The Issues People Have with AI Precede AI
Toward the end of his life in the year 1880, Karl Marx’s writings were well known and the term Marxism was in use. Part of what was considered Marxism by the French Worker’s party was the inevitability and necessity of violent revolution. Revolution had become associated with Marx since the Communist Manifesto was written in 1848, which was a very revolutionary time in Europe. However, toward the end of his life, Marx had come to see how his ideas could be misused.
For example, in 1877 Marx spoke out against the misuse of his concept of historical materialism. By then the broad ideas of modes of production and historical materialism were being used to make predictions as if they were deterministic. Engels characterized this use of the ideas as essentially lazy, an excuse not to study history. Marx wrote a letter to a Russian Newspaper in which he outlines precise examples of how his ideas are not deterministic.
In 1880 Marx was working with the French Worker’s party to advance reforms. Together with Paul Lafargue and Jules Guesde, Marx and Engels wrote a program of economic reform ideas, most of which have come to pass today. However, the concept of what to do with these ideas was quite different between Marx and Lafargue. While Marx viewed the suggestions as a realistic and achievable means to advance society, having come to believe in the possibility of peaceful societal evolution over the past 30 years; Lafargue and Guesde represented the ideas of the communist manifesto. Guesde believed the reform suggestions would be rejected and that this rejection would foster hopelessness toward peaceful reform ideas, which would galvanize workers in the direction of violent revolution. Marx characterized the ideas of Lafargue and Guesde as “revolutionary phrase-mongering”. For the Marx of 1880, if Lafargue and Guesde were Marxists, “what is certain is that I am not a Marxist”.
Throughout history, there are copious examples of how ideas can take on a life of their own outside of their originator to the point of being in direct opposition to what the originator believes or advocates. One of my favorite parables of this phenomenon is the film Rope by Hitchcock. Inspired by the real-life news of Leopold and Loeb, the film tells the story of a murder carried out as a perverse reification of an unserious philosophy espoused by an admired professor. Jimmy Stewart in the role of the professor is aghast when he learns that the murderers were inspired by his words, “giving them a life he never dreamed of”.
This subject is represented in an essay by artist Sarah Anderson in which she describes the appropriation of her work first by people she finds ideologically disgusting, and then by indiscriminate internet image harvesting to program AI image generation. The focus of the essay is on the emotional impact of having one’s work appropriated in ways one finds odious. This is an undeniably emotional issue. We send our words or drawings out into the world with some feeling of self-investment, and as she puts it in describing a typeface made from her own handwriting, to have these expressions made into a program can feel “particularly violating”, “as if the person who had created it was trying to program a piece of my soul”.
How we emotionally react to the world directly follows from what we believe about the world. I’ve never expressed an idea and seen it represented back to me in the same way I sent it out. The only source for any specific idea or expression is going to be a specific individual. As Sarah writes in the essay, every human brings their own experiences and humanity into what they do. There is no exact copy when a human being copies another human being.
AI is seen to be different because it’s a machine. Many artists whose work has been used to train AI have been upset when they see the machine replicate flourishes they see as characteristic of their style. Sarah provides an example:
The above image is garbage, truly. But when Sarah looks at this she sees it as having captured, “signature elements of her drawing style” such as a character with black bangs or a striped shirt. She cites the striped shirt as a nod to Calvin and Hobbes, and the influence of Bill Waterson can be seen in her style. For most of human history, we have been seeing art in 3 dimensions, in galleries and museums, on walls. Only recently have we been seeing art in printed books. And very very recently, we have been able to see the whole world of art in quick succession on the internet, on museum websites, or curations of museum websites. The internet is at its best a modern-day Library of Alexandria.
When one takes in art history in large doses, a different effect happens on the mind than walking through a museum from painting to painting or flipping through the pages of an art book. One is overcome with the common elements there are to art. Our brains have been likened to pattern-seeking machines. When one visually consumes art to get a historical perspective, the differences are subsumed by the commonalities. The whole of art becomes a variation on a theme, that theme being the human experience. Within movements like cubism or mannerism, one sees the influence of a place in time and culture. When looking at the work of a teacher and their students the influence is clearly observable.
Technique is not what art is about. Anyone can learn a technique given enough time and practice. To work in animation it was a requirement in many cases to be able to exactly replicate someone else’s drawing style. In my own experiments in drawing when I was young, I disciplined my hand by copying images of Bart Simpson over and over. At the time, being able to copy a style exactly was a possible route to employment. Many newspaper comics were subcontracted to artists beyond their creator to maintain the daily schedule. This is still something that has a place in the market. There are people who reproduce oil paintings for a living. This kind of art is all technique.
Anyone can learn techniques, but not everyone has creative ideas. The reasons I’ve enjoyed Sarah Anderson’s comics have little to do with her artistic style, though it’s pleasing. The concepts inherent to the comics are the entire source of my enjoyment. I like her comics because they are funny. It’s the ideas, not the technique, that bring value. This is what we find valuable in all art. It doesn’t matter how aesthetically pleasing a film is if aesthetically pleasing is all it is. It’s the writing of the comic, not the typeface, that is enjoyed by people.
The metaphor of the camera has been invoked a lot in the discussion around AI art, but I don’t think it’s commonly appreciated what a change this made. Prior to the camera, whole careers were sustained by portrait work alone. The ability to replicate reality in paint was what made good art. Impressionism and Expressionism were resisted as art because of this aesthetic preference. In paintings made before the ubiquity of the camera, you can find a lot of paintings of empty rooms. When I first came upon these I was puzzled; the whole message of the painting seemed to be, “hey, check out this room”. The realization that the ability to replicate reality well used to be what made an artist “good” changed my view of those paintings. Room paintings were showcases of skill. As the camera became easier to use, what defined a good artist began to change. Movements like surrealism and cubism were centered on what the camera couldn’t do.
The same thing is happening today. AI art is effective at creating a landscape or a character study, relatively. What it’s not effective at is portraying complexity and meaning. This is because it’s not generated from experience. AI art generation works by associating a word with an image. When you start up a generation the AI only has noise, and within that noise, the machine finds what correlates with the words you’ve entered, as best as it is trained to. There is no conception of why a hand looks like a hand or a dog looks like a dog because there is no conception of purpose. There will be a visual approximation of clockwork, but it’s not capable of generating clockwork that looks functional, because in order to do that one would have to know how gears fit together and move, not just what they look like.
Nonetheless, it’s easy to understand why an artist would be troubled by seeing the aesthetic of their art replicated. At a fundamental level, this is an issue of personal expression altogether, the possibility that we can be misunderstood, decontextualized, and misrepresented. Often this happens entirely in people’s minds; When we communicate our thoughts a listener can omit a word or replace an idea. Only in interactions do these mental edits get brought to light, and we aren’t always granted interactions to clarify our ideas. This lack of clarifying interactions was one of Socrates’ principal objections to the written word;
Writing, Phaedrus, has this strange quality, and is very like painting; for the creatures of painting stand like living beings, but if one asks them a question, they preserve a solemn silence. And so it is with written words; you might think they spoke as if they had intelligence, but if you question them, wishing to know about their sayings, they always say only one and the same thing. And every word, when once it is written, is bandied about, alike among those who understand and those who have no interest in it, and it knows not to whom to speak or not to speak; when ill-treated or unjustly reviled it always needs its father to help it; for it has no power to protect or help itself.
There are many concerns around AI altogether and AI art specifically. The specific concerns of decontextualization and appropriation are not AI issues. These issues go all the way back to the written word.
There is a lot that AI cannot do. What AI cannot do will become the province of art going forward. This has happened many times before. The ability to learn from and imitate human output is not the same as human beings learning from experience. The unique viewpoint and generative ability of the creative individual cannot be replicated.
I’ve personally defined art as what a person makes with the intention to be art. I use this definition because it makes no statement about quality. This definition isn’t changed by AI. As long as a human is directing the tool it’s no different than someone using a camera to take a picture rather than painting it by hand. The idea of an artist, and the fidelity to the idea, are central to the quality of the work. The reason that a lot of AI art looks like dross right now is that a lot of people take the limits of the machine as limits on their vision.