D&D General Would It Matter To You if D&D Books Were Illustrated by AI Instead of Humans?

[EzekielRaiden]

Aren't they? For all I know, they are created by "Wizards of the Coast". The internal process they use is unknown to me. I guess they have writers because... I am not that knowledgeable in AI and I don't know if it's possible.

At present, the absolute best AIs in the world, trained on literally gigabytes of raw text files, can manage about six or seven paragraphs before the text becomes completely unhinged.

I don't mean "ooh, you really notice the flaws." I mean that it starts talking straight-up bizarro weird stuff, like fourteen-horned unicorns, or repeatedly contradicting itself on what color something is, etc.

The problem is, the correlational structure upon which text-generation AI is built cannot handle the connections necessary for logical cohesion in a long-form work.* Combinatoric explosion takes over, and it scales much, much faster than technology can. We're currently still in the early days, where most of the limitations are in implementation, or training time, or learning the most effective ways of doing things. As it stands, writing even a single chapter of the kind of length expected of an RPG rulebook is completely impossible.

The irony, of course, is that it is much easier to fake images than it is to fake text. As the links others have given show, you can make completely realistic-looking (minus some eldritch horror side-faces and surrealist clothing/backgrounds) faces that are 100% fictitious. Making three pages of consistent, readable fiction? Completely impossible at present--and it's going to be hard to collect enough of a corpus to improve it much further. We've already used a scrape of a large portion of the internet. Where are we going to get the data to train GPT-4 on?

*Really, the problem is that "generative" text AIs, like GPT, are trying to substitute really, really, really complex correlation networks in place of actual knowledge. That is, they are trying to use a really, really, really complex model of English syntax--a model of "word X should come next, then word Y, then word Z, and NOT word B"--in order to successfully produce semantic content in English--that is, things which contain meaning to English speakers. That is, formally speaking, impossible. You cannot magic up semantic content from exclusively syntactic content. You can get some surprising, even shocking connections, you can get some truly impressive relationships. But you cannot derive the semantic from the syntactic, even in the limit of an infinitely complex syntactic model.

 

[Benjamin Olson]

That said, there are certainly huge ethical question marks hanging over the issue. Using generic or "self-orchestrated" AI art is one thing, but using AI art to mimic the style of a specific artist is something else entirely, like if someone created a Planescape-y setting using AI art that looks deTerlizzi-esque. That would be a definite no-go for me.

This was the most diTerlizzi-esque thing I could manage with my afternoon's experience with MidJourney:

I don't know if it filters out requests for the style of specific living artists, or if it just couldn't find enough of his art to get the style. But since I feel his planescape art has obvious Arthur Rackham influence I specified that instead, and just requested the most DiTerlizzi of all subjects, an elf girl relaxing in skimpy armor.

But fantasy artists aren't going out of business anytime soon. AI art can, after a few tries manage art of human (and human adjacent) subjects in a fantasy setting, albeit usually with some complications about items carried by them which it really can't figure out the rules for (the objects this lady had in her lap make no sense). An artist skilled enough to do touch-ups and corrections could get a lot of mileage out of it for such things. BUT even the most basic of fantasy creatures are too varied in the representations it's searching through to get satisfying output (although I think the output is very useful for brainstorming or creating rough versions to be redone by a proper artist). It just can't figure out the rules for things that only exist in people's imaginations, which is a huge part of what is needed for D&D books.

Behold, the results of my attempt to use similar means as those I used above to create a picture of a dragon flying over a fantasy city (these were some of the better results):

 

[Levistus's_Leviathan]

I normally wouldn't care if D&D book art was generated by AI if it was of sufficient quality . . . but also acknowledge that our society causes some artists to rely on jobs offered by WotC and similar companies for their work, so unless that issue gets fixed, I would not want WotC to transition to AI-generated art.

 

[Smackpixi]

I think it matters, it’s nice to have something real these days. I think that’s part of the appeal of RPGs, being with Real people, even if not in person. I think real artists is part of the D&D brand as well, publishing a history of your art and then just taking all the beholders you’ve ever commissioned and tossing them into a blender and hitting refresh till you get a new one…might actually be fun for the Twitter feed, but human created art is deliberate, and I’m pretty sure most creators of things want to be deliberate and not just accidentally cool.

But if your illustrations are just happened upon stock images randomally generated isn’t much different. It’s A make do until I can afford what I want…or it’s toss away background that isn’t important.

Art can’t be produced by a computer. It is not genuine, art is both a color palette, image, technique, medium, etc…but also the emergence of lived experience, it’s not only the object, but the creation of the object and the person behind that creation. People matter, their lives matter, a machine does not even if the people behind its creation do. It’s creations can be cool, and in viewing them you can give them meaning, but in and of themselves the machines creations are meaningless.

I’ve been to London and seen the real Rosetta Stone, a pretty unremarkable object, that in it’s time was also pretty unremarkable, but is an object of incredible importance, and being in it’s presence you can both imagine ancient people reading it and being bored, and it’s rediscovery by archeologists and the relavation of realizing what it meant when they found it. Objects have power and being next to them matters. I was also last month in the Milwaukee Public Museum with my son and saw a replica, and it was still interesting, and his chance to see what his history buff interest had previously only heard of. It’s different, less magical, but still, this is what what opened up understanding of history looked like, real boring.

So we return to the subject, mass produced books will never be art objects capable of bringing you into the presence of an artist creating a thing. And really art made on a computer will always deny you the opportunity of an object that can bring you into the room at the time of it’s creation, the experience an object can give. But reproduced images, be they lithograph prints, pictures of paintings on the internet, posters, or mass produced books, or limited run books still have meaning when the sources are made by people. This Owlbear has matted fur, why is it matted, why is it more bear than owl, why is it snarling, and so on, someone chose all of these aspects, and you can sit by the fire paging through your Monster Manual, engaging both with the choices made in the stat block and the flavor the choices made in the illustration imply. Why is the fur matted, do owlbears not lick themselves, what do the other creatures illustrated by this artist tell me about their ideas on owlbears.

That is if the art is made by a human. If it’s made by an AI, it’s just a whatever illustration next to the stat block. Looks cool, nothing to worry about, just an approximation.

Of course, there’s the aspect of all this, can I, can we, tell the difference? Could I have the Rosetta experience in Milwaukee if I believed it real? Can I tell the difference between a real Vermeer and a fake? I cannot. Even people who dedicate their lives to knowing the difference can be fooled, so is a believed experience relating to AI art where no human choices were made but I believed I was interacting with them any different than if it was real. If I can’t tell the difference between a real experience and one that is not, does it matter?

I believe it does, not because I can or cannot tell the difference, but because we as a society should protect and treasure real experiences. People who dedicate there lives to telling the difference between real and fake Vermeers don’t do it cause they’re up their own asses about Vermeer, though they probably are, but they do it to protect the authenticity of being in the room with a Vermeer, which they think is worth a thing.

if someone can’t tell the difference does it matter for them on that day of the experience, it does not. Does it matter for our society and the protection of reality? Yes it does. So long as people have bodies and lives that begin and end and hopes and fears and dreams connected to them, a full on embrace of the meaninglessness of the virtual will be bad, dumb, stupid and hollow.

 

[doctorbadwolf]

Will it bother me? Depends on what they do with the AI.

Do they use it to dupe/ape another artist's style, to in effect get DiTerlizzi art without having to pay DiTerlizzi for it? Then yes, it bothers me.

Will they use it but still claim that increased costs are due to having more art? Then yes, it bothers me.

Will they use it in lieu of actual usable content, since clicking "refresh" on a website is way cheaper than paying writers, editors, etc.? Then yes, it bothers me.

Can I think of any reasons to use AI art aside from those listed above? No I can't, so yes, it bothers me.

Warhol would be all over AI art.

Anyone bemoaning AI art isn’t thinking about it as a tool for artists, a therapeutic tool, and an accessibility tool.

AI art can eventually potentially translate our feelings about ourselves and others into visuals in ways that our internalized perception makes difficult, creating something different from what a given artist would do by hand, and allow people to explore their own perceptions through art who wouldn’t be able to without this tool.

What people aren’t remembering about these programs is that an AI that makes art is a tool into which art is input, and different art is output. It is a tool that creates what you ask it for, based on previous work and parameters you set.

It’s an incredible tool for art that will make art more accessible, and will lead to art being made that wouldn’t have otherwise, by people who wouldn’t have become visual artists otherwise.

 

[doctorbadwolf]

I normally wouldn't care if D&D book art was generated by AI if it was of sufficient quality . . . but also acknowledge that our society causes some artists to rely on jobs offered by WotC and similar companies for their work, so unless that issue gets fixed, I would not want WotC to transition to AI-generated art.

This is the only negative I see in this new tool. People need to pay rent.

 

[Levistus's_Leviathan]

This is the only negative I see in this new tool. People need to pay rent.

Yep. And that's not the fault of the tool, it's the fault of how our society is designed.

 

[Charlaquin]

I'm seeing a lot of "the artist deserves" arguments, but who really deserves anything? It's subjective. Either everyone deserves to be able to make a living doing the thing they love, or no one does. I say give everyone a universal income and let artists make are because they love it, let AIs make product because we want it, and everyone somewhat wins.

I mean, yes, I agree, but until we have something like that in place, I would still be bothered by an RPG company using AI generated art in leiu of paying artists.

 

[Charlaquin]

Warhol would be all over AI art.

Anyone bemoaning AI art isn’t thinking about it as a tool for artists, a therapeutic tool, and an accessibility tool.

AI art can eventually potentially translate our feelings about ourselves and others into visuals in ways that our internalized perception makes difficult, creating something different from what a given artist would do by hand, and allow people to explore their own perceptions through art who wouldn’t be able to without this tool.

What people aren’t remembering about these programs is that an AI that makes art is a tool into which art is input, and different art is output. It is a tool that creates what you ask it for, based on previous work and parameters you set.

It’s an incredible tool for art that will make art more accessible, and will lead to art being made that wouldn’t have otherwise, by people who wouldn’t have become visual artists otherwise.

This is the only negative I see in this new tool. People need to pay rent.

I agree, but for now it’s a huge negative. In theory AI could be an incredible tool for artists. But in a world driven by the pursuit of profit, it’s inevitably going to end up being used to cut costs.

 

[Charlaquin]

Yep. And that's not the fault of the tool, it's the fault of how our society is designed.

It’s the same problem as automation in any industry. On paper, automation should improve the quality of life for everyone. In the society we actually live in though, it just creates greater inequality, which speaks to the deeply perverse incentive structures we’ve created.