Workflow to Find Non-AI Art

[Moonsong]

These days, I often find myself with a problem that is becoming pervasive: it's become increasingly difficult to find non-AI art.

It goes like this: I am hurrying to get a pitch together for a client, or to publish the next installment of my serialized fiction on Substack. I want to find a picture of a 'teetering tower,' but as a part of my own personal code around this, I don't want to use something AI made, even if it is indistinguishable from a human work. I google 'teetering tower, fantasy, art, non-ai', and am immediately wading through a deluge of art published on sites with names like "dreamstime" or "peakspix" that is (sometimes obviously, and sometimes very not obviously) generated by AI. I try Artstation and face the problem of their horrendous search interface. I try Pinterest, get excited about an image, and then spent ten minutes trying to determine its source and authenticity.

There's a lot that can be said about this, and that has been said about it elsewhere. I'm not here to argue the merits of using or not using such art.

What I want to do is offer something simple that has worked for me in quickly finding pieces of work that were made by human hands using pre-AI-era tools.

1. First step is google image search, but with a critical addition. I add "before:2020" to the search. No spaces, exactly as written, just add to the end of whatever I am searching for.
2. Once I find something, I check whatever information is available easily as to the attribution. I'm searching for specific artists' names, and where it was first published.
3. If this is not evident, I reverse image search the image. These days, Gemini will often give me some information about the artist, if it is available on the open web.
4. Check the usage rights for the image, and if you use it publicly, attribute it to the artist.

I know there are probably much savvier ways to do this, but the reason I like this one so much is that it is highly flexible and very fast. It has worked well for me when I am dealing with extremely tight schedules, but do not want to compromise on my principles when it comes to using AI in my workflow.

I hope this is helpful! I'm very curious to know what other tools people suggest for this growing problem.

 

[skip.knox]

I used to be able to add exclusion strings: -getty for example. Because Getty and other entities have tied up billions upon billions of images. Copyright law does not prevent them from doing this, but it sure screws up searches. I'm a historian, and I go looking for images of, say, Emperor Frederick II. There is exactly one medieval illustration of ol' Fritz, plus some coins.

What do I find? Those images, plus plenty of 19thc silliness that I do enjoy, so long as I'm writing fiction and not history. But that image, which is something like seven or eight centuries old, has a watermark from Alamy or Getty or Dreamtime and a dozen also-rans. I can't use that in the classroom. For thirty years I've turned to the Internet for images to complement my history lectures.

And now, even the Boolean operators that Google once honored are now largely ignored. The before: operator does seem to work, though I had to roll it back to before:2000. The downside there is that older files tend to have lower resolution, which means they aren't going to project as well. ... ... I started this sentence a number of times but it kept ending up in an inappropriately passionate political rant. Inappropriate for a writing forum, anyway.

So, I'll just say thank you to Moonsong for the tip. Reverse searching on medieval illustrations is not one I've played with. A quick rabbit hole on stupor mundi showed me there many an hour I could spend there. So, double thanks!

As for other tools, does getting out of Googleverse help?

 

[A. E. Lowan]

Just a wee missive from the trenches. I've been doing this since I was cutting out magazine pictures and glueing them into notebooks. That took a lot of time. I'd also draw characters. That was great for character development because I'd be able to focus on them and really do a good deep dive to figure out who these people were. But that was college, and I got away from art for a long time. Just now I'm trying to get back in practice.

(I'd show off, but Website Dummy Boy won't let me upload.)

Google has enough of my information. I see no reason to make their lives any easier. I spend a fair bit of time on Pinterest looking for inspiration pics. The AI Boom-Chicka-Boom-Boom Part II: Electric Boogaloo Sucks to be You has already spread far enough that there is no way to avoid it. Good intentions are great, but it's like trying to put out a fire at the Louve with a Red Solo Cup. It's just now impossible. But I also don't see any reason to reward banditry, so I do the only thing I really can. I don't share inspiration pics in public, and when I do spot stolen images, I report.

 

[RoccO]

I find it funny how AI messes up the algorithm with their own creations. They have this neat thing where they can zoom in however. But the search for images gets quite accurate I know. My mother has more knowledge than me on ChatGPT, and I have seen YouTubers explore the merits of different search engines. The answer I have is unique but not unheard of, ask for images that are popular, because chances are the popular images are the ones you want.

I am however wary of the AI generated images taking over the market, purely because it requires no human skill or ingenuity. There are a number of mediums that simply move on, take modern painters, Bob Ross is the only one who remains a legitimate interpretation of modern art. Through the decades you see artists emerge in the graffiti or screen printing trade, but it will never be the renaissance period again, there will always be a new art form.

Whilst I was a terrible artist, I took notice of all the things you could write about the artworks, and I read other opinions to gauge what was interesting. That I think is similar to how AI would view the question. They will look for the best example, not the most accurate, based on what is popular, but not what is accurate. There I wind up where I started with the opinion that it is slanted, whichever way you look at it.