The Strange New World of AI Art

Last year I spent three hours refining a single image. I wrote and rewrote prompts, adjusted parameters, generated hundreds of variations, selected and refined and iterated until I got something that matched the image in my head. When I showed it to someone, they asked who the artist was.

I didn't know what to say.

The honest answer is complicated. The AI did the rendering. I did the directing. Neither of us did the thing alone. But "co-created with AI" sounds evasive, and claiming full authorship feels dishonest. We don't have good language for this yet.

The Wrong Debate

Most public conversation about AI art gets stuck on a binary: is it real art or not? This is the wrong question. It assumes art is a natural kind with clear boundaries, and that whether something counts depends on where it falls relative to those boundaries.

But art has never worked that way. Duchamp's urinal. Warhol's soup cans. Cage's silence. Photography itself. Every generation argues about what counts, and every generation's boundary-policing gets overrun by the next. The category "art" is a social agreement, not a natural fact.

The more interesting questions are about what AI art does to us. How does it change what we make? What we value? How we relate to images and to each other through images?

Tell me more about co-creation aesthetics

The Experience of Making

There's something strange about generating images with AI. It feels different from drawing or painting or photography, and not only because the motor skills are different.

When you draw, there's a direct loop between intention and mark. Your hand moves, pigment transfers, you see the result, you adjust. The feedback is continuous and physical. You learn what you can do by doing it, and your skill is inseparable from your body's trained capacities.

With AI generation, the loop is different. You describe in words what you want. The machine interprets your words through its own learned patterns. You see the result, which is never quite what you described, and often interesting in ways you didn't expect. You refine your description, try again. The feedback is linguistic and conceptual rather than physical.

This isn't worse. It's genuinely different. The skill being developed is the skill of description, of articulating what you want in ways the machine can interpret, of recognizing when an unexpected result is actually better than what you asked for. It's a new kind of creative competence. Some people are much better at it than others.

The Aesthetic of Excess

AI art has a distinctive look. Even when it's technically impressive, there's often something too much about it. Too many details. Too perfect symmetry. Too saturated color. An aesthetic of overflow.

This makes sense when you think about how the systems are trained. They learn from millions of images, absorbing patterns of what makes images visually striking. They converge on a kind of maximum stimulation: every technique for visual impact applied simultaneously. The result is images that are impressive on first glance but often exhausting on closer inspection. They don't know when to stop.

Human artists learn restraint. They learn negative space, understatement, the power of what's left out. AI systems tend toward the opposite: why have one dragon when you can have seven, each with more ornate scales than the last?

This is changing as the technology develops. More sophisticated systems can produce subtler work. And artists working with AI are learning to counteract its maximalist tendencies. But the underlying pull toward excess remains. It's built into how these systems understand what makes images good.

What Gets Lost

Something is lost when images become cheap to produce. Not quality, necessarily. AI can produce technically accomplished images. What's lost is the relationship between effort and output.

When a human artist creates something beautiful, part of what we value is the time, the attention, the developed skill that went into it. The image is evidence of someone's sustained engagement. This doesn't make it better as an image, but it makes it mean something different. It's a record of a human being's investment of their finite time on earth.

AI-generated images don't carry this meaning. They're not records of human investment. They're outputs of computational processes. They can be beautiful, moving, surprising. But they don't testify to anyone's life in the way traditional art does.

Tell me more about what gives art value

Whether this matters depends on why you care about art. If you care about visual experience, AI can deliver. If you care about connection to another human's vision and labor, AI complicates things.

The New Collaborators

I've started thinking about AI image generators as a new kind of collaborator. Not a tool exactly, because tools don't have aesthetic tendencies of their own. Not an artist, because they don't have intentions or experience. Something in between.

Working with them is like working with a very skilled assistant who has their own style. You give direction, they execute, but their execution always carries traces of their own predispositions. You push, they push back. The result is a negotiation.

This is actually how a lot of creative work happens. Directors work with cinematographers. Writers work with editors. Musicians work with producers. The idea of the solitary creator is always somewhat mythical. AI just makes the collaboration stranger because the collaborator isn't human. The concept of authorship gets fuzzy.

Where This Goes

I don't think AI art will replace human art. But I think it will change what human art is for.

When images are cheap, what becomes valuable is what images can't easily provide: the story of their making, the relationship with their maker, the context that gives them meaning beyond their visual properties. Human art will increasingly be valued for its human-ness rather than its visual accomplishment.

This has happened before. When photography made realistic representation trivial, painting stopped trying to compete and went abstract. When recorded music made perfect performances reproducible, live performance became valuable for its imperfection and presence. Every time a machine masters a technique, humans find new ways to make the human part matter.

---

That image I spent three hours on? It's still on my wall. I look at it sometimes. I know exactly how it was made, what I was trying to achieve, which happy accidents I kept and which I rejected. It means something to me that has nothing to do with whether it's "real" art.

But I have no idea how to explain that meaning to someone else. The strange new world of AI art is full of experiences we don't yet have words for. We'll have to invent them as we go.