Abstract This article presents Gates from the Body (1972), an installation by artist Juan Navarro Baldeweg, as a foundational artwork in the history of art and AI. It illuminates the context of the installation’s creation at MIT, where Navarro worked with the Center for Advanced Visual Studies (helmed by Gyorgy Kepes) and the Architecture Machine Group (helmed by Nicholas Negroponte) and encountered concepts drawn from cybernetics and AI research. Navarro took up the idea (sourced from Kepes) that art functioned as an analogic model, revealing structural equivalences across organic and technological worlds. He became fascinated (through Negroponte) by the perceptron, an image classifier, which was modeled on the optical neuron. In Gates from the Body, Navarro created an analogic model of a perceptron; incorporating sensors, wires, and illuminated boxes, the installation captured and processed images of its viewers, presenting those viewers with mechanically produced reflections of themselves. This installation probed the metaphorical equivalences between mind and machine underpinning AI research and revealed the viewer’s entanglement in cybernetic systems, disrupting humanistic conceptions of the subject and introducing cyborgian alternatives. Navarro’s project demonstrated that art and AI are assumed to operate in similar ways: as reflections of ourselves which make us aware of our subjectivity and incite us to probe its limits and potentials anew. This article reconsiders what constitutes “AI art,” suggesting that the term should encompass all art that is forged in the epistemological framework of AI (not just images produced using generative AI) and that it is especially useful for describing artworks that interrogate how AI technologies reflect and affect our understanding of the self.
Throughout history, certain art pieces have been labeled as rule-breaking or pioneering. Such artworks elicited strong reactions and rejection initially, but ultimately broke conventions and introduced new art kinds. Although revolutionary artworks play a key role in art history and society, established theories in philosophy of art have neglected this central topic. To address this gap in the literature, this article offers an analysis of pioneering artworks by drawing on tools from the ontology of artifacts while also introducing new concepts. I argue that the emergence of pioneering artworks is comparable to the emergence of prototypes in artifacts. These artworks enable the creation of new art kinds, much like prototypes provoke the emergence of new artifact kinds through the introduction of a new function. Furthermore, such artworks do not merely introduce a new function, as prototypes do, but often a function perceived as heretical, one that challenges artistic norms and provokes strong audience reactions. I conclude that a functional understanding of pioneering artworks that takes into consideration their disruptive force reveals their unique nature.
This is an exploratory essay about epistemological horror, which can be understood broadly as horror fiction that engages with epistemological questions, and more narrowly as horror fiction that engages with questions about the boundaries of human knowledge. It is suggested that there are at least two interesting categories of epistemological horror. One relates to the question of what should and should not be known, while the other relates to the question of what can and cannot be known. These categories can be broken down further into the categories of Promethean, Cartesian, and Lovecraftian epistemological horror. After explaining and investigating these categories of epistemological horror, I briefly consider the idea that horror fiction has epistemological value. Throughout the essay, I offer some thoughts about the general nature of horror fiction and the alleged paradox of horror fiction.
This article examines the status of artificial intelligence (AI) as an artistic medium. A definition is proposed according to which a medium is a manipulated grounding for a work's tangible qualities. AI meets this definition when used in making art, and is, more specifically, characterized as a mediated latent medium. Objections to the thesis that AI is a medium are also addressed. The article then asks what it would be for an AI artwork to exhibit medium specificity. Strong forms of medium essentialism are rejected, but a weaker alternative is proposed, according to which media have characteristic aesthetic potentialities. Focusing on recent generative models, five potentialities are then identified: statistical idealization, digital aberrations, recombination, hypermnesia, and AI automatisms (e.g., prompting). Accentuating these features allows artists to demonstrate that AI is a new, distinctive artistic medium.
What we call "musical instruments" have stable identities both as physical objects, qua the characteristics that define them (a "trumpet" say), and as musical objects, which is conveyed by the musical ends to which they are put (as in "jazz trumpet"); the musical object, that is, the musical instrument, remains the same even when the ends in question vary. An apparent exception to this rule, one routinely marked by luthiers, purveyors, and players alike, is the case of "fiddle" and "violin," where an identical physical object gives rise to two different instruments. This article asks how this transformation from physical object to musical instrument takes place and thus explains why a fiddle is not a violin, the difference being found in the "artifactual" or "culturally freighted" character of the instruments in question. The discussion proceeds under three main headings-"the music that is played," "how the instrument is played," and "performance and authenticity,"-and concludes with a brief observation on the implications of the argument for standard ways of ranking fiddle and violin along with the music and culture associated with each.
True crime is an incredibly popular genre of entertainment that crosses many forms of media including books, movies, TV shows, and podcasts. Is there something morally wrong with true crime stories or the enjoyment of them, given their focus on the pain and suffering of real people? In this article, I consider a range of concerns that implicate both the ethical and aesthetic dimensions of the true crime genre, and outline parameters for both the ethical creation and consumption of true crime.
There are many ways we describe visual artworks: beautiful, sublime, ugly, disgusting, and so on. What about AI art? AI art nowadays can be indistinguishable from human-made works. But this is not always the case; a lot of AI art is weird. This article makes a step towards an aesthetics of AI art. I propose that weirdness is central to this aesthetics. Weirdness, as I conceptualise it, is a form of norm violation. However, I argue that norm violation alone is insufficient to distinguish AI weirdness from other examples of norm violation in art (such as we might see in comedy or horror). Examining AI weirdness further suggests that it is not merely a violation of norms but the result of some failure to reproduce something that is convincingly human. The weirdness in AI art thus stems from norm violation through non-human failure. I address a potential objection to my account of AI weirdness: AI images are not weird; they are uncanny. I argue that while some weird AI images may be appropriately described as uncanny, uncanniness cannot sufficiently explain all cases of weirdness in AI images. Weirdness, unlike uncanniness, does not require uncertainty and can produce responses of amusement as well as anxiety. This, I argue, suggests that the weird and the uncanny are different, albeit overlapping, phenomena.
In this article, I introduce the concept of interpretational creativity (IC), a form of creativity in human artistic activity that has been neglected by discussions of AI creativity. IC, which I describe in relation to Persian traditional music and a few more art traditions, involves creating a new artwork that exhibits a novel relationship to a body of traditional content. Because IC is different from and orthogonal to other forms of creativity, specifically those discussed by Margaret A. Boden in relation to AI, I offer a new analysis of the capacity of generative AI systems to exhibit IC. Based on a technical understanding of how dominant generative AI models learn to produce artistic products, I argue that IC poses an acute challenge related to the need to master extensive cultural context and relationships that are not always descriptively articulable. Regarding the implications of my arguments for the broader problem of AI creativity, I argue that although that challenge can be considered as related to processing contextual information by AI systems, the problem, at a deeper level, concerns the lack of a capacity in AI to adopt a value-laden, cultural, and perspectival stance and to reflect on a tradition.
This article introduces and analyzes the notion of aesthetic bias in generative AI. I argue that AI models systematically privilege certain aesthetically relevant features while marginalizing others. Drawing on examples from popular text-to-image generators, I highlight recurring tendencies, such as a preference for beauty, spectacle, saturation, or symmetry. These aesthetic biases carry specific risks: they disguise aesthetic preferences as neutrality, threaten to homogenize artistic production and taste, and contribute to creating self-enclosed communities of appreciators, or "aesthetic bubbles." The upshot is that aesthetic bias in AI demands serious study.
