I don’t know about the novelty, but via Astral Codex Ten, here is an AI artwork characterized as showing “hidden but obvious spirals.” I agree with the obvious, not with the hidden.
I have to admit, I quite like this image. I find it surreal in an appealing way. But why would anybody characterize the spirals as “hidden”?
Now, this next one is supposed to show “hidden but obvious text,” and this is really interesting because this time I’m like, it must be well hidden, because I can’t see it. So this time, I’m saying the text may be hidden, but it is not obvious — at least, not to me.
Can anybody see the text? What does it say? Can you tell me where to look? All I see is a capital “N” at the top, and I’m not sure about that.
I’m not as opposed to AI art as AI text, which is (I’m very aware) because I’m a text generator myself, not an art generator. But I’m pretty sure I will never use AI-generated cover art for a book because (a) AI text generation definitely plagiarizes, so I’m quite willing to assume that AI art generators plagiarize as well; and (b) KDP asks you to declare that your books are free of AI-generated content (they permit, at least for now, “AI-assisted content,” and here is where they describe what they mean by these terms). Given this requirement, I’m 100% for sure not at all remotely tempted to use any cover generated by AI, and not only that, I’m asking cover artists to tell me flatly that they aren’t generating covers using AI tools so that I have a paper trail if necessary.
I didn’t bother for the Invictus cover artist because after all, that video where he shows the creation of the cover is pretty conclusive. No AI tools in use there. I would bet that cover artists are all going to be getting this question and all the reputable ones will be putting statements on their websites about it. I mean, I sure would. You know what else? I would produce videos just like the artist who did the Invictus cover to prove that I was doing the art myself. Maybe not for every cover I made because that would be a headache, but for enough to make it clear that I was an actual artist for real.
Please Feel Free to Share:
It is very obvious and spells New York, mostl with artificially heavy shadow.
I can read the word “New” at the top (and if you hadn’t mentioned the capital N I wouldn’t have known where to look). No other text that I can find; perhaps other people will?
Oh! Now I see York as well. Thanks!
For me it’s a function of distance. “New York” is subtle close up, and increasingly obvious as I pull my phone back.
At first glance the text in the second picture was hidden, but now I can’t un-see it. It Says New across the top and York at the bottom, all basically in black.
Thanks, everyone! I’ve got it now! Mike, it’s a LOT easier to see on my phone than on my monitor screen, and Pete, it made all the difference to know I should be paying attention to the shadows, not the buildings.
AI art training sets are, as a rule, entirely sourced without the artists’ permission, and often allow someone to specify “in the style of [artist name]”, so definitely unethical. In other arenas, AI training sets are required to be properly licensed, but AI art generators aren’t doing that currently. There will probably be legal challenges down the road, so steering clear of AI art is also safer, not just more ethical.
So that’s exactly how “training” works for art as well as text generators. Ugh.
With the new AI box on Amazon, I think authors are playing with fire if they touch AI tools at all. Huge risk. Huge.
Oddly enough, I could see the N and Y, and deduced (but could not see) “New York”, but when I opened the post to comment, I saw the rest.
Then, I could often shift the color of the dress while looking at it.