Why Shakespeare’s lines demand intention, not imitation—and why machines can only echo sound.

By Michael Cummins, Editor, September 3, 2025

The rehearsal room was cold enough that the young actor’s breath lingered in the air. He stood on the stage with a copy of Macbeth, its pages soft from use, and whispered the line under his breath before daring it aloud: Tomorrow, and tomorrow, and tomorrow. The words fell flat the first time. Too rehearsed. Too conscious. He shook his head, tried again, letting the syllables drag as if they themselves were weary from carrying time. Creeps in this petty pace from day to day… The repetition was not just fatalism; it was the sound of a man unraveling, his will eroded by grief and futility. The rhythm itself had to ache.

A machine could, of course, manage the cadence. A program could be tuned to repeat the word “tomorrow” with perfect solemnity, to stretch the vowels just so. Google’s WaveNet system can produce uncanny variations of stress, hesitation, even sighs—digital sighs—at precisely calculated intervals. DeepMind’s recent work on “expressive TTS” allows a line to be rendered in tones of grief, anger, joy, or boredom. There are demo reels online where Shakespeare is fed through these systems, and the result is surprisingly competent. But competency is not intention. What the young actor does—searching for futility in his own chest, summoning weariness from his own private reservoir—cannot be coded. Intent is not in the sound of the line; it’s in the act of dying a little as you speak it.

This is what Shakespeare demands, again and again: not just language, but will. His characters live on the knife-edge of consequence, their words pressed out by motive. Romeo, stumbling over Tybalt’s body, gasps, O, I am fortune’s fool! He has just killed his wife’s cousin, wrecked his future, and tasted blood he never meant to spill. It isn’t just regret—it’s horror, the shock of realizing you’ve become the villain in your own love story. No algorithm can know the sting of unintended consequence. An AI might shout the words, might even deliver them with trembling emphasis, but the cry comes from a boy watching his own destiny collapse. The line does not live without that recognition.

The experiment has been tried. In 2022, an AI-generated voice performed Romeo’s balcony scene at a conference in Vienna. Listeners were impressed—some even moved. But when the line O, I am fortune’s fool! rang out, the room chuckled. It wasn’t just that the intonation was slightly off; it was that the cry lacked stakes. It was Romeo without a pulse, Romeo without a body to bear the guilt. The line did not fall short technically—it fell short existentially.

Hamlet’s soliloquies are the most treacherous test. In Act II he marvels and recoils at the same time: What a piece of work is man… How noble in reason, how infinite in faculty. It sounds like admiration, but it isn’t pure. The words turn over themselves—what ought to inspire awe instead curdles into disgust. He sees hypocrisy in every supposed nobility, futility in every faculty. An actor must carry the irony in his voice, lacing admiration with loathing, as though the words taste bitter even as they sound grand. An AI might deliver a clean, almost clinical balance—“admiration” followed by “disgust”—like toggling sliders on a mixing board. But irony is not a switch. It’s a wound dressed in velvet.

When DeepMind released an expressive model that could generate “sarcasm,” the tech press hailed it as proof that machines could finally do subtlety. Yet what we heard was not a fractured human voice, but a pristine and empty performance. The algorithm delivered a raised-eyebrow cadence, the verbal equivalent of a painted-on smile—a gesture without the impulse to conceal. This is the core of the paradox: sarcasm and irony are built on a bedrock of paradox—they require a speaker to mean two things at once, to hold a contradictory feeling in their voice and body. A computer cannot hold a contradiction. It can only cycle between two different outputs. It cannot fracture its own will; it can only mask its lack of will with a calculated pose. It’s a perfect pantomime of motive, but it is not the thing itself.

John Barton, co-founder of the Royal Shakespeare Company, once said that “Shakespeare is inexhaustible because he leaves space for the actor’s choice. Every pause, every stress opens a door.” The line is telling: it is choice that keeps the plays alive, not just rhythm. Machines can render a pause, but they cannot choose it. They have no sense of opening a door.

Brook went further. In The Empty Space, he wrote: “A word, a movement, a gesture is empty until it is filled with the life of the actor who chooses it in the moment. That life cannot be faked.” Brook believed theatre was only alive because of its fragility—the possibility of collapse at any instant. An AI-generated Lear might roar flawlessly through every line, but the roar would lack the pulse of possible failure. For Brook, this pulse was theatre itself.

The question of intention extends far beyond Shakespeare. What of a writer like Samuel Beckett, whose characters mutter their way through a landscape of despair? Molloy, in his absurdist journey, seems driven by nothing but habit. Yet even his rambling, fragmented speech is an act of will. He confesses, he tries to make sense, he fails. The very act of muttering is a defiant choice against silence and nonexistence. The words tumble out of him not because of a calculation of probability, but because he is compelled by the fundamental, human need to bear witness to his own suffering. He wants to be heard, even if he doesn’t know why. The machine, by contrast, cannot be propelled by such need; it does not hunger or fear silence.

Borges provides another mirror. In “Pierre Menard, Author of the Quixote,” he imagines a modern writer who painstakingly rewrites Don Quixote word for word—identical to Cervantes, yet different in meaning because of intention. The same words in a different century become charged with irony. Borges understood that words are never just words; they are vessels for will, for history, for desire. An AI could reproduce Shakespeare endlessly, but reproduction is not creation. The ghost of intent makes the difference.

Shakespeare writes as if to test whether a human voice can hold the charge of intention. Lear’s roar against the storm is the most elemental: Blow, winds, and crack your cheeks! It is not just noise; it is betrayal breaking loose. A father disowned, a king humiliated, Lear rages not only at the storm but at the cosmos for his madness and grief. It is a voice already fractured, demanding nature itself collapse. A machine can roar, yes. It can pump bass through speakers, crack like thunder. But it cannot bleed. To speak Lear’s line without the tremor of betrayal is to strip it bare of meaning.

The theater knows this well. In 2019, the Royal Shakespeare Company tested an AI-generated “co-performer” in an experimental production. The system generated lines in response to actors’ improvisations, its voice projected from a disembodied orb above the stage. The critics were fascinated, but they noted the same flaw: the AI could surprise, but it could not intend. The actors on stage carried the burden of consequence; the machine was a clever ghost.

Harold Bloom once wrote that Shakespeare “invented the human as we know it.” What he meant was not that Shakespeare created humanity, but that he revealed in language the contradictions, desires, and paradoxes that shape us. Bloom’s point makes the AI test more daunting: if Shakespeare gave us the map of interiority, then any performance that lacks interiority—any performance without stakes—is not merely deficient, but disqualified.

And then there is Portia, standing in the court of The Merchant of Venice, her voice softening into moral persuasion: The quality of mercy is not strained… It droppeth as the gentle rain from heaven. Here intent is everything. Portia is not just lawyering; she is pleading with the very idea of justice, urging her audience to see mercy as divine, inexhaustible. Her belief must be palpable. A machine could roll the syllables like pearls, but eloquence without conviction is nothing but polish. What gives the line its power is the speaker’s faith that mercy belongs to the order of heaven. Without that belief, it’s rhetoric without heart.

Here the cultural anecdote is darker: in 2021, an AI-generated voice was used in a court training exercise to deliver witness testimony. The experiment was intended to test jurors’ susceptibility to persuasion by machine voices. The results were mixed: some jurors reported being swayed, others reported discomfort. What unsettled them was not the quality of the performance but the absence of belief behind it. To be persuaded by words without will felt like manipulation, not argument. One legal scholar described the prospect as “trial by ventriloquism”—justice bent not by human persuasion, but by hollow eloquence.

The ghost in the syntax grows clearest here. Machines can offer us form—eloquence, cadence, even dramatic surprise. What they cannot provide is risk. An actor saying The quality of mercy risks hypocrisy if he fails to embody belief. The line costs him something. A machine, by contrast, cannot fail. Every performance is safe, repeatable, consequence-free. And it is precisely consequence that makes Shakespeare’s words ache.

The paradox is that we, as listeners, are complicit. We project intention onto anything that speaks. We hear a chatbot offer sympathy, and we feel soothed. We hear an AI-generated sonnet, and we marvel at its poignancy. We want to find meaning. We bring the ghost with us. The ELIZA effect—named for one of the earliest chatbots—was discovered in the 1960s: people poured out their souls to a crude program that only echoed their words back. If we can believe that, we can certainly believe in an AI Lear. But the belief is ours, not the machine’s.

Could AI ever cross the threshold? Some technologists argue that with enough layers, enough feedback loops, emergent properties might arise that resemble motive. Perhaps one day a synthetic voice will “choose” to pause differently, to inflect a line with bitterness not because a human trained it so, but because its internal processes made that choice inevitable. If so, would that be intent—or the perfect illusion of intent? The philosophers divide: John Searle insists that no simulation, however perfect, ever achieves the thing itself; Daniel Dennett argues that if behavior is indistinguishable from intent, the distinction may not matter. The stage, however, resists the reduction. A pause can be “indistinguishable” only if we do not ask what it costs the speaker to pause.

The Royal Shakespeare Company, now experimenting with immersive technologies, has been clear-eyed about the limits. Sarah Ellis, their director of digital development, called the company’s work with Intel’s motion capture in The Tempest “21st-century puppetry.” She explained: “The actor is always driving the performance. The technology amplifies, but it cannot replace.” The line could have been written as a manifesto for the AI age: amplification without intention is echo, not expression.

Back in the rehearsal room, the young actor stumbles. His voice cracks slightly on a word, a small imperfection that carries more meaning than a perfect rendition ever could. The director, sitting at the edge of the stage, leans forward, attentive. The line is not flawless, but it is alive. The risk of failure is what makes the moment vibrate.

A machine could reproduce the monologue flawlessly. It could echo a thousand performances until the averages smoothed every edge. But what it could never offer is that tremor. The possibility of failure. The risk that gives intention its bite. For intention is always wager, always consequence, always stake. Without it, words are only words, no matter how well they trip on the tongue.

And that is Shakespeare’s test. Could AI ever deliver his lines with intent? Not unless it learns to bleed, to risk, to believe. Until then, it will remain what it is: syntax without a ghost. We may listen, we may marvel, we may even project a soul into the sound. But when the storm clears, when Romeo cries out, when Portia pleads, it will not be the machine we hear. It will be ourselves, searching for meaning where none was meant.

THIS ESSAY WAS WRITTEN AND EDITED UTILIZING AI