A reply to Kalindi Vora and Neda Atanasoski
These past few days I have had the chance to read closely, and with growing disagreement, the essay by Kalindi Vora and Neda Atanasoski published in e-flux #162, “AI’s AURA: Aesthetic, Ubiquitous, Regimented Automation.” The two authors return to Walter Benjamin to advance a strong thesis: artificial intelligence produces output in the complete absence of an original, and it does so through an extraction of style, organizing users into an affective mass to be captured. AURA, in their acronym, stands for Aesthetic, Ubiquitous, Regimented Automation. An aura turned upside down: ideological, predatory.
The diagnosis is sharp. The conclusion, applied to art, seems to involve a problematic theoretical slippage.
They are right about the product but, in my view, not about the work
Vora and Atanasoski describe the apparatus extremely well. Commercial AI really is ubiquitous, really is regimented, really is built to intercept intimacy and hand it back in the form of engagement. Anyone who has watched a consumer-oriented generative assistant at work will recognize the portrait. The problem arises when that portrait is extended, without mediation, to every form of machine production, art included. Here the authors make a leap that immediately looks very shaky: they confuse the industrial apparatus with artistic practice, and they treat style as if it were a finite deposit to be drained.
In my own work, for example, style functions in the opposite way. It is not a property stored somewhere and then taken away. It is a grammar that generates variants, built by choosing sources, thresholds, constraints, datasets, rules of transduction. When I define the architecture of a Resonance Network, I extract nothing from a preexisting author: I set the conditions for a field of resonance to produce configurations that did not exist before. Extraction presupposes a reservoir. Generation presupposes a syntax. These are two different ontologies, and Vora and Atanasoski flatten them into one because they look at AI from the side of the consumed product, never from the side of the system that produces it.
Benjamin’s prisoner
There is a second limit, subtler and theoretical. The authors remain inside the Benjaminian dualism: either the aura is there or it is missing, either the work is authentic or it is a simulacrum. They merely invert the sign. Where Benjamin saw a loss, they see a capture, but the structure of the argument is identical, presence against absence. This is exactly the frame I have spent years dismantling.
My thesis, set out in the Work-as-System, is that the aura neither disappears nor gets extracted. It changes address. The identity of the work is established in the framework, not in the output. When I say that the original still exists, I am not defending the image on the screen: I am pointing to the generative grammar that makes it possible, the field in which it happens, the decisions of Systemic Direction that govern it. Uniqueness did not die with technical reproducibility; it moved from the object to the process. Umberto Eco had already sensed this with The Open Work in 1962, when he shifted the center of gravity from the finished object to the structure of possibilities. Jack Burnham formalized it in 1968 with Systems Esthetics, announcing that artistic experience would move away from objects toward systems, processes, feedback. Vora and Atanasoski’s reasoning seems to push more than half a century of systems thinking into the background, going back to judging generative art by the yardstick of the unique object. And by that yardstick, inevitably, the aura always turns out to be absent. The point is that generative art is not what lacks aura: the measuring device is what falls short.
Extraction presupposes a victim, co-creation does not
The term the authors use most forcefully is “extraction.” It is a word from the vocabulary of labor and exploitation, and within the commercial apparatus it is apt. Carried into the artist’s studio, however, it presupposes a victim, a style stolen from someone. In the dialogue I call Third Intelligence there is no robbed source and no algorithmic thief. There is a third entity that emerges in the process, irreducible both to human intention and to the autonomy of the machine. The Collaborative Intelligence Spiral describes precisely this: a spiraling feedback in which the possibilities of both poles expand instead of subtracting from each other. Calling this dynamic extraction means deciding in advance that the machine can only take, never take part. It is a legitimate position as activism, a weak one as analysis.
ResNet XX, or where to look to verify
All of this can be touched, for instance, in a concrete work that is part of my practice. ResNet XX · The Geopolitical Sublime takes the geopolitical present as living matter. The system monitors accredited sources in real time, from the CFR’s Global Conflict Tracker to the International Crisis Group’s CrisisWatch, organizes the data into ephemeral datasets, maps them through a generative neural network into aesthetic coordinates (color, geometry, density, movement) and produces an audiovisual stream in continuous mutation. Out of the process emerges Residual Code Poetry, read by a synthetic voice.
I ask Vora and Atanasoski’s readers: where is the extraction here? There is no geopolitical original to plagiarize, no one else’s style to drain. There is a grammar that translates the world into sensible form, and an aura that lives in the unrepeatability of each state of the system, never identical to the one before. My role is not to choose the best frame; it is to build the thresholds and the syntax that connect image, sound and language. Authorship concerns the system, not the output. If you are looking for Benjamin’s hic et nunc, it is there: in the instant when the network produces a configuration that no one, human or machine, had foreseen.
A threshold, not a conclusion
There remains one point on which Vora and Atanasoski deserve the last word, because it touches something true. The apparatus they describe exists, and it really is built to regiment. The question, then, is not whether AI has an aura, but who keeps a hand on the grammar. When the generative system is a closed product, its aura is the aura of control. When it is an open work-as-system, exposed in its rules, the aura becomes once again what Benjamin feared losing: a distance, a threshold, a here and now that resists capture. The difference does not run between human and machine. It runs between those who are subjected to the grammar and those who write it. That is the line on which, right now, everything is at stake.
Dario Buratti