Why I'm not worried about AI art replacing artists

The discourse around AI art has settled into two camps. One says it’ll destroy art. The other says it’s just a tool. I think both are wrong, but the second is closer.

What AI image generation actually does

Midjourney, Stable Diffusion, DALL-E. These tools turn text prompts into images. The quality is now good enough that generated images win art competitions, fill stock photo queries, and produce concept art faster than a human illustrator.

What they actually do, at a technical level, is pattern matching at enormous scale. They’ve learned the statistical relationships between text descriptions and visual features from billions of image-text pairs. They don’t understand what an astronaut is. They know what pixel patterns are associated with the word “astronaut.”

That’s not a small thing. But it’s not the same thing as making art.

The photography analogy

When photography was invented in the 1830s, painters panicked. The painter Paul Delaroche allegedly said “from today, painting is dead.” It wasn’t. What happened instead was interesting: photography took over the documentary function of painting (portraits, landscapes, records), and painting was freed to become something else. Impressionism, abstraction, expressionism. Photography didn’t kill painting. It pushed painting toward what only painting could do.

The pattern repeats. Synthesizers didn’t kill orchestras. Digital recording didn’t kill live music. CGI didn’t kill practical effects (ask Christopher Nolan).

Each time a technology automates the mechanical part of a creative process, the creative process shifts toward the parts that can’t be automated. The parts that require human judgment, human emotion, human weirdness.

What AI can’t do

AI image generation can render light on skin. It can compose a scene following the rule of thirds. It can generate in the style of Vermeer or Monet or Greg Rutkowski. It can produce images that are technically flawless.

It can’t decide what’s worth looking at.

It can’t know that a photograph of a man sitting alone on a park bench in November says something about loneliness that no prompt can capture. It can’t recognize that the most powerful image in a series is the one that breaks the pattern. It can’t understand that the slightly out-of-focus background in a portrait is what makes the eyes arresting.

The seeing. The choosing. The noticing. That’s the human part. That was always the human part. We just used to conflate it with the rendering, because the rendering required skill that took years to develop.

Now the rendering is free. The seeing still costs a lifetime.

What will change

Some things will change, and we should be honest about that.

ArtStation illustrators who do commercial work (book covers, game concept art, marketing materials) will face pressure. Some of that work will be replaced by AI generation. The economics are brutal. An AI image costs cents. A human illustrator costs hundreds or thousands of dollars.

Stock photography, as an industry, will contract. Not disappear. But contract.

The changes are real and the people affected deserve empathy, not dismissal.

But artists? The people who make things because they need to say something that can only be said visually? They’ll be fine. They were fine after photography. They were fine after Photoshop. They’ll be fine after this.

The MoMA won’t close. Galleries won’t shut down. People won’t stop being moved by a painting they stand in front of for ten minutes because no JPEG has ever replicated that experience.

The tools change. The human need to create and to witness creation doesn’t. I’m holding on to that.

Related thinking: