The Responsibility of the Seller
I recently found myself shopping online for a novelty sweatshirt. The kind that was inevitably going to be an AI design from a POD seller, because that's just the way the world works now.
I already knew before searching that the artwork on said sweatshirt was probably going to be AI-generated. I'm not morally opposed to that. I simply wanted a design I liked that had the messaging I was looking for in the moment.
And I found one.
At first glance, I really liked it. The overall image was interesting and more unique. The subject was appealing, the colors were good, and it had the sort of visual quality that made me stop scrolling.
But then I took a good long look at it. Like, really looked at it.
And there they were, small but glaringly stupid failures that became obvious when I stopped looking at the image as a whole and started looking at the detail inside it.
My reaction was probably the same one that any good designer would have. Why didn't you fix this?
In the end, I didn't end up buying anything.
But that question stuck with me.
Because I know how easy it would have been to fix at least some of the errors.
And if I were selling that sweatshirt... I would have fixed it.
I don't actually expect AI image generators to produce perfect artwork. That would be a ridiculous standard. Anyone thinking it should has the wrong expectations (at least for now).
When I use AI to generate an image and it gives me something I like except for one obviously malformed object, my first response is "can I fix this, is it salvageable?"
That's just part of the process. Or... that's what the process always was.
Perhaps this is where my background in illustration makes me see the issue differently. I'm accustomed to digital artwork being a process of making something, looking at it, finding the things that don't work, and fixing them.
Sometimes the fix takes five seconds. Sometimes it takes an hour. Sometimes you decide the whole thing is a mess and needs to be redone. But there is an expectation that you look, you fix, you make it good.
So if someone is selling a piece of artwork on a sweatshirt, why would they upload something containing obvious visual errors? Didn't they notice them? Didn't they know what to look for? Did they not have the skills to correct them?
Maybe they generated hundreds of images and simply picked the one that was "good enough"?
Maybe they know the errors are there and have decided that most customers won't notice or care?
I honestly don't know.
I don't want to assume that every person selling sloppy AI artwork is lazy or dishonest. There are probably plenty of inexperienced people who simply don't yet know how to evaluate what they're producing.
But that raises another question.
If you don't have the skills to evaluate the output, why are you selling it?!!
That's not an anti-AI question. It's a question about responsibility.
If I decide to sell furniture, I don't get to say that I wasn't responsible for the wobbly table leg because the power tool made the cut.
If I sell a photograph, I don't get to blame the camera because I failed to notice that somebody was blinking.
If I sell a digitally designed product, I don't get to blame Photoshop because I forgot to proofread the text.
And if I sell an AI-generated image, I don't think I get to blame the AI because I didn't notice that something looks ridiculous.
The tool doesn't own the finished product.
I do.
Maybe "AI slop" isn't really about AI
There has been a lot of discussion about "AI slop" lately, and I totally understand why.
There is an enormous amount of AI-generated material being produced simply because it is incredibly easy and inexpensive to produce enormous amounts of it.
But I'm increasingly outraged at the idea that the problem is simply that it's AI-generated content.
Because I don't think the public necessarily rejects AI as categorically as some of the headlines suggest.
Research from Pew suggests a much more complicated picture. When Americans were asked what they would think if they discovered that a painting they liked had been made by AI, 49% said they would like the painting less. But 48% said it wouldn't change their opinion. That's hardly a universal rejection of AI art.
At the same time, there is clearly a desire for transparency. Pew found that 76% of Americans considered it very or extremely important to be able to tell whether content was made by AI or by people. Yet more than half said they weren't very confident in their own ability to tell the difference.
So the real dividing line isn't simply AI versus human. It's more about care versus carelessness.
When I look at an AI-generated image, I tend to scrutinize it. It's partly habit and partly because I've spent years looking at my own artwork critically. I notice when something doesn't quite make physical sense. I notice awkward perspective. I notice when an object has been drawn incorrectly, or in a non-human way.
But not everyone looks at artwork that way.
Most people shopping for a sweatshirt aren't zooming in on a listing image looking for anatomical errors.
They're looking at the overall impression.
I like that.
That's pretty.
That would look good with jeans.
My friend would love this.
Click.
Buy.
And maybe they never notice the problem. Maybe they don't care. Maybe they only notice when the sweatshirt arrives and the flaws are suddenly sitting six inches in front of them.
Or maybe they notice immediately and don't care because they still like the design.
That's their decision to make.
But it does make me wonder how much poor-quality work survives simply because people aren't examining it closely enough to notice. Or being bothered enough to complain.
And if that's the case, what incentive does the seller ever have to improve it?
The economics of "good enough"
Before generative AI, creating a complicated illustration could take hours or days. If you'd invested that much time into something, there was a fairly obvious incentive to make it as good as you reasonably could.
You'd refine it. You'd correct things. You'd fix the ugly bits. You'd make the investment of your time worthwhile.
Generative AI changes that equation. If I can generate fifty images in an afternoon, the cost of throwing away forty-nine of them is almost nothing.
And that creates an entirely new temptation:
Generate. Choose. Upload. Repeat.
The faster production becomes, the easier it becomes to stop caring about any individual piece.
Why spend twenty minutes correcting something when you can just generate another one?
Why spend an hour cleaning up an image when there are another hundred waiting to be generated?
Why worry about a small defect when the customer probably won't notice?
This is where I think the "slop" problem becomes less about the technology itself and more about the economics surrounding it.
AI has made it possible to produce an enormous amount of visual material very cheaply.
It has also made it possible to produce an enormous amount of mediocre visual material very cheaply.
And unfortunately, those two things go together.
AI can be a tool without being the artist
I am not opposed to AI being used in creative work. In fact, I use AI myself. If you follow me regularly you know this. You then probably also know that I'm not all-in on it. I still make my own work.
When using AI, what matters to me is what happens between the initial generation and the finished product. There is a meaningful difference between: "I told AI what I wanted and it made this." and "I used AI to help me create this, and then I made decisions about what worked, what didn't, what needed to change, and what ultimately deserved to be sold."
Those can look identical to a customer. But they're not the same creative process.
And I don't think the difference is necessarily measured by how many hours a human spent physically drawing the final image.
A person can spend twenty hours making something terrible.
A person can also spend twenty minutes making something wonderful.
The number of human minutes involved isn't the measure of creative value.
Judgment matters. Discernment matters. Taste matters. Editing matters. Quality control matters.
Knowing when something is wrong matters.
And knowing how to fix it, or knowing that it needs to be fixed, is a skill in itself. And there is a value in that, even if we can't see it.
The human still has a job
Perhaps this is the part that matters most to me. I don't think the human role in AI-assisted creative work disappears. I think it changes.
The human may no longer be responsible for drawing every single line. But somebody still needs to decide what the image should be. And that somebody should probably be a person. A person needs to decide which generation works, needs to recognize bad anatomy, needs to notice the impossible perspective or that the composition is cluttered. A person needs to realize that the color isn't working, or needs to say, No, that's not good enough. And then, ideally, a person needs to fix it.
That last part is important.
AI doesn't eliminate the need for creative judgment. It makes creative judgment more important.
When everyone has access to a machine capable of producing a superficially impressive image, superficial impressiveness becomes a pretty low bar.
The value skill becomes knowing what happens after the first impression. Does the image actually hold together? Does it make sense? Does it communicate what it is supposed to communicate? Does it look intentional? Does it survive scrutiny? Is it finished?
I don't need every piece of artwork I buy to have been made entirely by human hands and I don't think using AI automatically makes a piece of artwork worthless. I don't even think AI-generated artwork is inherently bad.
What bothers me is the idea that the existence of the tool somehow excuses the absence of care.
If you're going to use AI to make something you intend to sell, then look at what it made.
Really look at it. Don't just admire the overall effect. Inspect it. Question it. Be critical of it. Fix what needs fixing. Throw it away if it doesn't work.
And if you don't have the skills to recognize what is wrong, perhaps spend some time developing those skills before turning the generator into a business.
Because ultimately, I don't think I'm asking for some impossible standard of perfection.
I'm asking for the person behind the product to care enough to notice.
I don't need to know that a human personally drew every pixel.
I don't need to be told that someone spent 40 hours painstakingly illustrating something that could now be produced in a fraction of the time.
But I do want evidence that a human being was actually paying attention.
If AI is going to be part of the creative process, the very least we can do is remain creative enough to notice when it gets something wrong. And then fix the fucking design.
0 comments