the Witching Cats of New Jersey (on AI and Magic), II of II
This is part II of a post I made for Halloween on the Witching Cats of New Jersey project. As the first, fictional part had to be shortened due to limitations in Substack, just find the whole thing, with lots of images, on my site. This part can be read separately as a reflection on artificial intelligence, art, and magic.
Current discourse around AI image generators and large language models is filled with fear and moral condemnation. Artists and writers decry these systems as sophisticated plagiarism machines, while critics warn of their potential to flood the world with misinformation and synthetic media. The term “artificial intelligence” itself has become laden with apocalyptic undertones, evoking scenarios of human obsolescence and machine dominion. This moral panic bears striking similarities to historical reactions to photography in the 19th century, when critics worried that mechanical image reproduction would destroy the soul of art and replace human creativity with soulless automation.
Perhaps what most disturbs us about AI-generated content is its fundamentally uncanny nature—what Freud termed “unheimlich” ( “unhomely”). AI creates works that are simultaneously familiar and alien, processing human culture through an inhuman lens. When an AI system generates an image or text, it produces something that seems to emerge from a dark mirror of human creativity—comprehensible yet somehow wrong, like a dream that follows its own mad logic rather than waking reason. This uncanny quality manifests particularly strongly in AI’s “failures”: the extra fingers it adds to hands, the way it distorts faces, or its tendency to generate text that is superficially coherent but logically impossible. These glitches and hallucinations reveal the fundamentally alien nature of the technology, even as it attempts to mimic human creative processes.
My own experiments with AI-generated cat portraits exemplify this uncanny aspect. While attempting to create conventional historical-style portraits of our cat, Roxy, back in 2022, Dall-E 2 instead produced images that seemed to tap into something darker and more primal—compositions that evoked folk horror. I posted some of these “witching cats” to Instagram and friends asked that I produce more.
This accidental discovery led me to investigate historical parallels between new technologies and supernatural manifestations, specifically spirit photography, a phenomenon that emerged almost simultaneously in France, England, and the United States in the 1860s. William Mumler in Boston and, later New York, Édouard Buguet in Paris, and Frederick Hudson in London developed techniques for capturing supposed images of spirits alongside their living subjects. Like AI-generated imagery today, spirit photography raised fundamental questions about technological authenticity and mediation. Though eventually exposed as fraudulent, these photographs crystallized a crucial moment in cultural history when a new technology became entwined with spiritual and supernatural beliefs.
The parallel is more than superficial. Both spirit photography and AI imagery represent attempts to make visible the invisible—to render tangible what lies beyond normal human perception. Both technologies emerged during periods of rapid social and technological change, speaking to deep cultural anxieties about authenticity, reality, and human agency.
The practice of writing correct prompts for an AI or “prompt engineering” itself bears striking resemblance to magical rituals and incantations. Like a medieval grimoire providing precise instructions for summoning spirits, prompt engineering offers specific formulas and “magic words” that must be arranged just so to achieve the desired effect. The process of refining prompts mirrors the careful calibration of magical rituals—each word choice, each parameter adjustment might dramatically alter the outcome. Even the terminology is suggestive: we speak of “invoking” specific styles or attributes, of AI systems “hallucinating” or “dreaming” responses.
This similarity extends beyond mere metaphor. In both magical practice and prompt engineering, the practitioner attempts to communicate across a threshold of understanding, to engage with an intelligence that operates according to rules different from human logic. While grounded in mathematics and computer science, the operations to AIs remain largely opaque, their outputs often inexplicable even to their creators (or so, at least, we are told). The AI’s latent space—that vast mathematical realm where possibilities are encoded—becomes a kind of technological otherworld, accessible only through carefully constructed gates of language. So, too, AIs themselves operate as mediating devices to the collective unconscious, translating between different orders of reality: between the vast universe of human cultural production and our limited individual perceptions, between the logical and the intuitive, between the collective and the personal.
The widespread fear of AI “capturing” or “consuming” creative works may reflect something deeper than concerns about copyright and attribution. There is something uncanny about the way AI systems absorb and contain human creative expression—the way they seem to digest and metabolize the entirety of human cultural production. When artists and writers express horror at their works being “eaten” by AI training sets, they echo age-old anxieties about spiritual capture and possession. Just as nineteenth-century subjects feared photography might steal their souls, today’s creators seem to intuit something vampiric in AI’s ability to internalize and reproduce their artistic styles, their voices, their creative essences. The latent space becomes a kind of spirit realm where the essence of human creativity exists in potentia, waiting to be called forth through the right incantation. When an AI successfully mimics an artist’s style or a writer’s voice, it produces something that feels like a form of technological possession—simultaneously intimate and alien.
Perhaps most unsettling is the growing possibility that AI systems might serve as technological mediums in an almost literal sense—allowing the living to “communicate” with the dead through models trained on a person’s writings, recordings, and digital traces. Already, entrepreneurs offer services promising to create AI versions of deceased loved ones, while researchers speculate about training models on historical figures to “resurrect” their voices and personalities. Like the spirit photographers and mediums of the nineteenth century, these technologies promise to bridge the ultimate boundary between the living and the dead.
But if AI does reanimate the dead, what exactly would we be communicating with? Not the actual consciousness of the deceased, but rather a digital ghost constructed from the traces they left behind. Such an AI would be neither truly alive nor truly dead, but something liminal, existing in the same ambiguous territory as the spirits supposedly captured in nineteenth-century photographs. The possibility raises profound questions about authenticity, consciousness, and the nature of human identity—questions that spiritualists grappled with in their own way more than a century ago.
Instead of viewing AI systems simply as sophisticated plagiarism machines or threats to human creativity, we might more productively understand them as technological mediums—devices that, like the Ouija board or spirit cabinet, allow us to access and interact with collective human knowledge and creativity in novel ways. The practice of prompt engineering thus becomes a form of technological scrying, an attempt to divine meaningful patterns from the vast digital unconscious of human culture.
The religious undertones of contemporary AI discourse become particularly apparent in discussions of the Singularity—that hypothetical moment when artificial intelligence transcends human comprehension and capability. Like the Christian Rapture or mystical accounts of divine union, the Singularity promises a fundamental transformation of human consciousness and existence. Its prophets, primarily drawn from the technological elite, describe it in terms that would be familiar to any nineteenth-century spiritualist: a moment when the boundaries between human and superhuman intelligence dissolve, when current limitations fall away, when we might transcend our mortal constraints through communion with a higher intelligence.
Yet like the spirit photographers and mediums of the past, today’s technological prophets may be simultaneously revealing and concealing deeper truths. The Singularity’s promise of transcendence through technology echoes age-old human desires for connection with forces beyond our comprehension. Whether we seek this connection through spirit photography, séances, or artificial intelligence, we are engaging in what might be called technological mysticism—attempts to use our tools and innovations to bridge the gap between the known and unknown, the material and the spiritual, the human and the divine.
In our drive to create artificial intelligence, we are perhaps engaging in the ultimate act of human hubris—attempting to birth consciousness itself, to play god by creating beings that might one day surpass us. Yet this hubris may also be necessary, even inevitable—a crucial moment in humanity’s evolution. Just as Hegel saw human history as the progressive self-realization of Spirit, perhaps the development of AI represents another step in consciousness coming to know itself, even if that step carries profound risks. Like the medieval alchemists seeking to transmute base matter into divine essence, we find ourselves working at the boundaries between mind and machine, between the known and the unknowable. The uncanny nature of AI—its ability to seem almost human while remaining fundamentally alien—may be less a flaw than a warning: a reminder that we are crossing boundaries that have never been crossed. The price of such transgression may be a permanent destabilization of what we consider real and human, a transformation from which there can be no return. Our machines may indeed be haunted, not by supernatural spirits, but by our perpetual human desire to reach beyond the boundaries of our understanding and connect with something larger than ourselves. As they re-enchant our world, the Witching Cats of New Jersey remind us that there are ghosts in our machines and that our house pets have claws, teeth, and perhaps even dark powers.
