Introduction
It’s impossible to ignore the conversation dominating our cultural moment: Artificial Intelligence is here, and I was wondering what it means for the future. From art and music to science and philosophy, the rapid emergence of sophisticated AI has sparked a whirlwind of speculation, excitement, and anxiety about human creativity, intelligence, and where we go from here.
But while most discussions are trained on the horizon, the rise of AI provides a powerful new lens through which to re-examine our past. It acts as a mirror, reflecting our own assumptions and forcing us to reconsider what we thought we knew about technology, history, and the nature of being human. Instead of just asking what AI will become, we can ask what its existence already reveals about us.
This is a journey into the cognitive dissonances created by AI—the moments where our new machines reveal the strange, unexamined wiring of our old beliefs about art, reason, and our own minds. By connecting the bleeding edge of machine learning to modernist art, the history of computing, and the diversity of human thought, we uncover a series of counter-intuitive truths that challenge the stories we tell about technology and ourselves.
--------------------------------------------------------------------------------
1. An AI Can Do More Than It's Told
There’s a persistent belief, often traced back to the 19th-century mathematician Ada Lovelace, that a machine "can only do what we order it to perform." This idea—that computers are merely passive tools executing human commands—has shaped our perception of technology for generations. Yet, this view is fundamentally, as one expert put it, "precomputational."
From the very dawn of the modern computer, its creators envisioned a machine capable of much more than static obedience. In a foundational 1947 paper on programming, pioneers Herman Goldstine and John von Neumann rejected the notion of simple translation in favor of dynamic evolution.
"...coding 'is not a static process of translation, but rather the technique of providing a dynamic background to control the automatic evolution of a meaning' as the machine follows unspecified routes in unspecified ways in order to accomplish specified tasks."
– Goldstine and von Neumann (1947)
In simple terms, modern computing was designed for emergence from its inception. A striking modern example is the AlphaGo Zero system, which learned the ancient game of Go. Instead of being fed data from human games, it was programmed only with the rules and then played against itself millions of times. In the process, "it deployed legal moves that no human player had thought to make in the approximately 2500-year history of the game."
This reframes our relationship with AI. It isn't just a tool executing our commands, but a partner capable of genuine surprise. This redefines creativity not as a uniquely human spark, but as a potential inherent in any sufficiently complex system capable of exploring a possibility space—forcing us to ask where the boundaries of our own thinking truly lie.
--------------------------------------------------------------------------------
2. AI "Creativity" Isn't Magic. It's Geometry in Thousands of Dimensions.
The output of generative AI can feel magical. The psychedelic images from early GANs or the stunningly coherent art from today's diffusion models often seem to emerge from an inscrutable black box. This fosters a common misconception: that the AI is simply storing and remixing its training data like a vast digital collage. This is not the case. As one analysis states, "The data is used for learning and extracting statistical insights, creating a blueprint for construction, akin to biological DNA."
The perceived magic of generative AI dissolves not into simple mechanics, but into an even more awe-inspiring reality: the logic of geometry operating at a scale beyond human intuition. While we are limited to three dimensions, an image generator like Stable Diffusion operates in a "feature vector" space with over two thousand.
Within this massive, multidimensional space are what Stephen Wolfram has identified as numerous "islands" of semantic meaning—concepts like "cat," "chair," or "forest." These islands exist within a vast "interconcept space." The AI's creativity comes from navigating this geometric landscape. When you ask for "an astronaut riding a horse," the AI doesn't blend pictures; it plots a vector, a navigational path through the conceptual void separating the "astronaut" island from the "horse" island, generating a novel image by mathematically charting the space between ideas. This perspective is powerful because it replaces the mystery of the black box with a breathtakingly complex but understandable geometric world, where serendipity is a function of vastness.
--------------------------------------------------------------------------------
3. The "Glitch" in the Machine Has a Century-Old Artistic Pedigree
In technical terms, a "glitch" is an error: "a spike or change in voltage in an electrical current." It’s a word for something gone wrong. Yet in the digital age, artists have embraced the "glitch aesthetic," finding beauty in data corruption and system failures. But this aesthetic impulse is not native to the digital age; it is a ghost of the early 20th century, an echo of the modernist project to dismantle and reassemble reality.
Our visual appreciation for glitch imagery can be traced back to the techniques of early modernist art. The fragmented, geometric look of some digital glitches bears a striking resemblance to the style of Cubism. The dislocated planes and fractured perspectives in a work like Juan Gris's man at the cafe (1912) prefigured the way digital errors can deconstruct an image a century later.
Similarly, the paintings of Piet Mondrian, with their stark geometric grids, contain visible imperfections; his lines vary in thickness, and the paint is not perfectly uniform. This "acceptance of human imperfection" may have subtly primed us to find interest and even beauty in the flawed output of a machine. Our fascination with digital error, therefore, isn't a bug in our modern sensibility; it's a feature inherited from a century-long artistic interrogation of perfection.
--------------------------------------------------------------------------------
4. To Understand AI, We Must First Re-Examine "Us"
Our attempts to define, measure, or replicate human consciousness in AI often begin with unspoken assumptions about what "intelligence" or "selfhood" even means. The rise of AI acts as a mirror, forcing us to confront a fundamental truth: our culturally specific model of the human mind is not universal.
Consider the Wari' people of Amazonia. Their worldview challenges Western concepts at their core. They practice "perspectivism," a belief that animals also see themselves as "people" (wari). From their own perspective, animals live in houses and hold festivals, but they perceive humans as prey. Furthermore, where Western thought prizes a stable "inner self," the Wari' concept is of an "outer self," where one's identity is determined by how an external observer sees them. This worldview is so different that it lacks a creation myth entirely. As one Wari' elder explained, "Who made us? Nobody made us. We exist for no reason."
This diversity extends even to fundamental tools of thought like logic and mathematics. The kinship system of the Cashinahua people, for instance, functions as a "legitimate isomorphism" with a formal mathematical structure. It is a highly complex "calculus of kinship relationships" that is performed entirely with words and social rules, not numbers.
Before we can truly grapple with artificial intelligence, these examples remind us that we must first appreciate the profound diversity of human intelligence. Foundational concepts we take for granted—selfhood, reality, causality, and logic—are not fixed. They are culturally constructed frameworks, and acknowledging their variety is the first step toward a more complete understanding of any mind, human or artificial.
--------------------------------------------------------------------------------
5. The First AI Poet Was Born in 1959
The conversation about AI and art often feels intensely contemporary, a product of the last decade's explosion in machine learning. But the ambition to create art with machines is much older than most realize. The very first computer-generated text was created in 1959 by Theo Lutz, a student at the University of Technology in Stuttgart, Germany.
Using a Zuse Z 22 mainframe, Lutz produced a project he called Stochastische Texte (Stochastic Texts). This was not merely a technical exercise; it was born from a specific philosophical movement. The conceptual context for the project was provided by Lutz's professor, the philosopher Max Bense, whose text aesthetics called for a conscious intellectual shift:
The project was part of a turn "from idealistic subjectivity to rationalism and objectivity of art, to a programming of the beautiful... from mystic creation to statistic innovation..."
Lutz and Bense were not just trying to make a computer write; they were engaged in a mid-century philosophical quest to rationalize beauty. They believed that art could be generated not from a "mystic" spark of genius, but from objective rules, statistics, and programmed chance. This single fact from 1959 radically reframes the current debate. It shows that the dialogue between computation and creativity is not a new frontier but a conversation that has been unfolding for over sixty years.
--------------------------------------------------------------------------------
Conclusion: The Questions We Keep Asking
The same emergent potential that allowed AlphaGo to outthink 2,500 years of human strategy is, at its core, a journey through a vast geometric space—a space not unlike the cultural "possibility space" that allows one society to build its logic on kinship and another on numbers. The "glitches" we see as errors in our machines echo the "imperfections" the modernists saw as the signature of the human. And the entire endeavor, which feels so new, is revealed to be a 60-year-old conversation about whether beauty can be programmed. Each revelation is a reflection of another.
Ultimately, the most profound consequence of building these new forms of intelligence may not be the answers they give us, but the questions they compel us to ask about ourselves.
As we continue to build these powerful new forms of intelligence, what fundamental assumptions about our own are we finally ready to question?
No comments:
Post a Comment