News & Articles
Browse all content by date.
An AI-generated sunrise with lilacs in the style or Renoir.
Why does AI-generated imagery leave humans cold?
Before I begin I need to point out that I use two illustrations here that were AI-generated. This means they were thrown together in nanoseconds from work “scraped” — i.e. stolen — from original images posted by mainly human creators on the internet, with no regard for copyright. I use these for educational purposes only.
Anybody can use an image generator. You download one of a gazillion apps, many of which don’t charge you anything, though they steal information and probably content from you. You type in a prompt. Or, if typing is too onerous a task for you, you speak a prompt. In an incredibly short moment, you get a picture.
I entered these instructions: “Make a sunrise with lilacs in the style of the French painter Pierre-Auguste Renoir” into Photoshop’s Generative AI tool. The picture at right is what I got, along with a selection of other images that were just as bad.
This was not what I was looking for.
An AI-generated image produced by more specific instructions. I don’t own the copyright to this, either. Again, no one does.
Then I asked it to handle a subject that Renoir actually tackled. I made my instructions more specific: “Make a picture of a hand-painted Japanese vase with a large bouquet of white chrysanthemums and other flowers in soft pastels, including green. Put it against a grey background in a shaded interior setting. Please make it in the style of Pierre-Auguste Renoir.”
In this case, we can compare AI’s work with Renoir’s — or an internet reproduction of a Renoir, copied from Wikipedia Commons and in the public domain. Here it is:
“Spring Bouquet” by Pierre-Auguste Renoir, 1966, Fog Museum, Cambridge. Public domain, courtesy of Harvard University, available on Wikipedia Commons.
This is one of several attempts I’ve made to get anything that looks like art out of that tool. I never attempted to publish these images. I’m driven mainly by my belief that you shouldn’t knock something new if you haven’t tried it. (I’m also driven by intellectual curiosity, which is supposedly a gift, but one that gets me into lots of trouble.)
Why Renoir? I was hoping this tool is smart enough to look up examples of the artist’s work and move away from its usual style — which is generally eyeball-exploding balls of color, suitable for a five-year-old’s birthday card.
In other words, I was asking, to show me a frozen moment of light and soft gradations of color, the way things actually appear to a human being.
No dice.
I can predict a lot of responses to our psychedelic sunrise picture, if I put it in front of a general audience of readers, instead of you intelligent and visually sensitive people who are still reading after three minutes:
“Ooh, I love it!”
“I want to hang that over my sofa!”
“So cheery!”
And, of course, one follower, a good photographer, pointed out that I had used the wrong tool. He provided an example of a result from another app. It was better in the sense that it looked more like Renoir, but it was a picture of a young woman.
The results I got look nothing like Renoir’s art. But that’s not the problem. We all know that we collectively have lost our minds. But are we also losing any ability to discern or comprehend why these examples are not art?
Besides the fact that AI imagery is either silly and childish, or ugly and weird, like the nightmares of a gamer who’s had one too many Doritos, there are certain criteria for defining what art is, conditions that make it impossible for AI as it exists today to produce art that rivals the art we have from the past and the present.
Many critics have tackled this, but one clear, if lengthy description was provided by John Dewey, a brilliant, multi-disciplinary scholar whose work was very well-known in the early years of the 20th century. In 1932, in the first William James Lecture at Harvard, he presented ideas about art that would form the basis of Art as Experience, a book published two years later.
To make it as simple as possible, Dewey argued in the book that art-making is the transmutation of an experience into a particular medium, and art consumption is the experience of sharing the joy of creation. This, I think, may be the reason that people who read a lot of poetry may start to think of themselves as poets, people who listen to a lot of music start to feel like musicians, and people who look at lots of art may start to think of themselves as artists — art changes human perception, making us find interest and beauty in unexpected corners of mundane existence.
When we look at a painting, or hear a piece of music, or read a poem or a novel, we enter into the perceptual realm of an artist. Software does not have a physical body in the universe. It cannot experience anything.
What is experience? Dewey makes a clear distinction between experience in general and an experience.
Experience in general, for conscious beings, “occurs continuously, because the interaction of live creatures and environing conditions is involved in the very process of living. Under conditions of resistance and conflict, aspects and elements of the self and the world that are implicated in this interaction qualify experience with emotions and ideas so that conscious intent emerges. Such experience represents human awareness at its peak, not in inchoate states of boredom or ennui.”
But an experience, he writes, an experience worth transmuting into art, is different from the experience in general. In our time, we might say we had “an experience” when things happen that we recall as being “like a movie” — anything from falling in love to crashing your car and falling off a cliff. That kind of thing.
Dewey writes: “[W]e have an experience when the material experienced runs its course to fulfillment. Then and then only is it integrated within and demarcated in the general stream of experience from other experiences. A piece of work is finished in a way that is satisfactory; a problem receives its solution; a game is played through; a situation, whether that of eating a meal, playing a game of chess, carrying on a conversation, writing a book, or taking part in a political campaign, is so rounded out that its close is a consummation and not a cessation.”
Artificial intelligence is not conscious. It does not experience and it does not have experiences. It cannot experience being “finished” and cannot feel any satisfaction whatsoever from anything at all.
But another important distinction is this: The artist, in the process of creation, steps back, looks long and hard at a painting, plays the opening chords of a symphony, and reads recently written passages again and again. The artist perceives not only the experience that triggered the production of art but experiences the aesthetic experience of the perceiver or consumer or connoisseur of art in the process of shaping and perfecting the work.
That is what art is. A way of sensing, a way of perceiving, a way of being in the world, or in a world, that was experienced perceptually or imaginatively by a conscious being in such a manner that we can share that experience.
That is why looking at “Spring Bouquet” is an aesthetic experience, and looking at AI’s version is not.
A machine may produce images of a world that looks like ours in some way or in many ways. It may write full-length sci-fi novels or movie scripts. But the product leaves us cold, or it should, because we have been cheated, and those of us who have a scrap of humanity left know that.
If you would like to read a 1980 edition of Dewey’s book, you can find it at Archive.org: https://archive.org/details/deweyjohnartasanexperience
Duluthian Laura Marland is a photographer and writer on a variety of topics.
Tweet |