Paul Schrader says artificial intelligence gives better script notes than studio executives. He claims it generates better ideas than most screenwriters. The director of Affliction and Hardcore tested ChatGPT by requesting movie premises in the style of various filmmakers. The machine impressed him with rapid-fire pitches that felt both original and appropriate. This revelation struck him like Gary Kasparov realizing Deep Blue would defeat him at chess.
Schrader's wonder at artificial intelligence tells us more about Hollywood's creative drought than any technological breakthrough. The system that broke Orson Welles and kept Kenneth Anger out of sight (and cite) specializes in the production of insipid content using templates so rigid a computer can master them. This is nothing new: RKO, back in the day, treated screenplays like contraband. Every draft was retyped, approved, then burned. No loose pages, no fingerprints.
Jean Renoir, that French visual poet of human frailty, ignored the rules. His 1947 The Woman on the Beach survives as a 300-page script riddled with handwritten notes. A mess, but a human mess. Today’s AI doesn’t spill ink or coffee on the pages. It doesn’t rage against producers or scribble half-drunk revisions at three a.m. It just... generates. Clean, efficient, correct. Schrader’s right: Feed it “Paul Schrader film idea,” and it’ll spit back something moody, existential, drenched in guilt. Same for Tarantino, Lynch, Bergman. The machine knows the beats, the archetypes, the shadows we mistake for depth.
The streaming era already accelerated the transformation of screenwriting into paint-by-numbers. Writers analyze successful shows to extract formulas they can replicate in mass quantities. Schrader admits this mechanical approach works for procedural television. "If someone asked me to write a CSI episode," he says, "I'd watch a dozen to catch the template—the characters, dialogue, plot positions. I could knock that off easily enough." He acknowledges AI will likely do this better because "it's faster, cheaper and does not waste its time with any pretension."
This admission hammers home Hollywood's assembly-line mentality. When human writers already function like Small Language Models (SLMs), actual LLMs feel like an honest-to-god upgrade. The machine processes more examples faster and never deviates from the formula. It won't attempt artistic flourishes that might confuse the audience. It won't demand creative control or residual payments.
But standardized storytelling creates its own kind of poverty. Consider how easily artificial intelligence could generate endless variations on Harry Potter—the beloved boy wizard discovering his powers, gathering quirky friends, and battling dark forces across seven adventures beloved by precocious tweens and adult mid-wits who are now big mad about the author’s transition to TERFdom. The premise itself contains the blueprint. Slap 30 pages into a prompt, feed it into Sonnet 3.5 or DeepSeek, and let them cook.
True innovation resists such reverse engineering. Stan Brakhage's experimental films emerge from personal vision, not market analysis. Kenneth Anger's Scorpio Rising has no precedent in terms of mixing motorcycle footage with occult imagery and pop songs. Dušan Makavejev's Sweet Movie connects ideological and sexual liberation through shocking juxtapositions no focus group would approve. These filmmakers expand cinema's possibilities rather than exploiting proven formulas.
Therein lies the rub: ChatGPT can’t write Sweet Movie because it’s unwritable. That surrealist fever dream—a swirl of politics, piss, and perversion—defies templates. So do Brakhage’s hand-painted celluloid hymns and Anger’s homoerotic psychodramas. These works aren’t puzzles to solve; they’re seizures to endure. AI thrives on patterns, but what’s avant-garde if not a rejection of patterns? The machine can mimic a gritty slice-of-life vignette like Scorsese’s Mean Streets, but it can’t drunkenly stumble into After Hours. It can’t panic, can’t fail in ways that matter. Not yet, at least.
Schrader's ChatGPT argument succeeds because it’s operating within narrow commercial constraints, within the world that is. The machine analyzes patterns in existing films to generate variations that feel familiar and fresh. But this is more efficient imitation than genuine creation. It can remix elements from thousands of movies faster than any human. It can’t, however, draw on experience or challenge conventional wisdom or pursue a personal artistic vision.
The problem isn't that artificial intelligence gives better notes than executives. The problem is that executives judge scripts by such rigid criteria that a pattern-matching LLM can replicate their feedback. The problem isn't that AI generates better ideas than writers. The problem is that Hollywood has reduced storytelling to such predictable formulas that computation can replace imagination.
The fight, such as it is, isn’t between man and machine. It’s between memory and oblivion. Paddy Chayefsky’s redlined pages, Dalton Trumbo’s dog-eared scripts—these are relics of a faith. Faith in friction, in the sweat of creation. AI offers revelation without rapture: Enter the prompt, receive the vision. But what’s a vision without a visionary? Without the doubt, the second-guessing, the nights when the words don’t come?
When filmmakers celebrate machines for accelerating generic content production, they reveal how far the industry has drifted from cinema's artistic potential. The solution isn't faster, cheaper templates. The solution is remembering what human creativity can achieve when it breaks free of formulas. Otherwise we're just training smart machines to guess what we guess we want to see: “Guess the guesses.” That isn’t art. It’s Family Feud.