GUILLERMO DEL TORO’S BFI SPEECH IS NOT ABOUT AI.
IT IS ABOUT HOLLYWOOD.
Most people heard Guillermo del Toro’s speech as a warning about artificial intelligence. The more I thought about it, the more I became convinced that AI was not the real subject of the speech at all.
Every film begins in a room where nobody has seen the film.
That is the quiet absurdity at the heart of cinema. Before the audience, before the reviews, before the festival premiere, before the box office, before the later certainty that makes every success look obvious, there is a room full of people trying to decide whether something that does not yet exist deserves to become real.
This is where Guillermo del Toro’s warning about “image illiteracy” becomes far more interesting than another argument about artificial intelligence. At the BFI Fellowship ceremony, Del Toro gave the internet the line it wanted when he called AI a form of “natural stupidity.” But the phrase that matters more is “image illiteracy” and “cinema illiteracy.” That is not merely a concern about machines. It is a concern about reading, judgment and responsibility. And the most difficult thing to read is not the finished frame, protected by completion, interpretation and hindsight. It is the fragile possibility still trapped inside a script, a sketch, a reference, a nervous pitch, a director’s strange conviction or a creature nobody in the room yet knows how to love.
A finished frame arrives with allies. It has critics, numbers, audiences, awards, essays, memories, arguments and hindsight. A successful film has even more protection, because once it has worked, everybody can explain why it worked. The industry is excellent at retrospective literacy. It can read courage after courage has become profitable. It can read originality after originality has become prestige. It can read strangeness after strangeness has become a brand. What it finds much harder is reading those same qualities before they have acquired shelter, in the moment when they still look risky, awkward or commercially inconvenient.
That is the real test inside the greenlight room. The future film has no defence yet. It arrives as a possibility surrounded by substitutes: comparable titles, audience profiles, budget ranges, platform strategy, franchise potential, market data, concept art, mood boards and cautious language designed to make the unseen sound less dangerous. None of this is foolish. Film is expensive. Evidence matters. Planning matters. A serious production culture cannot run on instinct alone. But evidence has a permanent limitation: it comes from what has already happened, while the work under discussion belongs to what has not happened yet.
This is the contradiction Hollywood keeps trying to manage and often tries to deny. The industry celebrates original vision after it has proof, while building more and more systems to avoid betting on it before it exists. It loves the strange film once the strangeness has survived the market. It loves the singular filmmaker once singularity has become a name. It loves risk once risk has stopped looking like risk. The difficulty lies in the earlier moment, when nothing has yet been validated and someone still has to decide whether the fragile, half visible thing in front of them deserves protection.
That is why franchises, audience analytics, recommendation systems, market modelling, familiar intellectual property and AI driven development systems belong in the same conversation, even though they are not the same thing. They all offer comfort before commitment. A known world makes the unknown smaller. A previous audience makes the next audience feel less imaginary. A comparable title gives anxiety a spreadsheet. A model gives hesitation the appearance of method. AI adds speed, abundance and simulation to a desire that was already there: the desire to make the future answerable to the past before anyone has to take responsibility for believing in it.
Del Toro’s career sits almost cruelly against that desire. Today his visual worlds are readable because they have history around them. The name protects them. The awards protect them. The audience protects them. But before that shelter existed, many of his defining ideas were deeply vulnerable. A dark fairy tale inside fascist violence. Monsters who are not obstacles but witnesses. Catholic shadows folded into creature design. Children who understand horror more clearly than adults. A love story whose central premise could sound absurd if described badly in the wrong meeting. Those ideas were not obvious before they existed. Someone had to read them before they were safe.
That kind of recognition is not sentimentality. It is work. It is taste, experience, imagination and responsibility under uncertainty. It is what a good producer does, what a serious executive does, what a filmmaker does when defending a fragile possibility that has not yet learned how to defend itself. It is also what begins to disappear when institutions become more comfortable with systems that absorb responsibility than with people who carry it.
This is why Del Toro’s remark about being “door holders” rather than gatekeepers matters. Holding the door is not passive, and it is not a cute democratic gesture. It requires judgment, because a door holder must recognise what deserves entry before the crowd has agreed, before the evidence has hardened, before the market has supplied its permission. If the door only opens for what has already been validated, it is not being held open. It is being operated by the past.
AI enters this story late, and that is where it belongs. The machine can generate images. That fact is no longer science fiction. It is becoming ordinary. But more images do not create more image literacy. They make image literacy more urgent. In a world where visual output becomes cheap, fast and endless, the decisive skill is no longer the ability to create another image. The decisive skill is the ability to know which ones deserve attention, protection, development, money, time and belief.
Del Toro’s speech should therefore not be reduced to fear of a machine. It is a warning about an industry that may become so fluent in prediction that it forgets the older, harder act on which cinema depends: reading a possibility before the evidence arrives. Cinema does not begin when an image appears on a screen. It begins earlier, when someone recognises something before it has become visible to everyone else and accepts responsibility for making it real.
If Hollywood loses that literacy, it will not matter how many images it can generate, extend, test, optimise or repeat. It will still have forgotten the one thing cinema has always required: how to believe in an image before the world can prove it right.