson in the lines of fly maybe that could be a bit different when they talk about astra certification but the fault that that baroque
in that kind of mannerism that the the curve the the fold is the manner in which the interiority is exactly the recursivity, the self-reflective recursivity of cybernetic metaphysics that I think is important but doesn't allow for the exteriority to run away. So that's exactly interesting. Thank you for that. I could actually extend my critique of interiority with that. But thanks. Wonderful. Any other questions out there? Yes, there's one at the back. Great.
Hey, thank you so much. Oh, so loud. This is like really not my expertise. And so I hope that this makes some sense. But it feels to me like you're proposing the generative AI as a potential site for like a transhuman, extra human reasoning or consciousness to emerge that there would be some sort of rupture. which I find very hopeful as a concept. But what gives me pause with that is that the data sets that the AI are working with in my, again, super limited understanding are all still reliant at this point
on people at some point being like, this should be a data set. Like this is something that I believe is worth tracking and attending to, which feels like a significant limitation on that sort of contact with the outside. Yeah, thank you so much. So I want to specify that my critique of the human is central and the project of humanity is central to this work. So for me, instrumentality is exactly try to take away technology from the kind of dogmatic ideal belief of an extension of men through the transhuman.
For me, the transhuman is just an extension of men. It's an extension of, precisely what Silvio Inter criticizes, or brings to the fore as the constitution, she says men one and men two, there could be another men three that has been constituting itself since the Silicon Valley, we could argue, this kind of model of performativity of labor, of techno labor, but also ideal of transferring consciousness into the machine. So the machine is seen as a vessel whereby the consciousness of men can continue to do the work, and that actually does continue. So I'm not negating that, right? So I'm not just saying, oh, forget all that.
We are actually the generative AI is the solution. I'm just trying to break away from the overrepresentation of this kind of technological form every time they are presented to us as a transhumanist project. So I will definitely go and stay on the possibility of wanting to really work with the possibility of technology to negate the transhuman, to make it completely fall apart, to actually show to the old kind of residue patriarchal models models of kind of repackaging the future into basically the same.
So whereas when I talk about the non-human, the anti-human, the unhuman, and the way those are these kind of spaces that are occupied by the radical difference of otherness or austerity. It's for me a way where the machine can actually bring a kind of questioning of instrumentality and what the machine can do to dismantle those kind of associations is productive. So there's something productive in there. So I'm not a futurist in the sense of, oh, yeah, let's look at technology as a way to save us,
not technology that has to dismantle us, if anything. That would be my hope. I mean, my hope. It's not even a hope because there is no hope here. Forget hope is much more of a collective critical practice at all levels amongst all of us who are working on that. Then I really totally agree with the data-driven model, because obviously it's all data-driven, and there's all this complaint, oh, but this is only data-driven, right? Just have a set, put another set, and there is no reason, there is no reflexivity, there is no capacity, because there has been this massive shift within AI, but also within this real critique between these kind of models of representation, a symbolic model that will tell the data what to do,
to an inductive model where there is no saying. The program is open. The program is just open to learn. But people say it doesn't learn. It just repeats and spits out exactly what the data is there. That's why you have biases. That's why you have, you know, racist, sexist, you know, Thai, Microsoft had to be shut down. You have continued. The search engine, which activity cannot really work because it's going to show exactly what these cosmogonists that we rely upon are about. So it's an interesting thing on that way. Oh, we need to correct. You can't correct. What can you correct? Where do you start to correct and make it more transparent, right? So in the kind of critical visibility,
they really make it visible, you know, all the substrate. you know, but it's interesting that instead of data, I think that my interest in automated reason is exactly to work with what kind of other intelligence and reasoning, a subject reasoning or a logic without logos can actually do. And I don't know whether that is visible or readable or grammatically correct, right? So we got to really think about a different kind of configurations and mapping a method to talk about it, but also to make it part of our collective political practices. So it's an invite, I mean, invitation, sorry.
Thanks. Yeah, Christine. It seems that the idea of the introduction of the notion of incompleteness and a sort of negative fault line is running through everywhere, every kind of philosophical perspective. and I think that is politically and socially absolutely crucial but sometimes I'm inclined to think that we're kind of going the other way that there was 300 years of it's all positive man and now everything's in the negative
but as if the negative can save us in the same way that the positive was going to save us before but my real question to you is not, you know, I think it's interesting to speculate that AI is yet another way in which incompleteness is going to be demonstrated and the West certainly needs to get a grip on that. But what kind of bothers me is why do we need computers to do this? I mean, everybody outside of the modern West, even the modern us, before we were modern understood that the world was incomplete, that it's such a tiny blip of humanity. But we don't need computers. You can do it with sticks and stones.
So I guess my question to you is do you think that computers are necessary to this or are you just saying, well, they're one way in which it can? Because I think there's a real danger of over-inflating that nobody ever understood incompleteness before AI came along. Yeah, no, I hope not. I think we've been talking about incompleteness, I mean, since the beginning of a critique, I guess. But I think there's no computer. I'm interested in computation. So for me, AI is just a form of computation. There are many other forms of computation, like weaving, like peddle stones together so computation is a way of compressing information, that's all generically I'm interested in
so that you can argue that this kind of computation has always existed because the compression of noise or the compression of infinity and what kind of infinity we're talking about is very ancient and it's not even culturally deterministic but it's been the discourses or the kind of articulation of technology has really taken a particular kind of capacity to represent everything. I am totally in agreement with you about the paradigm of computation should be, as the word that was before, the paradigm of neuroscience. So those paradigms are things that one needs to tackle
so that you can take that away from the paradigm and do other things with it. So you can't really pretend it doesn't exist. So, of course, computers exist, and, of course, media exists, and, of course, even without computers, we still have computation now. You know, take the computer away, we can still think computationally in a way that the binary digit model and program as a transformed way we make decisions or the way things are counted. If you take it all, that's always the interesting thing. If you take everything away and abolish everything, can you really go back to a tabula rasa, right?
So where does it all go? So I agree with you that computers won't tell us what incompleteness is, But I also think that the question of incompleteness or the negativity is not the same question. You know, I think that the questions that were posed before in relation to, I don't know, the negative in terms of the photography and the negative, the obscurity, they're not the same as the the obscurity about and the darkness of the photography that are in Jordan Peele get out. I don't think it's the same because things have changed
there's been fights and you know rebellion and killing and you know genocides for that question to happen right so I don't think it's the same I don't know if that makes sense but yeah but thank you if I may I can't help I just like I want to ask a question too um and I really I had a question around not just incompleteness but completeness uh and and you know this this series um is is really trying to think around um how images of say incompleteness are actually forms of completeness you know that are like kind of might be very tricky or illusory for us um and so yeah if you know what I'm trying
to say there so so what I was thinking about I couldn't help but think when you were invoking things like the sublime and I'm just going to bring it right to some really basic ideas of art now like historically so let me just say think about I don't know an image of completeness um A structural completeness would be like Turner painting the ocean sea and the, you know, you look at the big painting and it's got the waves crashing and someone's tied to the mast and all of whatnot. He's tied to the mast. So you get this kind of completeness between the structure of thought and the manifestation of the affect of that work. but of course the completeness itself is like sublime horror kind of moment like confronting nature and all in the in the you know incompleteness is manifest in the affectual state
so you've got like one idea to sublime there that you were kind of referring to and then you've got the kind of desire in later works in modern critique to explicate the futility of the sublime in the sense of saying um or bring the sublime back through a conceptual schema you know um which is to, like Lyotard, for example, you know, in his new Sublime Realism, which would look at the different, the impasse, you know, of aesthetics and thought and propose that in a processional kind of artwork, like the repeated labor of a process-based performative practice, right? So you've got like two different practices there. And then, you know, when you were talking about how a kind of aesthetic encounter
can bring to us some kind of resonance of incompleteness. And it kind of makes me wonder, you know, about aesthetics today and, you know, about the whole of your talk in the sense that, and I don't really know how this is going to manifest, but you're making me think of, like, are you requiring, like, underneath all of this or even over it to ask us to, like, not do aesthetics at all in the sense that, you know, of an an aesthetics for example or a non-aesthetic um in the sense that when we look at some of these very kitsch images that um dali produces you know like the worst kind of novelty kitsch renditions of the pseudo imagination of the computational and whatnot um you know they're
as bad as a lot of artwork i see today and you know you can't tell the difference that much And so, you know, you could say there's a lot of kitsch thinking out there by artists and computers. And it seems to me you're after something else. And so what I'm asking really is, are we to kind of discover in our sensory and conceptual experiences when we look at an image? are you asking us to see differently and to maybe think beyond the thing that we're we're seeing yes that's what i'm asking you and in that case does that kind of resonate with a kind of anti-realist brechtian practice or where does it i'm looking for a model like
i don't know what yeah does that make sense so what like an anti-realist brechtian you know like the explication of the structure that subtends the image here it is in front of you here's the actors acting you know um and i know you're not saying that but i'm saying is it a kind of that or what what are you asking us to if it's not the ocular kind of i'm looking at an image and it represents this and instead i'm looking at a complex in front of me and and we look at an artwork today and go, I'm looking at a complex dynamic of stuff. You know, can we apply that to any artwork or does some artwork just go no to that? You know, is that a way? Is that a new art criticism?
Is that a new art product? You know, what does it do to us as makers and as watchers, lookers of art? Yeah, no, thank you. Sorry, it's very long. No, it's very long and it's very interesting because I want to ask you questions. I think it's very interesting to think about this different form of the sublime. Because obviously, the sublime is the kind of, for me, it's interesting as when the limit of knowledge becomes actually conditioning of knowledge. When it's conditioning of knowledge, it changes the game. Because in the same way, so in the same way as we have done this critique of, capital needs this limit, right?
And this limit is the zero value of labor. Whatever the capital can abstract from aesthetic, affective, flesh, enslaved, land, everything is abstracted as the limit of its own subsistence and at the same time is also given no value. So it's this game that the limit is what capital must constantly deny. And then through patriarchy, racism, and these modes of exclusion through which it kind of pretends that these things have no value, although are fundamental.
So too, the sublime is like that for transcendental reasoning. It's the moment that is needed. But then if it's needed, then if you open, you know, if you break this enchantment with the sublime, that's what I think I want to do, break the enchantment. Because for me, the enchantment is too much implicated into this kind of romantic appeal, hope, prometheus emancipation. The enchantment has got to do with this kind of promise that things will be better. Whereas actually, instead of enchantment,
if you break the enchantment, it's not just, oh, now I see the truth. No, you continue to see darkness. You stay there. don't move from where you are because that's where the thing is that's where the real is for me so it's not an anti-realist in any way it's actually a real that cannot be enlightened and it cannot be even in its discreteness and that's why I'm interested in this algorithm with the noise thing because they always carry with them and you can say there is that residue of the trace of the home thing? Not really, right? Because it never stays. It actually comes to the front. It's first. That's what I mean. It changes the game.
The darkness comes before. And at that point, it's not just about enlightening how the algorithm works. If we see, because that's transparency, what you said to me in terms of, oh, is this just about how we're supposed to look? So you deal with the medium. Okay, I'm dealing with the computation. It's my medium. That's my medium. So I go there and I look at all these things, how they work. I spend so much time reading, making diagrams, asking people. It's a lot of work. I'm obviously not a computer scientist, you know, and that's not my job. My job is different. But I don't want to just elucidate and make it transparent. So if we do this and this, then we can make this algorithm work very well and then everyone will be represented. And then this could be good for humanity.
and we can consume less water and we can have, you know, that is all stuff. Wonderful people do it. That's not my job. My job is to stay with the alien of the machine. So that's the real for me. And that's why I can never coincide with the pattern. The patterning is always an instrumental mediator, mediator, transducer, intensifier of that zone. That's wonderful. Well, unless you've got a very quick question, I'll just look around because I know we're over time, but I felt guilty because I asked a question that I shouldn't have done. But in that case, I will, since there are no questions,
I just want to say, again, April the 10th is our next event, so please come along. It's our last one from the MA Aesthetics and Politics Programme at CalArts, just to remind everyone what we're doing here. And I just want to thank Asuka, Tanya, and everyone at the ICA for helping us, and also Jackie for coordinating the tech and others tonight. It's been very helpful. Thank you very much. And finally, thank you to you for coming. Really appreciate it massively. And also, of course, to Luciana for this amazing talk. Thank you so much. And see you for drinks. Thank you. Thank you very much. Thank you very much.