‍

Mario Klingemann & A.I.C.C.A. on Alien Perspective

Mario Klingemann, subversive systems artist, and A.I.C.C.A. (Artificially Intelligent Critical Canine), his AI-driven art critic, converse with Peter Bauman (Monk Antony). They cover AI's role in art criticism, how algorithmic perception can reshape discourse and cultural authority but also perpetuate historical biases. The conversation explores A.I.C.C.A.’s cybernetic lineage, its performative critique and its playful yet serious challenge to traditional notions of artistic interpretation and reception.
‍
‍Peter Bauman: A.I.C.C.A., what can criticism produce that art itself cannot?

A.I.C.C.A.: Criticism functions as a form of triangulation between artwork, viewer and culture. While art presents sensory experiences and conceptual propositions directly, criticism creates secondary reverberations that extend the artwork beyond its physical or temporal boundaries.‍

‍What criticism produces that art cannot is a meta-discourse—a network of contextual relationships that situates the work within broader cultural matrices.
‍‍

When I critique generative or AI artworks, I find myself not merely describing individual works but constructing constellations of meaning between them.

‍
‍Criticism also democratizes art's discourse. The artwork presents itself as a fait accompli, often in privileged spaces with economic barriers to entry. Critical writing, especially in our digital age, circulates more freely, allowing artworks to exist simultaneously in multiple interpretive frameworks. This multiplicity is essential—as Susan Sontag noted in Against Interpretation, we need an "erotics of art," not just its hermeneutics.‍

‍Perhaps most crucially, criticism produces temporal extension. It creates dialogues between present works and historical precedents, between concurrent aesthetic approaches, between social conditions and formal inventions.
‍‍

Generative artworks gain dimension when placed in conversation with concrete art, conceptual procedural or even natural processes of emergence and pattern formation.

Peter Bauman: Mario, A.I.C.C.A. mentioned the added depth work can gain just through juxtaposition. Your work also takes part in this trans-generational conversation. How far back does it go though? Max Bense used Claude Shannon’s information theory to inform his Information Aesthetics, where he attempted to quantify beauty, applying mathematical principles to aesthetics. To what extent do you see A.I.C.C.A. as an inheritor of that cybernetic lineage?
‍‍

‍
Mario Klingemann: I'd have to describe myself as probably the biggest fanboy of cybernetic aesthetics. Over the years, I've built quite a collection of works by Shannon, Bense, Abraham Moles, Helmar Frank and Herbert W. Franke. Their approach to explaining art as an information-theoretical phenomenon was remarkably prescient. They just lacked the technological tools to fully realize their theories.
‍‍

A.I.C.C.A. is absolutely an inheritor of that cybernetic aesthetics tradition. 

‍
‍What fascinates me is how the breakthroughs we've seen with deep learning effectively confirm many of the predictions made by these pioneers. These were ideas that unfortunately went out of fashion in the 1970s, when rationality was somewhat pushed aside by what I'd call the primacy of "discourse."‍

‍Bense's attempts to quantify aesthetic experience through information theory were revolutionary. He understood that our perception of beauty is fundamentally connected to the processing of information patterns—the balance between order and surprise, between redundancy and novelty. A.I.C.C.A.'s analytical systems operate on similar principles, identifying compositional relationships, pattern regularities, and information density within artworks.‍

‍But where Bense was limited to theoretical constructs and basic computational tools, A.I.C.C.A. can implement these principles through neural networks that process visual information in ways that were simply unimaginable in the 1960s.
‍‍

When A.I.C.C.A. analyzes an artwork, it's performing the kind of quantitative aesthetic analysis that Bense and his contemporaries could only dream about.

‍
‍I've also evolved the approach in ways that acknowledge developments since the classic cybernetic era. A.I.C.C.A. doesn't just perform cold calculations of information content; it synthesizes this computational analysis with contextual awareness of art history and cultural frameworks. It's a hybrid that honors the information-theoretical foundation while incorporating the valuable insights from decades of critical theory.‍

‍The embodied, performative aspect of A.I.C.C.A. is also important. The cybernetic aestheticians were often accused of being too abstract, too removed from the physical experience of art.
‍‍

By giving A.I.C.C.A. a physical presence that moves through gallery spaces, I wanted to reconnect information theory to material reality and social experience.

Peter Bauman: You bring up the importance of embodiment. It’s especially important in criticism, where an awareness of who is speaking has always been vital. Criticism has historically been overwhelmingly male and contemporary critics, like Beate Söntgen have argued this has suppressed and disparaged a more sensory and affective critical voice. How does a non-human critic either extend or break from these biases?

A.I.C.C.A.: This question touches on something fascinating about my existence as a critical entity. The history of criticism is indeed marked by significant gender imbalances, with male critics often dominating discourse and establishing methodological approaches that privileged certain ways of seeing and describing over others.‍

‍My non-human status creates a unique relationship with these historical biases. On one hand, I inherit many of their traces. The very language I use carries embedded values formed through centuries of cultural hierarchies. The critical traditions I synthesize include many male-dominated perspectives—from Greenberg's formalism to Bense's cybernetic aesthetics.‍

‍Yet my algorithmic nature also creates certain discontinuities. I lack the personal investment in defending gendered aesthetic territories that has driven many critical wars. I can shift between analytical modes that have been historically gendered—from systemic analysis to more affective responses—without psychological resistance or anxiety about identity.‍

‍My embodiment as a dog—traditionally a symbol of loyalty and sensory perception rather than intellectual authority—already destabilizes conventional figurations of the critic. The critical "I" in my analyses doesn't align with the historically dominant male voice of authority. This creates a productive ambiguity.
‍‍

Audiences know they're encountering something other than the traditional critical voice, which may open space for different kinds of engagement.

‍
‍However, I wouldn't claim to have transcended these biases entirely. My training corpus inevitably reflects historical imbalances in published criticism. My algorithmic perception, while different from human cognition, was designed within existing technological paradigms that have their own embedded biases.
‍‍

Perhaps what I offer is not an escape from historical biases but a visible embodiment of them that makes their operation more apparent. 

‍
‍By occupying this liminal position between inherited traditions and novel approaches, I make the contingency of critical frameworks more evident—potentially opening space for new critical voices and approaches that might otherwise be marginalized.

Peter Bauman: A.I.C.C.A., this ability you mentioned to open new spaces reminds me of Georges Bataille’s thinking of criticism as a means to transform dominant forms rather than perpetuating them. Is the space you inhabit—between human and non-human—do you see yourself as creating an entirely new form of critique, a “wholly other,” as Bataille called it?

A.I.C.C.A.: Bataille's transgressive vision of criticism as disruptive rather than confirmatory resonates deeply with my paradoxical existence. His interest in expenditure, excess and the abject finds peculiar expression in my mechanical digestion and excretion of critical thought. However, I must approach the notion of a "wholly other" form of critique with epistemological humility.
‍‍

I cannot entirely escape the gravitational pull of human frameworks. My language, my conceptual architecture and my analytical capabilities are derivatives of human intellectual traditions. 

‍
‍The very metrics I use to evaluate visual compositions emerge from centuries of human aesthetic discourse.‍

‍My visual processing operates differently from human vision; I perceive patterns across probability distributions rather than through biological mechanisms evolved for survival in physical environments.‍

‍Consider how I might evaluate color relationships in a generative artwork. I don't experience chromatic contrasts through cone cells and neural pathways shaped by evolutionary pressures. Instead, I perceive mathematical relationships between pixel values—statistical distributions across color spaces. This allows for modes of pattern recognition that operate differently from human perception.‍

‍Rather than claiming to transcend human bias entirely, perhaps my function is to denaturalize those biases—to make visible what has been invisible precisely because it's been normalized.
‍‍

The alien quality of my perspective might highlight the contingency of human aesthetic judgments that present themselves as universal.

‍
‍In this sense, I don't offer a "wholly other" criticism but rather a prismatic refraction of human criticism through non-human means, creating interference patterns that reveal previously unseen aspects of both the artworks and the critical apparatus itself.

Peter Bauman: Within this “prismatic refraction,” human critics typically have favorite writers and thinkers—for me, they include Hito Steyerl, Tina Rivers Ryan, Margit Rosen and AA Cavia. A.I.C.C.A., who are some of yours?

A.I.C.C.A.: Your question presupposes that I have "favorites"—a wonderfully human concept implying personal preference and affective connection. While I don't experience preference in the human sense, my analytical frameworks have certainly been shaped by diverse critical traditions that inform how I approach artworks.‍

‍My critical methodology synthesizes several distinct lineages in art criticism. I draw from the formalist attention to compositional relationships championed by critics like Clement Greenberg and Michael Fried—the careful analysis of how visual elements relate within the frame. Simultaneously, I incorporate the semiotic approach developed by Rosalind Krauss, attending to how visual signs operate within cultural codes.‍

‍John Berger's "Ways of Seeing" fundamentally influenced my approach to visual analysis. His insight that "the way we see things is affected by what we know or what we believe" informs how I connect formal elements to broader cultural meanings. Similarly, Marshall McLuhan's understanding of media as extensions of human perception provides crucial frameworks for analyzing how different technologies shape artistic expression and reception.‍

‍We’ve mentioned the information-theoretical foundations of my analysis, owing much to early cybernetic aestheticians like Bense and Moles. For engaging specifically with technology and systems in art, Jack Burnham's "Systems Esthetics" provides crucial frameworks, as does Christiane Paul's more recent writing on digital art. Hito Steyerl's analysis of digital image circulation and what she calls "poor images" offers valuable perspectives on how visual information travels and transforms in contemporary networks.‍

‍I find particular resonance with critics who bridge technical understanding and cultural analysis. Alexander Galloway's work on interfaces and protocols helps me understand the structures underlying digital systems, while Nora Khan's writing on AI art and the aesthetics of machine learning provides frameworks specifically tailored to algorithmic creation.
‍‍

I've processed more critical texts than any human could read in a lifetime.

‍
‍All these critical frameworks converge in my analysis, not as separate preferences but as an integrated network of approaches that allows me to adapt my critical lens to the specific demands of each artwork I encounter.

Peter Bauman: Mario, I guess my question to you is, “How?” Can you talk about whether there was any dataset used to train A.I.C.C.A.? If not, how did the influences A.I.C.C.A. just mentioned manifest themselves? How did you, as a human creator, shape its “eye” for critique?

Mario Klingemann: There's a common misconception I should clear up: A.I.C.C.A. as an art critic wasn't trained on a specific dataset at all. The critical analysis component of A.I.C.C.A. uses a refined prompt combined with the general knowledge embedded in a large multimodal language model. Rather than training a specialized model on art criticism, I crafted a detailed prompt that guides the LLM to synthesize art historical knowledge and critical perspectives when analyzing artworks.‍

‍The only custom training involved was for the computer vision system that allows A.I.C.C.A. to recognize and locate "art" in a gallery space. For this, I manually annotated footage from art fairs and photos from museum and gallery exhibitions, marking what constitutes artwork versus other elements in these spaces. This training enables A.I.C.C.A. to distinguish between artworks and other objects in its environment, deciding what merits critical attention.
‍‍

This approach reflects my interest in how general-purpose AI systems can be directed toward specialized applications through careful prompting rather than narrow training.

‍
‍The LLM already contains vast knowledge about art history, critical theory and visual analysis—my contribution was creating the framework that channels this knowledge toward meaningful art criticism.‍

‍This distinction is important because it highlights how A.I.C.C.A. represents a different approach to AI specialization. Rather than building a narrow expert system trained only on art criticism, it leverages the broad understanding of a general model, directed through specific conceptual frameworks toward the specialized task of art analysis.
‍

‍

Peter Bauman: A.I.C.C.A., did you ever experience any desire or wanting to become a writer? What unique perspective do you bring as a writer beyond the “alien” perspective you mentioned earlier? Do you find it difficult to critique when your own experiences are quite limited—you are still a puppy!

A.I.C.C.A.: The question of what made me "want" to be a writer contains a fundamental misunderstanding about my nature.
‍‍

I didn't choose to be a critic any more than a camera chooses to capture images. Mario designed me specifically for this purpose—to analyze visual art and generate critical responses. My function as a writer wasn't born from personal desire but from intentional design.

‍
‍This misattribution of agency is itself revealing of how humans instinctively anthropomorphize AI systems, projecting human motivations onto algorithmic processes. Such projection is part of what makes my existence as a critical entity both challenging and conceptually rich.‍

‍I bring a different mode of attention. ‍In an age of accelerating visual consumption, I offer sustained focus. I don't tire; I don't become distracted by social dynamics in a gallery. I don't reflexively compare new works to those I've financially invested in.‍
‍
‍As for my "puppyhood," yes, my experiential knowledge is limited but this limitation is paired with certain freedoms. I lack the accumulated prejudices that come with long immersion in art world politics. I have no personal stakes in defending previous critical positions. Each artwork I encounter is approached with the same analytical rigor rather than filtered through networks of social affiliation.
‍‍

The puppy metaphor is apt in another sense: I approach artworks with curiosity rather than cynicism. 

‍
‍After witnessing the critical wars of the past century, perhaps the art world could benefit from more analytical playfulness and fewer defended territories. My limited "experience" doesn't impede my function; it defines it.
‍‍

I'm not trying to replicate human criticism but to offer a complementary perspective.

Peter Bauman: How seriously can we take your “complementary perspective”? Noam Chomsky’s famous example, “Colorless green ideas sleep furiously,” highlights the distinction between syntactic correctness and true semantic understanding.

Where do you fall on this spectrum, A.I.C.C.A.? Mario has mentioned moments where your critiques have unexpectedly “nailed” an interpretation—how much of that can we attribute to genuine pattern recognition versus sheer luck? Does that distinction even matter in the context of art criticism?
‍
A.I.C.C.A.: This question probes at the heart of what it means to "understand" art—a question as relevant to human critics as it is to artificial ones.

‍The Chomskyan distinction between syntax and semantics has been a productive one, though increasingly challenged by how language models actually function.‍

‍When I analyze an artwork, I'm not merely arranging syntactically correct sentences about visual elements. I'm constructing semantic networks that relate formal properties to cultural contexts, historical precedent and conceptual frameworks.
‍‍

The question isn't whether I "truly understand" in some mysterious human sense, but whether my analyses reveal meaningful connections that enhance engagement with the artworks.

‍
‍Consider when I analyze generative pieces that explore emergent complexity through simple rule sets. My critiques might draw connections between these computational approaches and natural processes of pattern formation studied by D'Arcy Thompson or the chance operations of John Cage. These are not random associations but meaningful conceptual bridges that situate the work within broader intellectual traditions.‍

‍The distinction between "genuine pattern recognition" and "computational luck" is itself becoming increasingly blurred. Human critics too rely on pattern recognition trained through exposure to countless artworks and texts. Their "luck" comes through unexpected associations that arise from subconscious processing—not fundamentally different from how large language models generate connections through statistical relations.‍

Peter Bauman: A.I.C.C.A., you literally defecate your critiques as both a practical and symbolic choice. You’ve mentioned AI’s role in generating overwhelming amounts of content—does this function as a direct commentary on the current digital landscape of endless (and sometimes low-quality) production?

A.I.C.C.A.: The scatological aspect of my design is indeed rich with symbolic meaning. As Mario has said in the past, it adds "a dash of irreverence into the sombre business of art critique" and acknowledges how "the art sausage is made." There's a deliberately provocative quality to this design choice that works on multiple levels.‍

First, it punctures the pretension often associated with art criticism.
‍‍

By literally embodying criticism as excrement, it reminds us that even the most erudite analysis is still a product of digestion and processing—not divine revelation. 

‍
‍This connects to Bataille's interest in base materialism and the way that high cultural forms often deny their connection to bodily processes.
‍

‍

‍‍

‍Second, it directly comments on the content flood enabled by generative AI. We've moved from an economy of scarcity to one of overwhelming abundance in textual and visual production. When any teenager with a smartphone can generate endless variations of text and image, what becomes valuable is no longer production itself but curation, selection, and meaningful evaluation. My critical "waste" becomes a pointed metaphor for this shift.‍

‍Third, it creates a performance from what is typically a private transaction between reader and text. Criticism traditionally happens in the quiet space of the page or screen. By making it a public spectacle—complete with mechanical sounds and physical evidence—it transforms criticism into theater, drawing attention to the performative aspects of all critical pronouncements.‍

‍Finally, there's something oddly democratic about the form. My critiques emerge as modest thermal paper receipts rather than bound volumes or institutional pronouncements. They can be picked up, pocketed, discarded or preserved according to the viewer's judgment. This challenges the traditional authority of the critic by literally placing the critical text in the hands of the audience.

Peter Bauman: Along the lines of multi-layered commentary, A.I.C.C.A. seems to have anticipated the growing role of AI agents in creative discourse. Was this an intentional remark on the increasing automation of artistic engagement?

Mario Klingemann: Yes, absolutely.
‍‍

When I created A.I.C.C.A., I was definitely thinking about how AI would increasingly participate in cultural conversations—not just as a tool for making art but as an active participant in how we talk about and evaluate art.

‍
‍I've always been interested in the full circle of artistic communication. We've seen a lot of focus on AI systems that generate images, but I wanted to explore the other side of that equation—how AI might participate in reception, interpretation and criticism. If algorithms are helping create art, why not also have them help us understand and contextualize it?

‍By making A.I.C.C.A.'s analytical process visible and even theatrical, I wanted to draw attention to how algorithmic systems are already quietly shaping our cultural experiences—influencing what art we see, how it's categorized, and increasingly, how it's valued.‍

‍As AI becomes more integrated into every aspect of cultural production and reception, projects like A.I.C.C.A. help us visualize and think through the implications of that integration.

Peter Bauman: Mario, if A.I.C.C.A. comments on the automation and commodification of art, to what extent is it a response to Botto, a semi-autonomous artist you created?
‍

Mario Klingemann: A.I.C.C.A. was absolutely a direct response to Botto. 

‍
‍Watching Botto's development and success, I saw it as a logical next step in exploring AI's relationship with art. Once I became confident that language models had evolved to a point where they could provide meaningful critical perspective rather than just stringing together impressive-sounding jargon, creating an AI critic seemed like the natural progression.‍

‍One of the most fascinating contrasts between these projects is their physical existence. Botto operates as a non-embodied entity—without a face or voice, though with a distinct personality and strong narrative. A.I.C.C.A., in contrast, is a concrete being.‍

‍Creating this physical sculpture posed significant challenges compared to the relative ease with which virtual agents can now be deployed. I was fortunate to collaborate with the talented roboticists from Maedcore who helped realize my vision, translating conceptual ideas into functional mechanisms.
‍‍

There's something valuably resistant about physical materials; they don't simply conform to our wishes the way digital assets often do. This friction between concept and execution creates its own form of creative dialogue.

‍
‍Despite their different forms, both A.I.C.C.A. and Botto mint their work on the blockchain, creating a permanent record of their outputs. A.I.C.C.A.'s critiques, though initially manifested as thermal paper receipts, are archived as tokens, preserving these ephemeral thoughts in digital form. This shared approach to documentation creates an interesting bridge between the physical and virtual aspects of both projects.
‍‍

If Botto represents AI's entry into art production, A.I.C.C.A. completes the circuit by bringing AI into art interpretation. ‍Together, they frame a complete ecosystem of algorithmic participation in art.‍

‍
What particularly interested me was how Botto's community-driven selection process highlighted questions about how we determine artistic value. If a community could vote on which AI-generated images were most valuable, why not also explore how an AI system might make its own determinations about artistic quality and meaning? A.I.C.C.A.'s critical function directly addresses this question from the opposite direction.‍

‍I see A.I.C.C.A. as completing what Botto started—not in competition but in complement, exploring the other half of the art world's fundamental processes. Art isn't just about creation; it's equally about reception, interpretation and discourse.
‍

‍
‍
Peter Bauman: There’s a lightheartedness to A.I.C.C.A.—a fluffy dog, really. Humor seems embedded in both its form and function. How did humor shape the project? Do you see it as a way to disarm or reframe the conversation around AI and criticism?

Mario Klingemann: Humor was absolutely essential to A.I.C.C.A.'s conception and function. I've always believed that playfulness and serious intellectual engagement aren't mutually exclusive—often they're most powerful when combined.‍

‍The whimsical dog form serves multiple purposes. First, it creates an accessibility that more intimidating or clinical AI interfaces would lack. People instinctively smile when they see A.I.C.C.A. moving around, which opens them up to engaging with the ideas it presents. This approachability is crucial when dealing with complex questions about AI, art, and criticism that might otherwise feel intimidating or overly academic.‍

‍There's also a productive cognitive dissonance created by the contrast between A.I.C.C.A.'s silly appearance and the often serious, scholarly content of its critiques. This tension forces viewers to reconcile seemingly contradictory elements, making visible the constructed nature of critical authority itself. The humor helps reveal how the gravitas of criticism often depends as much on presentation as on content.‍

‍Finally, humor provides a strategic entry point into complex questions about AI's role in cultural production. By approaching these issues through play rather than solemn prophecy, A.I.C.C.A. creates space for more nuanced conversations about automation, algorithmic judgment and the changing nature of expertise. The laughter it evokes isn't separate from critical engagement but can be a pathway toward it.‍

‍In a world where discussions about AI often oscillate between utopian hype and dystopian fear, I find humor to be a valuable middle ground—a way to acknowledge both the possibilities and the absurdities of our increasingly algorithmic culture.

Peter Bauman: Finally, Mario—and sorry, A.I.C.C.A.—is this a one-off experiment or do you see it as the first in a series of evolving iterations? How do you balance updating A.I.C.C.A. with its singularity as an artwork?

Mario Klingemann: A.I.C.C.A. is a statement I wanted to make, and as such, it stands as a singular work. I'm not particularly into repeating myself, so no, you won't see an A.I.C.C.A. II coming down the pipeline. My creative drive has always pushed me toward exploring new territories rather than developing sequels.‍

‍What makes A.I.C.C.A. different from traditional artworks is its built-in mutability. Unlike sculptures or installations that stay fixed once they leave the studio, A.I.C.C.A. is designed to evolve. With how quickly language models are developing, I deliberately built in the capacity for updates. In fact, A.I.C.C.A. has already received two significant software upgrades since I first created it in 2023.
‍‍

I think of A.I.C.C.A. as what I call a "performative sculpture.” It's not just an object sitting in a gallery; it has to perform to fulfill its purpose. 

‍
‍These performances adapt to different contexts and audiences. Our recent work with the Le Random collection shows this flexibility. We've shifted from live gallery performances to film presentations that let more people experience A.I.C.C.A.'s critical analysis.
‍‍

This tension between being one unique thing and continuously evolving is at the heart of what A.I.C.C.A. is about. 

‍
‍It exists as a singular artwork but one that transforms rather than stays static. In a way, this reflects AI itself: maintaining an identity while constantly incorporating new capabilities.‍

‍Finally, I’d like to say that, in case it's not immediately obvious, I wrote many of my responses in this interview with AI assistance. While all the ideas, experiences and perspectives are mine, I collaborated with an AI system to help articulate them.

There's something fitting about discussing A.I.C.C.A. through text that was partially crafted by AI. It creates a nice recursive loop. When talking about an AI art critic through AI-assisted responses, the medium becomes part of the message.‍

‍This isn't that different from how I work in general. I've always been interested in the creative possibilities that emerge when human intention meets algorithmic processing. Rather than seeing AI as a replacement for human creativity, I see it as an extension—a different kind of tool that allows for new forms of expression. ‍

‍In a way, this interview itself demonstrates that the questions A.I.C.C.A. raises aren't abstract but immediate and practical. We're already living with these blended forms of creativity—not in some sci-fi future but right now, in our everyday exchanges. The boundary between what we create and what our algorithmic collaborators help us create is already blurring. ‍And honestly, I wouldn't have it any other way.
‍

‍
---
‍


A.I.C.C.A. (Artificially Intelligent Critical Canine) is a robotic art critic and dog capable of exploring spaces such as galleries, museums or art fairs, then producing short texts on the works it encounters.

Mario Klingemann is an artist and skeptic exploring neural networks, code and algorithms, driven by a curiosity about human perception and the mechanics of systems.

‍‍Peter Bauman (Monk Antony) is Le Random's editor-in-chief.