I’m pretty sure that you know French Quarter hasn’t really been destroyed. I created this synthetic image from one of my photographs that I scanned into my image generator. This what it came up with on its first try. Further evolutions only made it worse. Yes, I experimented but I like the first iteration from the very start.
This is a hybrid image. What does hybrid mean? There are no copyright considerations. The image you are looking at is based on my original photograph. Apparently, that’s rare.
Most creators are generating images from whole clothe. They type in a few words, the program adds more words that are mostly technical and the software generates an image based on what you wrote.
Here’s where the problem lies.
Who owns this newly created image? Who’s copyright was violated? Ah, copyright. A set of good, but complicated laws that are a mix of black letter law and court rulings.
We know that the software can’t hold the copyright. It can’t create. Only humans or one of my dogs can do that. My dogs are cocker spaniels. They think they can do everything. That aside, aside…
Every photograph anyone makes is covered by copyright laws the minute you press the shutter button. To collect damages the image must be registered with the copyright office.
All of that said, the way that Generative AI gathers data is to scrape from every available source with no regard for ownership. After all, every image is delinated into ones and zeros.
I’m fairly sure that the other Storyteller, the one on Wordpress, has been scraped multiple times. It’s very image heavy because for the last twelve years I’ve been posting at least one picture daily. What a gold mine.
Of course, only ones and zeros are scraped. Once the data is collected it is dumped into large cauldron with everone else’s ones and zeroes and converted into a gumbo.
For instance if I’m using Nightscapes, which is a Dall-E’s product, and type in a few prompts it spits out a brand new image. Who owns the copyright? In theory I do because I created it.
Remember Generative AI can’t create. It can only shuffle and sort.
The question becomes this. I might have assembled it but whose data did I use? It could be one photographer or it could 4,672,980 and eleventy photographers.
Everyone of those photographers holds a copyright on the image that scraped. How do they get paid and by who? At what rate?
Some people have been kicking around the idea of using the Spotify model. When a song is streamed the artist receives 0.00174 cents per song. At that rate it will a decade before I earn 32 cents.
If you are the writer and own the rights to the song there is more compensation. Well, I’m the photographer and I own the rights to my photographs. How do I get paid?
Eventually, this will be settled in court. Either we’ll be ignored outright, or a system will be created that pays us a teeny tiny little royalty everytime our data is used to create an image.
Musicians are even more worried. Spotyify is already using Generative AI to create new and very boring songs curated into playlists that they might call “Music For a Sunny Foggy Smoggy Rainy Day.” It’s blander than elevator music. The musicians are getting organized to fight this as we speak.
I believe anyone who makes any art that can be duplicated or mass produced should pay close attention.
Now.