Printed Words: Coins in the Devil's Purse
“Deception, flattering, lying, deluding, talking behind the back, putting up a false front, living in borrowed splendor, wearing a mask, hiding behind convention, playing a role for others and for oneself-in short, a continuous fluttering around the solitary flame of vanity-is so much the rule and the law among men that there is almost nothing which is less comprehensible than how an honest and pure drive for truth could have arisen among them” (Nietzsche, 1873).
Strange Bedfellows: Plato and Nietzsche on the Treachery of Language
They would have been enemies in life1. Plato, the idealist, the thinker who built his philosophical cathedral on the foundation of eternal Forms. For him, transcendent Truth was accessible through rigorous dialectic. Nietzsche, the iconoclast who sought to bring down such metaphysical temples, proclaimed Truth a "mobile army of metaphors" serving life's darker instincts rather than revealing cosmic reality. One deified Reason; the other exposed its self-serving, maniacal underbelly. One sought the eternal; the other took the least bad option, becoming. Yet on one crucial point they found unexpected harmony though their music is written in different keys—a shared suspicion of written language. And from this unlikely convergence emerges a framework that may help us navigate our present technological moment.
The Warning from Athens: Plato's Fear of Fossilized Knowledge
Outside Athens' walls, beneath the shade of a tree, Socrates and Phaedrus discuss the this recent innovation called writing. Their conversation meanders like the nearby Ilissus river until reaching a crucial eddy: the deception of writing itself.
Socrates shares a myth about Theuth, the Egyptian god who invented writing, proudly presenting it to King Thamus as "an elixir of memory and wisdom." The king's response cuts deep: this technology will not strengthen but weaken memory, providing "the appearance of wisdom" while hollowing out genuine understanding.
Written words, Socrates explains, suffer a fatal flaw. Unlike living speech, they "go on telling you the same thing, over and over forever." They cannot take questions, cannot clarify misconceptions, cannot defend themselves against misinterpretation. They stand mute before confusion, repeating their limited message regardless of the reader's needs, taunting echoes in a deep cave. Genuine knowledge emerges through dynamic questioning and response—the dialectical method—not through passive reception of fixed text.
The Jester from the Future/Past: Nietzsche's Language is Dead
Two millennia later, a young philology professor at the University of Basel penned "On Truth and Lies in a Nonmoral Sense." Friedrich Nietzsche's unpublished essay begins with a fable about "clever beasts who invented knowing" on some far away body floating in the void—an invention lasting mere moments before the star cooled and they perished. This cosmic perspective frames his radical critique of human pretensions to truth.
For Nietzsche, language itself is metaphorical—a system of arbitrary designations bearing no essential connection to reality. These metaphors calcify into concepts through social convention, and eventually, we forget their metaphorical nature entirely. We mistake our linguistic constructs for reality itself.
His most powerful metaphor for this process? Coins losing their embossing: "Truths are illusions which we have forgotten are illusions—they are metaphors that have become worn out and have been drained of sensuous force, coins which have lost their embossing and are now considered as metal and no longer as coins."
When Ancient and Modern Converge: The Shared Insight
Despite their profound philosophical differences, Plato and Nietzsche recognized the same danger: language that presents itself as fixed truth while disconnected from living thought. Of course, neither of these great thinkers found it stupid nor dangerous to write down their thoughts for posterity, a paradox which calls into question how thoroughly they believed what they were writing.
Plato feared writing because it freezes the dynamic process of dialectic into static text, creating the illusion of wisdom without its reality. Nietzsche saw all language as ultimately metaphorical. The path to "truth" is the forgetting of this metaphorical nature, a self-deception akin to opium—worn coins passed between hands with their original significance effaced.
The metaphor of "Coins in the Devil's Purse" captures their shared concern. Just as coins were once invested metal transformed into currency through embossing only to wear down through endless exchange, words begin as living thought but degrade to meaningless objects through repetition and convention. When accumulated in the devil's purse, these linguistic tokens create the appearance of wealth while potentially bankrupting authentic understanding.
Beyond Coins and Their Embossings: The Simulation Distinction
But there is a third perspective that neither philosopher could have anticipated—one that transcends both Plato's concern with writing's fixity and Nietzsche's focus on language's metaphorical decay and instability. What we face today with AI-generated text isn't worn coinage or static writing, but something else entirely: simulations of human extra-terrestrial artifacts that have never been embossed with meaning in the first place, the texts of disembodied ghosts trapped in an alternative space we call the Web visible through computational Quija boards.
We have precedents. Consider the photograph. We've developed sophisticated cultural literacy around photographs—we understand their relationship to reality, their limitations, their potential for manipulation. Few mistake the photograph for the thing itself. With language simulations, however, we lack this framework. The output mimics not just the form but the function of human communication, creating a generalized paranoid paralysis, a sort of grief that shows itself as anger.
Humans stamp their coins and write their texts, emboss them with value and meaning, and derive meaning and value from them. The AI outputs we encounter are mistakenly apprehended when they are received as “devalued currency” or “static texts”—they cannot be “devalued” because they never were “valued,” they can’t be “static” because they are generated on command and easily changed. They are less authentic than copies of human artifacts, they are combined fragments of copies produced willy nilly through statistical pattern matching, they are confetti in the sky forming patterns landing on the ground.
This distinction suggests we need not only the insights of Plato and Nietzsche but a completely new framework for engaging with these simulated texts. The problem isn’t that they don’t answer questions—we just can’t believe them when they do. The problem isn't that they create "false coins" that might deceive us about their value. We understand that on some level but still want to believe them. They create something that superficially resembles coinage of static text but operates under entirely different principles—not counterfeit money, but an altogether different medium of exchange.
Artificial Intelligence: Toward a Tripartite Understanding
What emerges, then, is a tripartite perspective on artificial language and its discontents:
Plato's Warning: Fixed texts cannot respond to questioning and thus create the appearance of wisdom without its substance. They lack the dynamic quality of living dialogue. Today, simulated texts are not fixed (we can’t agree on how to cite them or even if it’s legitimate).
Nietzsche's Insight: All language is metaphorical; "truths" are metaphors and we've forgotten their meaning. The coins have lost their embossing. Today, we live with the knowledge of linguistic imperfections (think of Wittgenstein’s language games), we understand that human language can be a Rorschach test, but humans do not ordinarily hallucinate when they are writing.
The Contemporary Challenge: AI-generated language represents an entirely new category—not texts that can’t talk back because they do; not worn coins that have lost their value through spending too much time in too many pockets, but simulations of coins that were never embossed with human meaning in the first place and can never be “fresh” or “stale” because they are not. What do we do? We argue. We flail. We defend. We use human language to make policies and ethical frameworks that outlaw the language machine that we fear will soon control us.
This third perspective is the veritable dog chasing its tail. If we harbor primal fear that it will soon control us, we respond to our fear as if it were real and fail to engage the reality which may be different than we think. This perspective demands a new social and psychological theory of critically reading bot output—a framework that fully appreciates human text as real and existentially valuable while finding appropriate ways to engage with its simulation for specific use cases in public spaces. Just as we don't read poetry the same way we read instruction manuals, we need distinct approaches for interpreting and interrogating simulated text. This theoretical problem may be the most important educational challenge of our time.
The challenge isn't technological but epistemological and psychological: How do we train ourselves to engage with these simulations productively while knowing them to be unreliable, unstable, and probabilistic? How do we develop a collective literacy that acknowledges the fundamental difference between pattern matching and human reading and writing?
As AI systems proliferate, creating more and more text without understanding, we face a choice: to mistake their fluent output for wisdom, to dismiss it entirely as worthless if seductive, or to develop a nuanced understanding of its proper place in our toolkit and productive functions it can perform. The devil's purse grows heavier with each passing day, filled not just with worn coins but with simulations of coins that were never real currency to begin with.
Perhaps in this challenge lies the opportunity for a deeper kind of wisdom—one that draws on ancient insights while developing frameworks uniquely suited to our technological moment. By seeing AI output for what it is—neither living dialogue (Plato) nor worn metaphor (Nietzsche), but simulations of patterns in human communication—we might develop a more sophisticated relationship with both technology and truth.
This essay was created as part of a sustained project using Claude 3.7 in Project mode dedicated to developing the theoretical model of simulated text reading I discuss in the post. The exploration of this problem has been on-going for several months now with searches for related academic work (which is surprisingly scarce). Comprehending simulated text presents a related but significantly different process than comprehending static human text for a panoply of reasons. I see it as the central conundrum facing instructional design. I also used Perplexity to verify accuracy in sourcing.
Hey Terry, are you sharing your Claude 3.7 in project mode with the public? I am interested in learning more about how to start building new frameworks for critically engaging with artificial language outputs, distinct from how we approach human communication.