Why we expand information to compress it later.
Many years ago I read a book called Hitchcock on Hitchcock, and one small passage in it has stayed with me longer than almost anything else I have read about films, which is one of my passions. Surprisingly, I have come to think that the ideas expressed there may not be only about films. I don’t have a copy of that book anymore–indeed it is a great book– but let me tell you how I remember it. The details may have drifted a little over the years, but that will not change the point.
This is how Hitchcock, in one of the chapters, explains what it is to make a film. A filmmaker wakes one morning with the idea for a new movie. He writes it down in approximately a single page, starting with something like: a retired detective with a fear of heights is hired by an old friend to investigate the man’s troubled wife. He falls in love with her and … He brings that page to a producer, who reads it and sees something great and original in those few lines and says yes, we will make this. Then the machinery begins to turn. A writer is hired to grow the one page into three hundred; the three hundred pages pass to the people who cast the actors, design the sets and the lights, dress the actors and decide on the placement of the cameras. The dialogs are finalized, the actors give the words a face and a voice; the composer lays music beneath the silences, and so on. A year or two, and a lot of money, goes by, until at the end there is a finished film, shown in a thousand theaters round the country and possibly the world. And now, Hitchcock says, picture an ordinary man coming home after a long day, and his wife asks what he did, and he tells her he had two free hours between meetings and slipped into a cinema. She asks what he saw, and he says he saw a good film, about a retired detective with a fear of heights, hired by an old friend to follow the man’s troubled wife. He falls in love with her and … If that story told by a stranger who happened to see the movie in his free hours, is the same story written on the page the filmmaker began with, then he or she had done his job. Of course that does not guarantee box-office success of the movie.
What Hitchcock’s test really proves
What that test does and does not tell you? It does not tell you the film was beautiful, or that it moved anyone, or that it will be remembered. It tells you something narrower, and in its way harder to achieve: that the idea survived. That across a year of work and hundreds of hands, nothing essential was lost between the first page and the last frame, and the thing that arrived in a stranger’s head was the thing that began in the filmmaker’s mind. That’s a mind-to-mind communication, going through a complex expansion of the information in the mind of the filmmaker to be compressed again in the audience’s mind. However, a film can pass Hitchcock’s test and still be dead on the screen, and a film can fail it in a dozen small ways while doing everything else well. The matching summary is not proof that the film succeeded. It is proof that the filmaker did his part of it, which is to carry one idea, intact, through an enormous and noisy apparatus without letting it dissolve along the way.
Continuous refinement
That apparatus does not run by itself. The production of a film is not just an idea that is simply handed off to that apparatus and later collected. A good film is the product of a long and intense conversation between the person who had the idea, in this case the filmmaker, the director who wrote that idea on a page, and the big apparatus. The filmmaker is there in the morning when the light is wrong, and in the editing room when a scene that read beautifully on paper turns out to die on the screen, and at the piano when the music says something the story never intended, and again and again he sends everyone back, saying not that, closer, no, once more, until the thing in front of her matches the thing she first saw in his head. The one page was the beginning of her work, never the end of it. And of course, during the process, some things may also change on the initial idea, the original page. Much of the film, in truth, is made in the hundred corrections between the idea and the thing that finally carries it.
Why the page is expanded
But why bother? Why expand the page at all? If the idea fits on a single sheet, why not simply give the audience the sheet?
Believe it or not, part of the answer comes from Claude Shannon, the Bell Laboratories scientist who first formulated the theory of information in the late 1940s. He realized that to push a message through a noisy channel intact you have to add redundancy; you have to say more than the bare minimum, so that what matters survives the trip. And humans, and in particular the audience of a film, are a noisy channel: a room full of people each carrying a hundred distractions in with them. A single page read aloud to that room would be misheard, half-heard, gone by the time they reached the street. The three hundred pages, the ninety minutes of film, the slow machinery of a movie production, are there so that the idea expressed on the one page arrives whole.
More than the story
And there is a second answer, and it is the deeper one. The machinery does not only carry the idea. It makes something the idea by itself never could, which is feeling. You can tell someone the plot of Vertigo–by the way, one of the best movies by Hitchckok–in a single sentence, the way the man told his wife, and they will know what happened, but they may feel nothing at all. You cannot summarize dread. You cannot compress the slow turn of the stomach as the camera seems to fall away, or the ache of a face seen again that should not exist. Those are not information to be transmitted, they are experiences to be built, and they are built out of precisely the redundancy that is not present in the summary: the held shot, the repeated motif, the scene that rhymes with an earlier one, the time the film is willing to take to make you wait and surprise you. The one page tells you what the story is. The three hundred help make you feel it, and the feeling is the part no summary can ever carry home. It is why the man’s wife, holding the plot in four neat sentences, has not really seen the film. She has the idea. She was not in the room when it became an experience.
And of course this is true, to different degrees, in all the artifacts created by humans. Artifacts, like a poem, a book, a painting, carry the original information that was in the mind of the artist, plus the expression of feelings. And that feeling is created by the redundancy added to the initial idea. And in some cases, by the subtraction of redundant elements in the original idea. As we move towards more abstract art, the expression of feeling becomes more and more important than the initial message. In documentaries, technical and nonfiction writing, news articles, the balance is more towards the facts, but the emotional factor does not disappear. Both elements are still there, maybe in different proportions.
The thing no theory can measure
So there is a shape to all of this. A film, a poem, a writing, carry an idea faithfully from one mind to many, and it turns that idea into something felt, and these are two different achievements, and a great artifact needs both. But beneath the two of them lies a third thing, and it is the one that no craft and no theory can supply.
Shannon’s information theory, for all its power, could measure almost everything about a message: how much it carried, how surprising it was, how faithfully it would survive the noise. But it could never measure whether the message was worth sending. And nothing else can, because there is no formula for the quality of a message. The most surprising signal you can transmit, in terms of Shannon’s theory, is the one with the highest entropy, or confusion: pure random noise, unpredictable in every part and meaning absolutely nothing. Novelty is not quality. A page can be perfectly original and perfectly empty. All the machinery in the world, running flawlessly and faithful to the last frame, will still only deliver whatever was on the page to begin with, and whether that was worth delivering is a question the machinery cannot reach. Two things slip past every instrument in the same way: the worth of the idea a work begins from, and the amount of its maker that ends up inside it. I call both of them quality, because they are near enough to be one thing.
There is an idea in Robert Pirsig’s Zen and the Art of Motorcycle Maintenance, a wonderful book that I read many times since it came out in 1974. An artifact, and even a machine or a mechanical part, assembled by someone who cared, who was fully there for each small decision, is not the same machine as one bolted together by a mind that was elsewhere, even when every part is identical, because the caring shaped a hundred choices too small to name.
And let’s come to the point I wanted to make from the beginning: writing is no different. A page can be produced with almost none of its author in it, and that absence is exactly what we feel when fluent prose reads as correct and hollow, technically right and somehow unalive, with nobody home. The long back and forth, the insistence on not that, closer, no, again, is how a person gets into the thing, line by line, until enough of them is present that a stranger on the other side can feel it was made by someone at all.
The machinery, in an afternoon
What I have been thinking, throughout this whole excursus inspired by Hitchicock, is how writing is now made with AI. Look again at the shape of it. Someone starts with a page, a thought that would fit in a paragraph or two, and hands it over, and back comes many pages, ordered and fluent and full. It is Hitchcock’s example exactly, the single idea given to the apparatus that expands it into something a reader can sit inside, except that the writer and the crew and the composer and the year of production have all collapsed into an afternoon at a keyboard. An AI model used for writing is the whole machinery at once, a kind of ghostwriter, editor, copyeditor, quick and tireless and impossibly well read, taking a page and handing back an expanded version of it. But there are many ways to use the power of AI today. You can have something written by AI, or you can write something with AI. And these are two completely different things that produce two completely different results.
Some hand over a thin idea and keep whatever the machine returns, mistaking the expansion for the work and publishing the result as though volume were the achievement. The others do what the movie director does. They begin with an idea worth expanding, and then they refine, in the same patient back and forth, saying not that, closer, no, once more, throwing out what sounds good but is not true, restoring the point the AI machine quietly rounded off, and steering, draft after draft, until the words on the page carry the idea that was in their head. The AI will add the redundancy that will carry the message across a noisy channel. It will even reach for feelings, and now and then it will land, but the feeling AI produces tends to read as faintly empty. Knowing the difference is the writer’s work and no one else’s.
Which leaves the last thing, and here I will state the absolute plainly, because the whole essay has been walking toward it. However fluent these machines become, and they will become more fluent than any of us now imagine, they cannot give you the one thing that was always underneath: an idea worth expanding in the first place. What Pirsig hints at. They cannot give you the judgment to keep the expansion true to it across a hundred corrections, and they cannot give you the feeling that turns a faithful summary into something a stranger carries home. Thousands of people can now grow three hundred competent pages from a thin idea in a single afternoon. That was never the scarce thing. The scarce thing is the page worth growing, and the patience to grow it well, and the taste to know when it has finally become true, and all three of those were, and remain, entirely human, entirely yours. The machine will build the film. Whether it was worth making is the one question it cannot answer, and the one you cannot hand to anyone else.