AI Image detectors are quite interesting, because they provide false information in a wide assortment of ways. In fact, I think they are a little TOO extreme in what and how they classify collage-type images. It is evident that they cannot determine the difference between art that was made with AI and art that is human created that happens to have some AI elements within it’s creation.
Something I have played with is the use of minimal AI in collage images. Where you have 90-95% of the work being completely original with a few small AI-generated pieces brought in as minimal components. Let’s say you are looking for a very specific mechanical component. Stock agencies don’t have them and anything you can find online are likely copyrighted and protected. So, you can generate them using whatever AI model you choose. Now yes, this does get into ‘what do these models train these image on’ discussion, but that’s a conversation for another day.
I have brought up the ethical use of using AI-generated content within collages before and I have been struck down by art purists, NOOO you can’t do that. Yet these same people support collage artists incorporating copyrighted magazine images or newspaper content into their work and they defend that as art. This led me into thinking, how can one use small pieces of AI-generated art as an element of a collage. Not as the entire generated piece, just as a small component. And, I think there is a balance that can be discovered.
The following piece is one that was created using AI, but in only one aspect – the woman in the lower left corner.
I added the screen-capture of this image with the Photoshop layers to show you that this is an actual composition – it is NOT generated by AI. The central woman in the photo was taken from old vintage photographs I own. All of the other components are vector images, hand-drawn script, original textures, and more. The woman in the corner, much like much of the image, was processed with a halftone effect and some significant contrast and coloring changes.
As anyone would probably agree, this composition was NOT generated in AI. I can show the layers, I can show the process, I can show all of the components that went into this, I can show original photographs and content where the other images originated. Only one corner of this image was generated by AI. And this is where the problem comes up.
Now sure, I can replace that woman with any other woman from the thousands of vintage photographs I own. But that defeated the purpose in this series. I wanted an AI element in this, to show that you can use it with reason. Not as a full image, but as a tiny little inconspicuous part.
So, I ran this through AI detectors. The result was “This image was 100% created by AI or the majority of the image was created by AI.” As I am showing you my layers, this was not generated by AI. This makes me think, how ACCURATE are these AI-detectors? We are looking at maybe 5% of the image that was generated by AI. And it isn’t even the focus of the image itself, just a part sitting in the corner.
This led me into a second experiment, to see how flawed these AI-image detectors are. And, as I suspected, they are very flawed.
The original content for the next image was, indeed, made by AI. However, I printed the image, hand-cut a stencil from that original image, and spray-painted the stencil using a couple colors. Hand-cutting a stencil and spray-painting the stencil elements are ALL human aspects of art. The following image does NOT have ANY AI art in it whatsoever. Only the source image was created by AI. The output is 100% human.
This begs the question, at what point is AI-detection going too far? And what does this mean for artists who use different types of content in artistic ways? Do collage artists now need to use ONLY their images, not sourced from magazines or newspapers? Does this mean a stencil artist cannot use the image of a famous celebrity as a starting point? Of course not.
When I say that artists CAN use AI in their work, this is exactly what I mean. Not as a ‘prompt this, here’s the image, poof it’s art.’ But it can be used as a tool, as an element, as something that becomes part of the art itself.
The piece above has a few AI pieces in it. The faces are, indeed, all AI. However, this is part of an After Effects animation sequence I created. I used a really old AI model to created these weird, distorted faces, images I would not be able to find in real life. Here’s the thing, they were not meant to be the artwork itself. They were meant to be motion, to be recognizable features, to be a small part of something much bigger. None of this composition was created in AI. Only the parts of the faces were. And they are moving, they are morphing, they are constantly changing. It is a kaleidoscope with a vast assortment of components. Was this image created by AI? No, I created the animations, the motions, the movements, the effects, the visuals. Only parts that are meant to morph and change second over second, frame over frame. But, you put this still into AI detectors and the result will be, this was 100% created by AI.
Now, in all transparency, I don’t like AI art at all. It is, however, something that is with us and will not be going away. Instead of turning my back and shaking my fist, I want to find ways where I can find a balance using the platforms. Some will say there is NO ethical use of AI in art because stealing (fair enough, but please hold those standards to other forms like magazine collage). There are others (egads) that say AI IS art no matter what (no). And I personally think there there is a balance somewhere in between. Somewhere that critics can say yes, they may have used some AI in the piece but the artist was still the originator of the art and directed the vision with their tools and skills.
When I create artwork, I know exactly what I am doing and why I am doing it. I don’t do things for ease, I do things for a purpose. And I create with intention. I see AI as just another tool like a stencil, a paintbrush, a sheet of handmade paper, a can of spray paint, or a 3D model. No tool defines my art, only I can define my art. Yes, I put the work in, I design, I paint, I create.
But, there are people who use it as an “insert prompt here.” Yes, this is the major problem.
The terrible thing is… AI detectors cannot determine the difference. They only identify a part of the design that may have been created by AI, so the whole thing is now labeled as AI artwork. I am NOT an AI artist and I NEVER want to be labeled as one – oh please Lord no. I just happen to use minimal components as tools every now and then. Judge me, I suppose.