What We Didn’t Imagine

I have always thought that making something was largely a matter of choices.

Some years ago, I started painting on canvas, with acrylics as the medium. Each time I started a piece, I began with an idea of what I wanted the painting to become.

But it never quite worked that way. Each painting wound up different from what I envisioned when I started, some of them a lot different. Somewhere along the way, the painting would begin to tell me something I hadn’t anticipated, and I would listen to what it said. The finished work became the result of an internal conversation between two forces; intention and discovery.

While I am no longer painting, I have been thinking about that conversation lately. The reason is that a new tool has emerged; artificial intelligence.

There is considerable resistance to this medium among visual artists, and I understand why. For someone whose life has been built around mastering a particular craft, a machine that can produce a compelling image in seconds can seem not just like a new tool, but an assault on the value of the craft itself. I’ve heard colleagues dismiss AI-generated work as “AI slop,” and I’ve seen people refuse to use any generated imagery in professional communications—as if the mere presence of it contaminates the seriousness of the work. That reaction isn’t trivial; it comes from a deep sense of identity and from the fear that something essential is being bypassed.

Ted Chiang makes a thoughtful version of that argument. He suggests that art is fundamentally about choices: the thousands of decisions an artist makes, from the largest elements of a work down to the smallest. A painter makes those decisions with every brushstroke. A writer makes them with every word. If a machine makes most of the decisions, what exactly is the human artist doing?

It is a good question, and it is compelling because it honors the labor of craft. But my experience with generative AI has led me to a somewhat different question: Must the artist make every decision, or must the artist know which decisions matter?

Consider what happens when I use Midjourney for the images on my website.

I don’t just type a description and accept the result. I begin with something I am trying to say. The machine gives me possibilities—sometimes dozens of them. Most are not quite what I want. Some are interesting but miss the point. Occasionally one makes me stop.

That moment is important.

I may not have imagined that image at all. In fact, it may be nothing like what I thought I was looking for. But when I see it, I recognize something in it. I know that it belongs.

That happened recently when I was looking for an image for an essay about my parents. We (and I say “we” advisedly) tried symbolic images, landscapes, empty chairs, and various artistic styles. None felt right. Eventually we arrived at a stylized portrait of my parents on their wedding day in 1943. Suddenly, there they were: my father in his Army uniform, my mother in her dress and hat.

The image was romanticized—almost impossibly so. But it fit the essay perfectly, and had just the right tone.

Why?

Because the essay was about two people. And the image finally was, too.

The machine did not know my parents. It did not know the promise I had made to my father, or the seventeen years that followed, or what it meant to bring my parents together again. It could not know why an empty chair for an illustration felt like a cliché to me. Those things came from my life.

But it could show me something I hadn’t imagined.

That is the part of this tool that interests me most. It does not eliminate the artist’s choices, but it changes the nature of some of them. The artist becomes not only a maker, but a chooser, editor, director, and—perhaps most importantly—a recognizer. Intention begins the work, but discovery and recognition completes it.

This choice isn’t new. Artists have always discovered things they did not intend to make. The unexpected qualities of the material is part of the process. A painter responds to what appears on the canvas. A photographer responds to what happens in front of the camera. A filmmaker discovers something in the editing room that wasn’t apparent during the shooting.

Artificial Intelligence makes the unexpected arrive much faster, and in greater abundance.

That may be unsettling. It may even change what we mean by artistic skill.

But I don’t think it eliminates art.

There is room for the painter who spends six months making a canvas, and there is room for the person who works with a machine to discover an image neither of them could have imagined alone.

The important question is not “who made every stroke or mark?”; it is “who knew what the work was trying to become?” For me, that is the most important artistic question of all.