Yes and no. I agree. But the new difference between artistic vision and the final product has become much more stochastic.
Yeah, I think I could agree with that.
I do believe creative people will always have a place in creative industries, but things will change, because of AI. It happened before, and it will happen again.
One of the main reasons impressionist and post-impressionist art evolved in the 19th century is because of the advent of easily acquired, inexpensive, vibrant pigments and the invention of tubed paints. No longer were artists limited by muted hues, and they could now go on the road with their transportable paints into nature, not solely limited to a studio.
They were also no longer constrained to create art on commission only by those who could afford to pay them—nobility, the church, etc. Art salons came into vogue (in Paris, to start), and they could now create for the common man. This allowed them to paint common folks and common things, in common places, for common people. They began to paint with emotion, not just painting nobles on commission doing noble things.
The same type of evolution occurred with music. There’s a reason nobody in, say, ancient Greece, or Rome ever played baroque, or classical, or romantic, or jazz, or polka…music. They had no concept of such things as polyphony back then, nor the instruments to play it.
Beethoven, for example, almost single-handedly ushered in the romantic era of music (no longer was music constrained to adhere to rigid ornamentation and church-approved standards). Beethoven broke through those barriers and composed with raw emotion, first and foremost with his 3rd sympony.
Audiences at first hated Beethoven’s 3rd, because they didn’t understand it. To them, it was pure chaos (like rock and roll was to our parents). But Ludwig didn’t give a shit, he soldiered on with his emotional compositions, and the audiences grew to love them. It was a watershed occurrence. The floodgates were open.
It takes something (s), or someone (s) to break the barriers in art (and science) to evolve to the next level. If you’re at level A, you can’t jump to level D without first going through levels B and C. There’s a time and a place for everything. Often there’s a lot of stagnation between levels until something big happens to usher in the next level. Beethoven was something big. So was Einstein, and…
AI is a big thing, too, in the world of art (and language, etc.). It’s a game-changer. It’s going to shake things up quite a bit for a while, but it’s here, and it’s not going away, like it or not. The creatives will survive, so long as they embrace AI as a tool to express their artistic visions.
It’s like baseball—anyone can play it, but few will ever play for the N.Y Yankees.
All the people hellbent against the very existence of AI
are going to have to start singing a different tune with it integrated into Photoshop. Demanding it not be used will be like demanding that people not use the clone brush. (I wonder how many will squawk “tech-bro, tech-bro!” at Adobe?)
except that the Yankees never caused millions of people to lose their job…
Artists will lose jobs, but not all of them. Until they develop means to have AI apply different media ( like oil paint) to canvases, I suppose it will just be artists who work with art printed on paper, or posted online whose jobs will be in jeopardy, at least for the foreseeable future.
But, the more adaptable artists will use AI merely as a tool to help make their art even better and more quickly produced. They can use the AI to create the reference shots from which they draw, paint, or whatever, the actual final product—like artists who take photos of scenes they want to paint.
And, it still takes artistic talent to compose a great reference shot (with proper lighting, perspectives, etc.). These artists will save time being able to compose their references quickly using AI, rather than sketching them out, or going out in the field to take photographs. AI will be a benefit to these artists.
So, AI will not decimate all artists, just those who can’t adapt. It may even result in producing more artists being able to compete in the art market. For example, an artist who is disabled and unable to go out and take proper reference photos in the world will now be able to do so in their homes on their computer. It levels the playing field for them.
It is regrettable that some artists will lose their jobs, but that’s what happens whenever there’s a paradigm shift in a particular field. I’m sure some classically trained artists lost jobs when impressionism came into vogue, and some classically trained musicians lost jobs when Romantic era music took over, and some Newtonian-taught physicists lost jobs when Einstein ushered in relativity and Max Planck ushered in quantum mechanics, and so on and so forth. The world changes, but those who can adapt prosper.
I’m not sure, but I suspect that the people who lost jobs due to the camera were not “artists” but draughtsmen.
Even without having a unique artistic vision, the skill of being able to render a realistic image e.g. of the proposed site of a new quarry or of a machine part for a catalogue, was a relatively rare and thus valuable one. Similarly, if you wanted an enticing image of your product to stick on a poster, you needed a commercial artist to produce one. Some of those would also be artist-artists creating their own works and doing the commercial work to fund the masterpieces, but others would be in it for the paycheck.
Of course, taking good photos is also skilled work, and requires technical grounding etc. so while draughting skills may have been obsolete, skilled work was not. But the point is that when we’re talking about art, we’re not necessarily talking about Art. I think it’s quite likely that “producing original cover art for publishers” is going to become a more rarified trade quite quickly. Sure, prestigious bestsellers will get a human artist, but the budget end of the market not so much.
Ironic you should say that since the cover mentioned above IS from a NYT best-selling series.
A few months ago, there was another “Publisher uses AI Art for cover” story with a sci-fi novel. A couple similarities jumped out at me between that story and the one a few posts up:
(a) The art is pretty generic. The recent story is a stylized wolf head, the previous one was a generic spacesuit guy walking towards a swirly portal.
(b) The publisher took them from stock image sites. The intent was never going to be “hire an artist for this”, they just combed image galleries until they found something that looked good and that image happened to be AI generated. No one was robbed of their chance to create a bespoke paperback cover.
(c) There WAS some graphic artist interaction involved in changing the colors on both, cropping, laying out the title, etc. It wasn’t exactly just Right-Click/Save and put some words on the front in MS Paint.
I’ll admit I haven’t looked at many scifi/fantasy covers since the mid-90s but, back in the day, the art usually had something to do with the content and the castle or spaceship or guy in armor would be representative of how they were described in the novel. AI generators have a lot more issue with that – a dwarf woman is easy, a dwarf woman with a black streak in her blonde hair and wearing armor made of braided leather and silver chains and with a blue tattoo over her right eye is basically impossible, even with inpainting. So I’d say that part of the problem is that either tastes have changed or standards have gone down or both, but MidJourney can kick out “Rose laying on a dagger” a hell of a lot easier than “Distinctive spaceship described on pages 35-37 which is the symbol of the series”. Again, it’s not as though some cigar-chomping publisher said “We need a wolf head! Get Williams on DALL-E now!”, they just sent someone to trawl stock image sites and didn’t care who/what made their wolf head picture. Most of the art department was already out of work by this point.
I mess around with MidJourney a lot and agree that it’s superior to basically any other AI out there right now for a lot of applications (I haven’t tried the new Adobe stuff). It’s probably at a draw for rendering photo-realistic people (in large part because there’s so many finely trained porn-centric SD models) but for most other things and for artistic looks, it’s in the front of the pack. However, it has a lot of holes in it once you try to get exactly what you’re looking for and it lacks inpainting which means you need to import results into Stable Diffusion then hope the inpainted bits match well enough to not ruin the effect.
Stable Diffusion often has less convincing results but has a shitload of tools available as plug-ins to where you can influence color, directly pose figures, hand draw things to get rendered, etc. This all requires a lot more skill and knowledge than just “Hurr hurr I typed ‘elf girl’ and got a book cover” so there’s probably a definite niche for artists who want to learn to use ControlNet and still have an eye for composition and color. Or to blend the two and hand-craft your very intentionally detailed space pirate but use AI to paint in some generic (but directed) lunar background to save time.
I think this is the critical point. Assume these publishesr aren’t idiots - they know at a stock image isn’t as good in terms of what you would call quality metrics - e.g. style, textual references etc. but it is good at being cheap. And they’re already trawling stock image sites! That tells you plenty about the appetite for spending on human artists.
Now with a little extra money for the time taken to do a prompt that is mildlly more sophisticated than “elf girl” you can get something that offers a lot more quality for little or negligible extra cost. How can you say no?
I have thought about this before in my plan to go a few years into the past and sell my AI art as my own creations. I see that there are printing technologies that lean in that direction.
https://www.erinhanson.com/blog?p=new3dtexturedreplicas
Now I need to add “find a way to buy these undoubtedly very expensive printers” to my agenda, alongside “find a way to travel into the past”. (I’ll also have to come up with a system for adding paint stroke depth maps to images.)
Here is one of the breathless “publisher gets caught trying to sneakily do something evil and wrong” articles on it:
Ironically enough, in this bit of angry editorializing, the Seqential Art creator thinks book covers are one of very few places AI art will be used.
I’m not sure when that day was. I’ve happened to have read extensive descriptions from a couple of authors on how cover art happens:
http://www.antipope.org/charlie/blog-static/2010/03/cmap-6-why-did-you-pick-such-a.html
From one of them:
The artists who design the covers tend not to read the books they’re working on because they don’t have time. They’re working on so many books in a year and their deadlines are so tight they barely have time to read the cover brief. On top of that sometimes the book they’re working on hasn’t been written yet. (Or, at least, not finished.)
One cover that comes immediately to mind for me is the original paperback cover for the US release of Hyperion:
Pretty cool–except the Shrike was supposed to have four arms. (Check out some of the covers from other countries–some got it right, some didn’t.)
Eh, back in the 80s and early 90s. But I’m just speaking from my own experience and memory and not willing to go to bat for it across the full spectrum of c.1987 Waldenbooks
The quote about artists not reading the books doesn’t mean a lot. They don’t need to read the whole book, they just need an accurate description of the scene/character they’re supposed to represent.
It’s all academic anyway since none of it really affects the main point: Publishers are currently using stock art for their covers, even for best-selling authors/series, which is either aided by or resulting in very simplistic book covers. It’s hard to argue that AI is threatening the industry of drawing bespoke paperback covers featuring a stylized pistol sitting on top of a manila folder or the back view of a woman facing a glowing triangle.
Coincidentally, I was getting some cool results last night using “Dystopian Spirograph art” as part of my prompts 
Publishers are currently using the same stock art.
Also, maybe if he used AI, Battle Bunnies wouldn’t have been abandoned three years ago. ZING!
Not really, particularly for first print stuff that didn’t have a big name attached to it. Science fiction books would often get a generic space ship, even if it wasn’t that kind of science fiction. Female characters would be hypersexualized, regardless of how they were written in the book. POC characters would regularly be portrayed as white. Publishers chose book covers based on what they thought would sell the book. They didn’t care if it matched the contents, because by the time the reader figured out the cover wasn’t representative of the contents, they’d have already bought the book.
Could be. I also might be over-weighing my memories of TSR Press stuff where they’d have Elmore/Caldwell/Easly/Parkinson do art that could get repurposed into module covers and stuff. Or Xanth novels. Or whatever else I’m remembering reading then referring to the book cover and thinking “Ah, so this is that moment…”
Anyway, as I said, irrelevant to the main point of “Oh no, an AI-generated book cover!” anyway.
ChatGTP is being sued for libel.
You can’t make that stuff up. Oh wait.
Why do I get the idea that this case was filed not on its merits, but rather for the lawyerly bragging rights?