Another View on Generative AI
In Today's Journal
* Admin Note
* A New Short Story
* Another View on Generative AI
* Of Interest
* The Numbers
Admin Note
The Journal Website (by WordPress) is getting more and more iffy. Yesterday I was unable to upload the two Traits charts even though they're small files.
If you currently view TNDJ on the website, I strongly recommend that you switch over to reading it in your inbox or on the Substack.
A New Short Story
"Mistaken Identity" went live yesterday at 10 a.m. on my Stanbrough Writes Substack. Go check it out.
If you enjoy the story, please click Like. That helps with my Substack algorithms. Then tell Everyone else. (grin)
Bradbury Reminder
Today is Saturday. Just a reminder to get your Bradbury Challenge story info in to me before the Journal goes live on Monday.
Another View on Generative AI
a guest post by Sean Monaghan
I'm feeling concerned about the advent of AI into the creative sector. ... As a writer I'm feeling it as editors and slush readers at magazine to which I submit find themselves awash in AI-generated stories.
Apparently these stories are fairly obvious, but they still take time to work through. Sometimes it might not be for a page or so that it becomes clear that a story has been generated by AI.
But these are early days. I’ve seen the rapid improvement in the quality of AI art, and I suspect that better AI writing may not be far away.
I have seen some pretty memorable quotes lately, which have seemed to crystalize some of the general concerns about AI in the creative realm. I quite liked this one:
“The underlying purpose of AI is to allow [the] wealthy to access skill while removing the skilled from the ability to access wealth”
The art is amazing, but I'm struggling with the moral side of [using AI], let alone with the lawsuits. From what I see, these companies have effectively stolen artists' work.
The companies are in the business of creating derivative works, which is something protected by copyright. But the companies claim 'fair use' (it's not) and it seems the onus is falling back on the artists to prove that the works are derivative.
Copyright exists to protect the rights of creators so that they can generate income from their intellectual property. This includes the rights to derivative works.
A musical version of one of my science fiction action stories may bear little resemblance to the original work, but it’s still a derivative, and I will have received a licensing fee from the musical’s creators.
In general, it seems to me that AI companies are creating derivative works without paying such licensing fees.
For all my books I have used licensed art, generally from Dreamstime, and on occasion from Pixabay. [Ed. note: I also recommend checking out Unsplash.] For a brief period I had some AI images on covers.
But when I realized how they’d been generated, I pulled their availability and changed out the covers for properly licensed art before putting the books up for sale again. All of this is relatively inexpensive, but I’m clear on the licensing.
I found another quote, from Elena Perez (@thisiselena.bsky.social) which to my mind nicely sums this up:
“Imagine if people went to retailers and just took stuff and called it ‘generative shopping’ and then, when the retailers said that’s illegal, people said 'regular shopping is too expensive so we have disrupted it' and then the retailers were like 'Oh! Carry on, innovators!'”
I know that the online graphics and text site Canva, which has an AI generator, have used their own library for training, but it appears that they have also gone beyond, and they state that they cannot guarantee that their AI creations do not violate copyright.
Similarly Adobe have firefly, which is fun, and based on their own stock collections, for which the contributors were compensated.
My suspicion with that one, though, is that it was an 'opt-out' in updated T&C, rather than an ‘opt-in’. Too easy for that kind of thing to slip by.
To me this is bears comparison to Napster (which outright stole musicians’ recordings) and Spotify (which compensates musicians with money calculated at tiny fractions of a cent).
Musicians I know are generally not fans of Spotify but what are you gonna do? Adobe and Canva are like Spotify, while it seems to me that MidJourney, Wombo, Dal-e and their ilk are like Napster.
For most of my books, there is legally licensable art readily available. Sometimes it’s imperfect, but it does the job.
One of mine in particular, Tramp Steamers, is a space adventure novel set on a world where trade is plied by vessels that ride anti-gravity fields over vast prairies. If you google the art of Ian McQue, you’ll see where I got the idea.
Ideally I would have licensed one of his images for my cover, but as a little indie writer, that’s outside my wheelhouse for the moment. But there was nothing that compared available from the regular art sites I use.
Ultimately I made my own cover, using my Bryce license. It’s way too amateur, of course, and so I’ve looked more deeply into getting an AI image using my Adobe license. I even had a go, and the results were far better. But that niggle inside me about how the actual creators are being compensated didn’t sit well.
I am so, so tempted, you know.
But with all those lawsuits and all these industries being disrupted in, I feel, a lackadaisical and even immoral way, tempted as I am, for now, I'm going to wait for the dust to settle.
Talk with you again soon.
Of Interest
So you want to be an indie author – now what? This will probably be helpful to some of you.
The Numbers
The Journal…………………………… 150
Writing of Blackwell Ops 37: Temple-Schiff
Day 1…… 2012 words. To date…… 2012
Day 2…… 2487 words. To date…… 4499
Day 3…… 4597 words. To date…… 9096
Day 4…… 2790 words. To date…… 11886
Day 5…… 3430 words. To date…… 15316
Fiction for February………………….. 20796
Fiction for 2025………………………. 142151
Nonfiction for February…………….... 6360
Nonfiction for 2025…………………… 38340
2025 consumable words…………….. 173981
Average Fiction WPD (February)…….. 2971
2025 Novels to Date…………………….. 3
2025 Novellas to Date…………………… 0
2025 Short Stories to Date……………… 3
Novels (since Oct 19, 2014)…………..... 107
Novellas (since Nov 1, 2015)…………… 10
Short stories (since Apr 15, 2014)……… 274
Short story collections……………………. 29
Disclaimer: Whatever you believe, unreasoning fear and the myths that outlining, revising, and rewriting will make your work better are lies. They will always slow your progress as a writer or stop you cold. I will never teach the myths on this blog.
Writing fiction should never be something that stresses you out. It should be fun. On this blog I teach Writing Into the Dark and adherence to Heinlein’s Rules. Because of WITD and because I endeavor to follow those Rules I am a prolific professional fiction writer. You can be too.
I just this minute noticed pixabay removed their AI content selection from the top right menu. I was afraid they had removed the selection to turn off AI content entirely. I subsequently discovered the AI setting has been moved to the All Images / Orientation / Size / Color part of the menu selection. It can be set to "Authentic Only". Thank goodness for that.
So, All gAI art no matter where it is from uses improperly acquired art. Because, the original images they were "trained" on were for scientific experiments, and thus were fair use for educational purposes. And then they opened the models (programs) to the public and started charging...which is not educational use. So, even if you have Stable Diffusion on your own computer and you train it on your own images that you created by photo, hand, etc, the base that it was trained with was improper. And yes, it adds noise to the picture, like added blur, but how do you think the programs that remove blur and sharpen your picture work?
So, Adobe Stock was inundated with AI stock images for more than a year before "AI" became a thing. And the artist/photographers just claimed it as theirs, so it would have their name on it. Same for most/all of the other places. how do I know? Being on the more arty side for a long time you hear about it. How good it had to be to be accepted. And yes, later on they purged whatever they could identify. But it's still contaminated, and that is what they are training their Firefly on. Same with other places, especially the ones that have their own gAI engines now.
So, more and more of the stock places are being "tainted" by being fed/scraping other gAI images, and by doing so less and less of the "real" images are still there.
Have you noticed that more and more gAI "art" images look the same? 9 out of 10 times I can spot a gAI image without it having funky fingers or eyes looking cross eyed, etc. It's the sameness that is the give away. Photos - a little harder, because photos are already digital numbers, right inside the phone camera or digital camera. So it's all being mixed together to become grey goo eventually.
I mean it's a lot like the old "chinese knock-off" bags or shoes of the past. They can look pretty good, look pretty close, but they aren't "quite" right.
The other problem with gAI is that you never get exactly what you want. So you have to "settle" for "close enough". Sure, I played with it in the early days - in the beginning with ArtBreeder, and Disco Diffusion and Wombo. And it was cool - like Spirograph, or splashing paint on a paper/canvas or doing acrylic pour and scrape or spraypaint art. But even after playing with Leonardo and MidJourney and BlueWillow and Limewire and Playgrounfai and CGDream and DeepDream and ArtFlow and the list goes on and on...I never ONCE got the image that was in my head. No matter how many tweaks and words and little tricks I tried.
Which means to me (and this is MY opinion) that it's not my art. It's very pretty and often detailed Pareidolia. (The meaning of PAREIDOLIA is the tendency to perceive a specific, often meaningful image in a random or ambiguous visual pattern. -Merriam-Webster)
My art may come from different sources, 3D, photos, hand-drawn, digital patterns (I use Daz Studio, PD Howler, Filter Forge, my graphics tablet, Affinity, Rebelle, GIMP and Corel and even a touch of Blender - have used pen and ink, graphite, watercolor, and colored pencil in the past) to make my cover and make my art. My inspiration, for art as well as stories, come from photos, taking drives, reading articles, pareidolia in the patterns around me, and yes, sometimes, *gasp*, even playing around with gAI which is like doodling to me because it's like throwing torn up magazine pages on the floor and seeing what it looks like. That can be inspiring to my brain, which is a real intelligence informed by soul/spirit.
Legally, I don't have to worry about that because I'm creating my own things (although I do have to watch out because some unscrupulous people will rip off from movies or video games and then sell those assets - and I don't play video games or watch many movies, so I often don't know that it's been done...so I do have to do some due diligence - and don't get me started on fonts...that's a whole 'nother rant!) .
But, most of those places (like Canva and Creative Fabrica) will hang you out to dry if there are legal issues, because there is a tiny line somewhere in their ambiguous ToS that they can point to and say, "tough shite - not our fault - you should have done our job for us and checked it out better". So, even if you have PAID for a commercial license, you are not guaranteed that you actually have one that will protect you. There are predatory lawyers out there who specifically look for certain images or things very close to those images and then sue the users, and then they settle before even going to court for hundreds to thousands of dollars. So buyer beware.