What if GPT Does to Writing What the Camera Did to Painting & Drawing?
GPT:Prose :: Camera:Drawing ⊃ ??
Before 1838 or thereabouts, if you wanted to create a visual image—of a landscape, or a person, or whatever else you might have in your head to picture—it was a difficult process, the sort of difficulty that economists sum up with the term “expensive” although often that expense was in human labor, knowledge and connections, and other things that can be (but needn’t be) purchased with actual money. You needed to hire an actual artist who had to take a reasonable amount of time using laboriously acquired skills to hand-craft the picture, whether portrait or still-life, painting or drawing or etching. It was expensive, and therefore it was rare: only the rich and the well-connected (e.g. friends of artists) would have portraits. Most people passed in and out of life without any record of their appearance; most places would crumble into ruin without leaving a solid record of their glory.
And then this happened:
This did not, of course, replace painters and draughtsmen overnight. At first subjects had to sit still for minutes at a time to have their portrait taken, for example. Color photography had not yet been developed.1 But the process was underway: James Polk became the first sitting president to have his photograph taken in 1849.2 This doesn’t mean that photographs instantly replaced painting—the official presidential portraits remained painted through Obama—but all of a sudden there were new ways to make images.
Which meant that something that had been always bespoke, expensive, rare, was on its way to becoming common. It was a slow process: arguably it didn’t end until the widespread disemnination of the smartphone, less than fifteen years ago. But each step was imporant: family pictures became a more common thing, and then a routine thing. Almost everyone alive today (at least in the developed world) will have a photograph of their grandparents. The camera, in other words, made simple and automatic the process of taking basic images.
What this did was precipitate a crisis in art. What were artists for if not making a representation of reality? If you wanted to know what someone looked like you got a portrait painted. What else was there?
The answer, of course, was that art could be understood as something other than a representation of reality. Indeed, this understanding was (selectively but not entirely untruthfully) read back into the history of art, such that it was seen has having always been about something other than just representing reality. It was expressing a personal vision; making an aesethtic object. One got a portrait not to remember someone’s face but as a glimpse of the subject’s soul by way of the personal vision and aesthetic craft of an individual.3
As for the camera, its uses became multiple. One could use it for routine portraits—think driver’s licenses. Other people used them to create artistic visions, manipulating what was (in some sense) a direct and simple technology in ways which produced personal, artistic effects.
So whereas in the early nineteenth century, we had only one way to produce a portrait:
By the early twentieth century, we had three: artist’s vision; artist’s vision using a camera; and straightforward photograph:
The creation of an easy way to make visual reproductions did not make obsolete the artistic creation of reproductions, but it changed their purpose, their nature, their look, as well as expanding the technological means by which they were made.
Clever Reader, you are, I suspect, way ahead of me.
Perhaps ChatGPT, and its cousins and descendents, will do for the creation of text what the camera did for the creation of visual images.
Right now, its capabilities are crude: just as early photographs were not up to the standards of contemporary aritsts, what ChatGPT writes now is a standard, average text. It will improve, just as photographs did, and presumably in the future we may use such devices for all common recordings, whatever the textual equivalents of drivers license photos are. (Corporate memos and fundraising letters are already being done; more, surely, is to come.)
But I have some hope that artists in words will not be rendered obsolete by these machines. They might change their style—indeed calls for invisible prose may well be going, in some sense, in the wrong direciton, just as art over the modern age increasingly resemembled less and less a window on the world and more and more an unmistakable interpreation of it. But there will still be a call for sentient, creative, thoughtful artists to craft words.4 And their may be other types of artists—AI-whisperers, perhaps—who can bend the seemingly automated process of text creation to the ends of art in ways far different than those which we lesser mortals use it to create selfies.
One objection might be that now AIs are creating art in various styles; you can take a photograph and have it rendered in the styles of Van Gogh and Modigliani. But even apart from the question of how good those imitations and renderings are, they are not—or, at least, not yet—creating styles of their own. And, of course, in the modern period, creating a style became (in part for reasons related to the developments I have sketched here) a central part of what an artist does: if the work of art is not reproduction but a personal vision of something, then the creation of the form of that vision is going to be a central part of the work.
A final note: there is other ways photography has interacted with other visual arts aside from pushing it in different directions and being used as a tool to make art. For instance, visual artists use photographs as references, whether taking them themselves or (nowadays) using google image to call up a picture of some object or place or person they wish to draw or paint. And there are yet other ways that photographs can be made into separate artistic objects, ways which are not the making of artistic photographs per se, but in which more ordinary photographs become incorporated into something greater. Here, too, there may be parallels with this new type of camera, the camera that instead of creating an instant, often un-artistic, rough and ready image, creates an instant, often un-artistic, but rough and ready text.
I would try to guess them, but they could, I suspect, no more be imagined today than large art projects using photos could have been imagined in 1850.
There is no question that ChatGPT will change writing and writers. But there is, I think, hope that it will not render them obsolete.
Although there were other ways to get color photographs.
Ex-president John Quincy Adams beat him to be the first president ever, in 1843.
It goes without saying, I hope—hence I am saying it merely in a footnote—that this is a one-note, simplistic explanation for a complex historical process. Lots of other factors came in to play—romanticism comes to mind, but there are a great many more. Photography played a role: I am focusing on that to make a point, not to claim that that’s all there was to it.
Note I am just talking here about machines like the GPTs; if AI gets to the point of actaul creativity and sentience (a possibility about which I know no more, and quite possibly less, than you do, Well-Informed Reader), that will raise an entirely different set of issues to those I am addressing here.
A minor nit-pick: Obama's official white house portrait is an oil painting - both the one you linked to, and the extraordinary portrait by Kehinde Wiley that hangs in the National Portrait Gallery. I'm a little confused as to which image of Trump - there are both paintings and photos - is "the" official one. I don't think there's an official Biden painting yet, but I assume there will be; sometimes they're not painted until after the person is out of office. (Trump's might also be still to come).
In terms of artists incorporating new tech into their work, you mentioned people using photos as reference. But it's becoming increasingly common for people to generate AI images for reference, too.