Mod+ Ai generated content in RPGs

[jamie_graham]

I'm a long time lurker. Not posted before and I post on forums extremely infrequently even when I do. Too much going on in the real life unfortunately.

A bit of background: I work as a Graphic Designer, Web Designer/Developer. I'm also an artist when I have free time. I've also been building my own RPG. I've got to the point where I am considering the art/design aspects of it. Not sure if I'll sell if or give it away at this point.

I'm interested in your thoughts on AI Art specifically but on the AI onslaught on Creativity in general hence why I am breaking radio silence and particiapating. As you can see my background gives me a vested interest in the future AI Art in general due to my background and now I'm pondering the design and art for my own possible product even more of an interest.

AI generated Art/Illustration and Design

So my feeling is this cat is well and truly out of the bag! The genie cannot be put back in the bottle. AI Art and design content is here whether we like it or not.

Sites like: https://www.kittl.com/ are already eating into the lower end of the design market. The big 5 publishers already use AI Art. The big stock photo companies are already including AI art in their collections. I'm pretty sure it's already being used in the wider design industry within workflows to cuts donw time on layouts for example and in visuals if not in the actual artwork for print.

It's only a matter of time before AI Art cannot be distinguished from art created by a human (in fact it may already have crossed that line reading some of the latest on Ars Technica the technology website this morning).

I think AI will have four major effects in the design/illustration sectors (if it's not already!):

1. Products that could not previously afford art for lack of a budget will be able to afford it. This is probably a positive in that art enriches our experience whatever you might personally think about the use of AI art. This is an area where no/fewer artist are being hurt as there would not have been art in the product anyway. This covers most small indie RPG products which usually do not have the budget to extend to commissioned art.

2. The lower end of graphic design will be covered by products like Kittl. These use templates and 'guided' AI designs to create reasonable usable designs for posters, t-shirts, logos etc. I can see this expanding to business cards, stationery, branding even brochures in time.

3. The mid to upper tier will still need a creative eye. Designers/Illustrators/Artists may use AI as a tool, but it will be their creative direction and eye that guides that tool.

4. Weirdly I can see the premium for verified non-AI art/illustration/design by a verified human may actually go up. It's interesting seeing a backlash against the use of AI art. There are RPG products that have used AI generated art on Drivethrurpg and I've read comments where people have stated up front they are not buying that product because of that use of AI generated art. I think that backlash could put a premium on art that is verifiably hand made by artists/illustrators especially in traditional media, but not exclusively. The premium end of the market may go this way, but I suspect the allure of AI will mean its use will eventually become pervasisve and all encompassing.


The combination of the above will probably put Designers/Artists in all sectors at the lower end of the market out of business or they will have to embrace AI as part of their workflow... Kettl for instance seems to be aiming itself at designers... There will likely be fewer designers/artists/illustrators as a consequence. Anecdotally some of my illsutrator colleagues who have a simpler or vector styles are already saying their workload has dropped precipitously in the last six months, they are worried! There may not be a connection, it could just reflecting the economy (in the UK specifically although many of them work internationally) in general. But it is worrying all the same.

I'm not sure where I will fit into all this as a Designer/Artist. I'm guessing at some point I will probably be forced to adopt AI as part of my workflow or become uncompetitive. I don't like that, I've spent decades honing my skills but I think there is probably no avoiding it. I guess I'll just need to look at it as another 'tool' like Photoshop or Procreate... and probably sooner than I thought!

What are your thoughts on AI generate Design/Art/Illustration. Does the above sound plausible/probable?


AI Generated text content – This is a subject that has largely gone under the radar in most RPG forums, at least I've not seen a great deal of discussion on it. I'm not sure whether that means it's more acceptable . I can pretty much guarantee that a huge proportion of RPG text content has had AI input on it.

Almost every Copywriter I know uses AI to either polish what they have already written or to generate seeds from which they then generate a finished piece of copy. I am aware of Authors who are already using it to polish their writing. Use of AI in text content is difficult to near impossible to detect. I'm pretty certain authors are using it for RPG products right now and in already published material. Most of the big 5 book publishers already have their own AI model or are using an existing AI text generation model and I cannot see the big RPG publishers not following them in the long run. The big 5 publishers have been using AI for some time now as the tech has got progressively better. I guess most people see this as acceptable or at least something they cannot realistically protest against, but I would be interested in your views.

As an addendum to the above, so far I have resisted using AI to polish my writing!



And now for my own predicament!!

I have been tinkering with my own illustrations for my RPG. While a little rusty I can do them. However as hinted above I am time strapped. So this is one of the areas where I seriously started thinking about the potential of AI generated Art to shortcut my process. To rapidly iterate a compositions for example rather than sketching them out.

What are the ethics of using AI to generate composition that I then work up into finished art of my own? Whats the threshold on it being acceptable 10/90 in favour of the AI generated with only 10% of my input polishing it? Or should it be the other way nearer 90% my input and 10% AI? Is there a threshold under which you might find it acceptable at all?

Do I declare that my art is partially AI generated or inspired by an AI generated image?

Are there other issues I am not thinking of in this scenario?

 

[Exploderwizard]

AI art can be technically very good. However, AI cannot draw, and AI cannot paint. I think art pieces produced in a physical medium have more soul than computer produced art whether by a person or AI.

 

[Urdlen's best boy]

Honestly my thoughts about AI art :


AI art makes me nauseous to look at. Garbage at composition and lights. Soulless. The more you put work in polishing it the less disgusting, but I really don't see the point in having it at all.

As for the ethics of it, what platform are you using? Go look how they harvest images and whether you agree with them that's legitimate. Also look if it's compatible with the platforms you want to sell on. What I think? None of it passes the sniff test but you do you. As for whether you should mention it or not, if you want to be transparent yes.

My advice? Get a good model photos subscription. It doesn't cost that much and there are some for all kinds of things and you get iterations with actual lighting and all. Good models, a crayon and paper or photoshop are all you need to sketch fast.

 

[Bunch]

I'm using AI art as a image board for my scifi heartbreaker. As it will likely never see the light of day it's probably moot but baring massive success as a Kickstarter it's probably what I'd use if I published it.

 

[Tommy Brownell]

Is AI the reason WotC’s output is now purely bland garbage?

 

[jamie_graham]

AI art can be technically very good. However, AI cannot draw, and AI cannot paint. I think art pieces produced in a physical medium have more soul than computer produced art whether by a person or AI.

Unfortunately the models are advancing so fast that they can create artworks that look like they are traditional media that are indistinguishable from art created by humans, with composition that is pretty expert. This will only get better.

There was even an AI model linked up to a robotic arm that painted traditional media paintings as part of an Art installation! I don't think that will catch on it was weird and creepy!!

I'm not in favour of it by any means but it is realty and whether I like it or not creatives like me are going to have to live with it one way or the other...

There was a time I would have wholeheartedly agreed with you on the digital art created by humans statement, but now I use Photoshop, Procreate and Artrage as part of my process I realise these are tools or media with different weaknesses and strengths to traditional. It still needs a creative to competently make marks within it. AI however is another level.

I've lived through enormous changes. I went through the transition from paste up artwork to digital artwork. That transition made a huge amount of people redundant and jobless with skills that were essentially almost useless. Transparenceny retouchers, Typesetters with machines that were so expensive you had to take a mortgage out on them. Whole sectors of the creative industry went out the window almost overnight.

AI will do the same to my profession yet again.

In terms of RPGs I'm interested in the discussion about where the line is between acceptable and unacceptable. I'm interested in seeing what the RPG market will accept in the long run. I get that its probably going to be fluid, as to what a majority of people will accept, but inthe long run there is no stopping it.

 

[jamie_graham]

Is AI the reason WotC’s output is now purely bland garbage?

Have they used a large amount of AI artwork? I think at least one artist has been identified that uses it in his workflow for WOTC but I'm not aware of others. I suspect the bland nature of their work has more systemic causes at WOTC, though use of AI could be a factor...

 

[Tommy Brownell]

Have they used a large amount of AI artwork? I think at least one artist has been identified that uses it in his workflow for WOTC but I'm not aware of others. I suspect the bland nature of their work has more systemic causes at WOTC, though use of AI could be a factor...

No idea. You said the “big 5 publishers” and I assumed that was RPG publishers.

 

[jamie_graham]

I'm using AI art as a image board for my scifi heartbreaker. As it will likely never see the light of day it's probably moot but baring massive success as a Kickstarter it's probably what I'd use if I published it.

Personal use seems like a better usage scenario.

Yeah this is one of the issues I struggle with. On the one hand AI is hugely democratising, in that now the bar is lowered for almost anyone to have art with their product. Which feels like a good thing in some ways.

On the other hand many of these models have been trained on artists work without their express permission or an extension of permissions that did not even mention AI. And anecdotally some illustrator friends of mine are already feeling the pinch if their drop in work is caused by AI...

 

[Urdlen's best boy]

Is AI the reason WotC’s output is now purely bland garbage?

Probably not but it sure helps them make it even worse.

Unfortunately the models are advancing so fast that they can create artworks that look like they are traditional media that are indistinguishable from art created by humans, with composition that is pretty expert. This will only get better.

That's just not true.

 

[Silverlion]

AI creations are not 'art' they may be images (for visual creations) but aren't actually art. There is no creativity there at all. No artistic vision, style, or any of the other key things that make it art.

Don't get me wrong. I love using AI imaging for prototyping visual ideas and things like character looks, but I won't publish anything using those images. Rather I'll find a living creator to use several images and notes to create their own piece of art.

I in the past referred to it as art, but it really isn't, and someone I was speaking with used images so I've adopted that.

I see a HUGE amount of pushback from almost anyone with a creative bone in their body on using it for game products. I realize profit margins are razor thin without AI, but I also realize that it isn't worth it to ostracise some of the best and most creative people in our field. Oh, there are people already using it for images and maybe for writing, but I run into the fact that at best it is a very very clever regurgitation of ideas often using the "by the numbers" approach. This means it will produce at best 'average' results as in producing by the most common factors. This means no innovation in mechanics, no inventive use of language, and no truly unique imagery. I know this having produced thousands of images with AI*.

Of course that won't stop people from using it just like most of the models are trained mathematically on stolen art doesn't stop people from using it.

Does that mean it won't ever be usable? No, someday it may be truly able to work as a tool for someone to create truly amazing new things, using bits it can create but making something wholly new with it, rather than copying and pasting as/is.



*I'd like to note that I often use it to generate details based on my own sketches and artwork so that at least it has my input and is not entirely based on stolen intelligence. It should also be noted that due to using it for my own stuff, I can identify a huge amount of AI visual works at a glance.

 

[jamie_graham]

No idea. You said the “big 5 publishers” and I assumed that was RPG publishers.

Oh sorry that's not what I meant. My partner works in mainstream publsihing I'm guilty of using her jargon!

The big 5 are the mainstream book publishers: Penguin Random House, Hechette Livre, HarperCollins, Macmillan, scholastic etc.

Where they go more niche sectors like RPG publishing will probably follow.

It wouldn't surprise me though if WOTC aren't already using it even if its just internally. They are the only RPG publisher that comes anywhere nears the level of The Big Five...

 

[jamie_graham]

That's just not true.

Hate to break it to you but DALL-E 3 is already there. I hate that fact believe me, but it is what it is.

In terms of composition. I would argue that like any tool, composition is probably its greatest weakness at the mo. But the models are learning exponentially. And you can get the basics of composition or at least a composition that will pass muster simply by following the maths of the Golden Rule/ Fibonaci Sequence. My guess is a few more generations and it will be indistinguishable from human generated art.

It's already famously fooled the art/photography world several times even with the older models and a little bit of judicious human intervention.

 

[jamie_graham]

Also just like to say that there is AI art and then there is AI art.

AI art generated using prompts written by someone with no arts/creative background is probably going to result in competent looking garbage most of the time. Any successes are probably more accidental than deliberate.

AI art generated using prompts from a creative with years of training, experience using the AI model and iterated upon with further prompts to fine tune are the works that will be more indistinguishable from human generated art.

So yes a lot of AI generated art looks pretty rubbish although kind of competently executed. But some is amazing... again unfortunately.

 

[Silverlion]

Hate to break it to you but DALL-E 3 is already there. I hate that fact believe me, but it is what it is.

In terms of composition. I would argue that like any tool, composition is probably its greatest weakness at the mo. But the models are learning exponentially. And you can get the basics of composition or at least a composition that will pass muster simply by following the maths of the Golden Rule/ Fibonaci Sequence. My guess is a few more generations and it will be indistinguishable from human generated art.

It's already famously fooled the art/photography world several times even with the older models and a little bit of judicious human intervention.

I can identify a lot of what Dall-E 3 produces, I've not seen everything of course but I can. It's also worth noting that even Dall-E 3 has trouble with things like medieval weapons (still) making straight blades, sane-sized maces, and the like. Sure that will get better eventually but it is far from there yet.

 

[Shipyard Locked]

That transition made a huge amount of people redundant and jobless with skills that were essentially almost useless. Transparenceny retouchers, Typesetters with machines that were so expensive you had to take a mortgage out on them. Whole sectors of the creative industry went out the window almost overnight.

Out of curiosity, what happened to these people made redundant and jobless, from your observation? What did they transition into?

 

[Urdlen's best boy]

Hate to break it to you but DALL-E 3 is already there. I hate that fact believe me, but it is what it is.

In terms of composition. I would argue that like any tool, composition is probably its greatest weakness at the mo. But the models are learning exponentially. And you can get the basics of composition or at least a composition that will pass muster simply by following the maths of the Golden Rule/ Fibonaci Sequence. My guess is a few more generations and it will be indistinguishable from human generated art.

It's already famously fooled the art/photography world several times even with the older models and a little bit of judicious human intervention.

It's still not anywhere as good as an artist or anywhere near passing for arts made by humans and I doubt it ever will. Humans don't look at a series to pixels to determine what the next one will be. It's a fundamentally different process, and in art, process shows.

There's been people claiming "AI is as good as humans now" and every time I checked their exemples it's been evident they had a poor grasp of not just composition but lighting, anatomy, shading, artistic themes and style consistency. Further the machine process is always evident. I'm open to being wrong but I'll require an actual exemple of a purely AI pic that's not obviously machine and at this point I don't think I'll ever see one because it's not possible.

 

[Urdlen's best boy]

That's a painting by the master Pisanello. You can see the process, the intent, in color and in brush, of every little detail, the care behind the composition, the thematics and story behind the image. The physical technique, the artistic purpose. AI will never be able to do that and that's why all it produces is dead.

 

[SavAce]

Unfortunately the models are advancing so fast that they can create artworks that look like they are traditional media that are indistinguishable from art created by humans, with composition that is pretty expert. This will only get better.

There was even an AI model linked up to a robotic arm that painted traditional media paintings as part of an Art installation! I don't think that will catch on it was weird and creepy!!

I'm not in favour of it by any means but it is realty and whether I like it or not creatives like me are going to have to live with it one way or the other...

There was a time I would have wholeheartedly agreed with you on the digital art created by humans statement, but now I use Photoshop, Procreate and Artrage as part of my process I realise these are tools or media with different weaknesses and strengths to traditional. It still needs a creative to competently make marks within it. AI however is another level.

I've lived through enormous changes. I went through the transition from paste up artwork to digital artwork. That transition made a huge amount of people redundant and jobless with skills that were essentially almost useless. Transparenceny retouchers, Typesetters with machines that were so expensive you had to take a mortgage out on them. Whole sectors of the creative industry went out the window almost overnight.

AI will do the same to my profession yet again.

In terms of RPGs I'm interested in the discussion about where the line is between acceptable and unacceptable. I'm interested in seeing what the RPG market will accept in the long run. I get that its probably going to be fluid, as to what a majority of people will accept, but inthe long run there is no stopping it.

I think that in the long run, the RPG market will accept it all, much like the general public will choose to purchase particle board flat pack bookcases at Walmart over a bookcase made by the Amish craftsman down the road, or is OK with auto-tuned vocals, even if a great singer can pull out a more emotive performance than what auto-tuned vox can, etc.

A lot of the push-back I think is in this sense of "I'm a human and that means something dammit!", like in the way the heliocentric model of the solar system threatened a "we're the center of the universe" vision the geocentric model supported, or how some people's beliefs about the nature of man was threatened by the theory of evolution. Chess players had to give it up a long time ago now. People are going to "I'm a professional wine taster", "I'm an audiophile who can tell this system doesn't use the best speaker wires" kinds of woo about ineffable senses of warmth or soulfulness. A feeling like "this is where our human soul lives, and they are trying to deny it and take it away."

Taking AI visual art in particular, I've seen a lot that is currently as passable to casual regular folk as a lot of boring artists. It is certainly better than some people who released products using Poser 3d model art back around the days of the d20 boom (or those supernatural romance novels where you get ravished by a Yeti or whatever). No human artists felt particularly threatened by that art, so there was less upset and more just mockery. Also, there is "bad" art that has a certain human appeal to it, like the earliest RPGs, it's almost like a 'zine punk rock appeal. DIY, getting it done. AI art isn't currently where it is like a great artist with a strong aesthetic and vision for a particular project. A lot of humans don't have a strong aesthetic or vision for their own products as it is, though. When it comes to lighting and anatomy, etc., I'm sure it will continue to improve. And it's not like Rob Liefeld always had a perfect sense of lighting and anatomy, yet he was able to sell millions of dollars worth of comics. Basically, an AI not being able to match the lofty heights of the art masters will not be a barrier to it being useful to folks who never aspired to those lofty heights themselves.

Take someone like Kevin Crawford, or I imagine, many RPG creators. They have an RPG, they have some ideas for the art they want to fill a number of slots in their layout. A number of human artists are recruited to fill those art slots. Is there a strong aesthetic being asked for, to create some kind of memorable unique art? I think that a lot of the time the art is serving a more utilitarian purpose. It's getting the job done, but isn't some grand achievement of the human spirit.

So... I think there will always be a place for the fully hand crafted things, but AI tools will continue to improve and the end state of all of this is that fully hand crafted things will end up niche, and the most mass appeal things will be made with generous helpings of AI assistance.

 

[Stumpydave]

I love the thought that computers and computer programming has advanced to the point that AI art is where it is.
I'm less than happy that the models have been trained on other people's art without acknowlegement or payment.
I'm also less than happy that creatives are seeing their work opportunities being pulled away from them by a corpo-rat with a subscription to the Painto-matic 5000.
Its the same 'less than happy' that has me giving the film studios side-eye for wanting to scan their extras and background artists and use that image in perpetuity.

See AI isn't really the problem, its how its being used.
You might love AI 'art', you might hate it, but chances are it will continue to improve until it is indiscernible from human produced work.

 

[jamie_graham]

AI creations are not 'art' they may be images (for visual creations) but aren't actually art. There is no creativity there at all. No artistic vision, style, or any of the other key things that make it art.

Don't get me wrong. I love using AI imaging for prototyping visual ideas and things like character looks, but I won't publish anything using those images. Rather I'll find a living creator to use several images and notes to create their own piece of art.

I in the past referred to it as art, but it really isn't, and someone I was speaking with used images so I've adopted that.

I see a HUGE amount of pushback from almost anyone with a creative bone in their body on using it for game products. I realize profit margins are razor thin without AI, but I also realize that it isn't worth it to ostracise some of the best and most creative people in our field. Oh, there are people already using it for images and maybe for writing, but I run into the fact that at best it is a very very clever regurgitation of ideas often using the "by the numbers" approach. This means it will produce at best 'average' results as in producing by the most common factors. This means no innovation in mechanics, no inventive use of language, and no truly unique imagery. I know this having produced thousands of images with AI*.

Of course that won't stop people from using it just like most of the models are trained mathematically on stolen art doesn't stop people from using it.

Does that mean it won't ever be usable? No, someday it may be truly able to work as a tool for someone to create truly amazing new things, using bits it can create but making something wholly new with it, rather than copying and pasting as/is.



*I'd like to note that I often use it to generate details based on my own sketches and artwork so that at least it has my input and is not entirely based on stolen intelligence. It should also be noted that due to using it for my own stuff, I can identify a huge amount of AI visual works at a glance.

Hmmm not sure I agree on that. Art is subjective and the value we ascribe to art really comes from the viewer rather than the creator. If the viewer does not know there is an actual live human painter or author of that work, but appreciates it just as much as they did a similar human generate peice, is it art, is it not art?

I don't know the answer to that.

It's like the thought game of virtual reality. If we are living in a virtual reality indistinguishable from realty is it of any less value than actual reality. If it's indistinguishable from reality then for all intents and purposes it is reality. Or is it?

There is an argument that there are certain natural laws to a pleasing meaningful piece of art. I've already mentioned the fibonacci sequence which is encoded in the structures of almost every living thing on the planet. It also happens to be encoded into many paintings and compositions via the Golden Rule because it is pleasing to the eye. We have a theory of colour which is also pretty much hard encoded into our eyesight.

I'd like to think there is something unique to human creativity but I suspect that there are certain patterns and formulas that underly it which may not be unique.

 

[Mosker]

AI creations are not 'art' they may be images (for visual creations) but aren't actually art.

• Indeed: I've "trained" myself to refer to it as AI imaging.
  
  
• AI art--best excuse not to pull the trigger on a purchase without having to go through the creator and/or publisher's social media posts and have to ask yourself, Do I really want to give them my money?
  
  
• Full disclosure: I have bought one big book for a brand close to my heart, from a small press whose creator claims (and I believe them) that others (who would want to remain unnamed) did get paid for polishing the images. However, the use of AI art is a primary reason why I am not creating for said property be it for fanzines, or serious standalone products.

 

[Baron]

I don't condone the use of AI in art or writing. (There's a blog I looked at recently and I swear the guy is just generating the posts with an AI.) "Training" the software on the work of actual creators without permission or recompense (and how do you recompense someone for taking their style and cutting into their livelihood anyway) means I will refuse to purchase anything from those perpetrators. If you choose to use this software then there, you have my answer. I hope you'll at least be transparent. I understand Drivethru requires that transparency. And WOTC had something of an AI art scandal recently. The quote I copied was:
Hasbro/WOTC causes an uproar when their latest book, Glory of the Giants, features A.I. artwork. "An embarrassed representative for Wizards of the Coast then tweeted out an announcement about new guidelines stating explicitly that ""artists must refrain from using A.I. art generation as part of their creation process for developing D&D art."" – Slashdot

 

[Exploderwizard]

A couple of folks in my gaming group used AI to generate a brief character character story. They told me they were doing it, and even if they had not I spotted it right away. Devoid of soul or character, the writing was bland and tasteless like oatmeal made of shredded cardboard.

 

[Art's Puzzlebox]

The combination of the above will probably put Designers/Artists in all sectors at the lower end of the market out of business or they will have to embrace AI as part of their workflow...

1. Working Artists are already losing work, including myself and several other freelance artists I know. The damage has already been done.

2. Painting the situation like it is inevitable (you "have to" embrace it) is kind of short-sighted. At this moment, ai companies are facing a variety of lawsuits from authors, artists, actors, and companies (like stock photo companies) for the misuse of copyrighted data. AI companies are seeing resignations of top AI evangelists who are leaving because they don't agree with the current methods of data scraping. The technology is not regulated at all and laws will take time to catch up to the reality of the situation. Ai companies are now complaining that if they actually had to do things by the book, and compensate artists like they should have at the start, they would go broke (which is false, given that ai companies are potentially being valued in the billions).

If the courts decide that current methods of data scraping are not kosher in the future, all of those shiny ai -gen images you "made" by typing prompts may be considered completely radioactive at some point in the near future, leaving you exposed to potential copyright infringement.

3. AI generated images are not valid for copyright. This is the biggest issue Pro-AI users try to ignore, but it is the biggest issue for users in the future when making a commercial product. Do you want to own the art you use in your project and protect your IP? If so, AI-gen images aren't kosher for your project. Do you mind if anyone can take your ai-gen images and reuse them, even in a competing product? By all rights a larger company could take your work and repackage it as their own without any recourse (like what Wotc wanted to do during the 1.1 OGL debacle).

4. Artists & their supporters are currently developing the tools to defend their work from companies and individuals who seek to misuse ai-generators and training models. Tools like Nightshade can effectively poison training models, permanently destroying their ability to make sense of scraped images. If those who make and use ai tools continue to refuse to treat artists with respect, and refuse to give proper credit and compensation to the artists from whose work they are derived, then those artists will have no choice but to defend their livelihoods & their work.

So my feeling is this cat is well and truly out of the bag! The genie cannot be put back in the bottle. AI Art and design content is here whether we like it or not.

AI-tools require human inputs on a massive scale. If artists poison their own work to protect themselves, there won't be any new training models to create, and that will be the end of ai-generated "art" as we know it.

These companies had the option to use data existing solely in the public domain, but they chose greed, plain and simple. This decision can eventually come back to bite them, and anyone who uses these tools in the future.

Be advised that you are taking a risk that you don't have to. There is already a massive amount of art available in the public domain, all of it for free. Several successful games use public domain and stock art and do just fine (Mork Bork and Into the Odd are good examples of games that use public domain and stock art).

Not to mention, you can find affordable artists that don't charge an arm and a leg if you are willing to look for them. Painting the situation as inevitable is basically like saying you have no other alternatives available to you, when your options have actually never been so diverse.

 

[Urdlen's best boy]

Yeah it's not the same as any technological advancement, unless theft is "just an economic innovation".

And I maintain, it won't even stop to be visible. Just hearing the "pro" side, I know they've got nothing. Not directed against OP personaly but "moar calculation power" and "fibonacci sequence" are complete memes and besides the point.

Artists don't actually use the "golden ratio" or whatever. It's just a pop science meme. People who use that as a selling point are just looking to sell subscriptions on a pipe dream - the idea that art is fundamentally solvable by math. It's not, and people who say otherwise have no clue what they're talking about.

People will do what people do tho. It's enough for to sell slop to the average philistine and requires no effort. I guess it is what it is. Incredible that not so long ago D&D was


and now it's "perfectly good art just as if done by a real human"



Thanks technology! (edited due to 2nd image bugging)

 

[Lychee of the Exchequer]

I don't know if AI images (I won't call them art) are gonna be widely used.

If they are, and if the book buyers adopt it en masse, I think it could pervert our tastes and aesthetic sense as a society, until we feel a little bit more isolated and lonely in this end of capitalism society than we already are. You can't feed machine-things to people in search of meaning and beauty without deadening their souls.

This is for the image part. For the text part, I'm not too worried: unadulterated AI text is garbage. If some writer is willing to use it to feed his prose, he will suffer from gradual loss of writing and creative skills. And that won't sell well : you can't always fool or interest people with bland and inept prose, which warmed over AI text outputs tend to produce.

AI doesn't create anything. It just produces simulacras which may fool people into thinking they are the product of human ingenuity. The danger of AI isn't Skynet - the danger is the human mind smothered by mass-produced pseudo-thinking: day of the never-having-been-alive zombie bots.

But I'm an optimist: RPG aficionados will continue to entertain themselves playing with real human friends and love-generated RPGs hobby games, and D&D One VTT-enhanced bollocks will be an abysmal failure .

One thing to warm my heart: I'm amazed by the creativity currently displayed by the RPG scene compared, say, to the Hollywod crapazoïd movie industry. Human quirkiness, creativity and talent will survive the corporatist structure.

 

[TristramEvans]

Welcome to The Pub J jamie_graham

Unfortunately do to the controversial nature of AI art in online discussions at the moment we've had a longstanding rule that any discussion of AI art is confined to a single ongoing thread: https://www.rpgpub.com/threads/a-i-art.8317/ (as described in the Beginner's Guide to The Pub: https://www.rpgpub.com/threads/beginners-guide-to-the-rpgpub.8516/)

However, since this thread has a number of active replies what I've decided to do is lock the original ongoing thread and this will now be the continuing thread. As such we will be changing the designation to Mod+, meaning simply that there will be stricter Moderating of personal or group attacks.

 

[jamie_graham]

Out of curiosity, what happened to these people made redundant and jobless, from your observation? What did they transition into?

Good question. I'm not entirely certain. Some retrained in other sectors. Some managed to transition over to Artwork Departements. Over time however even those disappeared. Most Design studios that I worked in from the early 2000 onwards the Designer was responsible for doing all the Artwork

There were also designers as well who couldn't transition. I started on a drawing board as a Junior Designer saw where things were going so went to Art College so I could better learn my craft and learn how to use the Mac because I could see that no-one would be using a drawing board in five years time. One guy I knew in that studio gave it all up became a bus driver I think, another took early retirement.

1. Working Artists are already losing work, including myself and several other freelance artists I know. The damage has already been done.

2. Painting the situation like it is inevitable (you "have to" embrace it) is kind of short-sighted. At this moment, ai companies are facing a variety of lawsuits from authors, artists, actors, and companies (like stock photo companies) for the misuse of copyrighted data. AI companies are seeing resignations of top AI evangelists who are leaving because they don't agree with the current methods of data scraping. The technology is not regulated at all and laws will take time to catch up to the reality of the situation. Ai companies are now complaining that if they actually had to do things by the book, and compensate artists like they should have at the start, they would go broke (which is false, given that ai companies are potentially being valued in the billions).

If the courts decide that current methods of data scraping are not kosher in the future, all of those shiny ai -gen images you "made" by typing prompts may be considered completely radioactive at some point in the near future, leaving you exposed to potential copyright infringement.

3. AI generated images are not valid for copyright. This is the biggest issue Pro-AI users try to ignore, but it is the biggest issue for users in the future when making a commercial product. Do you want to own the art you use in your project and protect your IP? If so, AI-gen images aren't kosher for your project. Do you mind if anyone can take your ai-gen images and reuse them, even in a competing product? By all rights a larger company could take your work and repackage it as their own without any recourse (like what Wotc wanted to do during the 1.1 OGL debacle).

4. Artists & their supporters are currently developing the tools to defend their work from companies and individuals who seek to misuse ai-generators and training models. Tools like Nightshade can effectively poison training models, permanently destroying their ability to make sense of scraped images. If those who make and use ai tools continue to refuse to treat artists with respect, and refuse to give proper credit and compensation to the artists from whose work they are derived, then those artists will have no choice but to defend their livelihoods & their work.



AI-tools require human inputs on a massive scale. If artists poison their own work to protect themselves, there won't be any new training models to create, and that will be the end of ai-generated "art" as we know it.

These companies had the option to use data existing solely in the public domain, but they chose greed, plain and simple. This decision can eventually come back to bite them, and anyone who uses these tools in the future.

Be advised that you are taking a risk that you don't have to. There is already a massive amount of art available in the public domain, all of it for free. Several successful games use public domain and stock art and do just fine (Mork Bork and Into the Odd are good examples of games that use public domain and stock art).

Not to mention, you can find affordable artists that don't charge an arm and a leg if you are willing to look for them. Painting the situation as inevitable is basically like saying you have no other alternatives available to you, when your options have actually never been so diverse.

Those are all valid points.

By the way I am a working designer/illustrator and I have friends directly affected by this who suspect they are already losing business. I'm not advocating for it, or personally happy with it. However I also don't share your optimism about the future. I think the momentum behind AI generated works is so massive, the money savings that comerecial entities could save is so enticing to those businesses, the lobbying so intensive it's almost inevitable in my view, as depressing as that might be.

In response to your points:

1. Yep absolutely. I see it with sites like kittl they are directly undermining the very business I am in and possibly in he long run my own artistic endeavours.

2. I don't wan t to embrace it. However I suspect eventually I will be forced to whether I like it or not. I am hoping for the best, preparing for the worst. Part of that preparation is discussing it and trying to work out how it could feasibly be used in a workflow with a more minimal impact or as a tool for the process rather than the end result.

3. IA artworks cannot be be placed under copyright under US jurisdiction, for the moment. In the UK the government are still deciding how to clarify the situation. In other jurisdictions it is different. The by defaul the copyright of AI works lies with the user of the AI software i.e. those who enter the prompts or it lies with the provider of the AI software depending upon the wording of their copyright laws. Or the copyright is undefined/uncertain. That can and possibly will change.

If theyr remain essentially public domain that is still not a problem for many businesses. There are many commercial entities that will be fine with that if it saves the considerably on their bottom line. It is already the case with Royalty Free stock photography that other epeopla can use an image you've used in your product. You mention after these points the use of Public Domain images in Mork Borg. All AI generated images are essentially Public Domain under US law. For many works a unique image that you or your client owns is not a priority.

4. This is a very valid point. There are some AI companies that are recognising the ethical problems with using IP without permission and are switching to Public domain and explicit permissions. I think these companies will find a legal or semi legal way around it eventually or they will lobby government successfully that is is fair use. I don't agree with that but unfortunately we live in a world where that is often the reality.

The problem I see, is that when big business gets involved when huge savings can be made by using something like this then you and I are just steamrollered. I hope that's not the case, and I am involved in trying to prevent it.

However to paraphrase Sean Lock, I suspect the attempts to prevent what is happening will be like 'turning up at an earthquake disaster zone with a dustpan and brush'.

I really hope I'm wrong. But I've lived through these kinds of paradigm shifts before and the tech has pretty much always won.

 

[jamie_graham]

A couple of folks in my gaming group used AI to generate a brief character character story. They told me they were doing it, and even if they had not I spotted it right away. Devoid of soul or character, the writing was bland and tasteless like oatmeal made of shredded cardboard.

Yeah I think any text made up whole cloth in AI is pretty much garbage without a huge amount of time invested in refining the prompts at least. Probably not even then!

How AI is actually used in a professional context for text content however is quite different though to what you describe above.

For example I have a copywriter friend who writes copy for a consulting business. She will generate about six or so 'seed' content for tweets for example on a specific subject using a number of prompts. She will run those by her client.

If the client chooses one or two he likes she will then work them up, rewriting them using her preferred grammar/tense for example and any jargon and house style that makes sense for the brand/audience.

By the time she has finished it's often quite different to the original 'seeds' she used the AI for. It's also undetectable as AI generated.

Or if its a longer piece she will write up the whole section of text herself, she will then feed that into the AI with a number of prompts that will 'improve' it. She will get back about 4-6 variations from which she will choose one, do a bit of further editing and send it on to the client for approval.

This is how AI is in practice used in a professional context for text content. She confimred to me that is how it is being used in journalism and mainstream publishing. You would be unable to detect that AI had any input in it at all. I'm pretty confident you will have read text content on a news website, brochure, tweet, blog post etc. that has had AI input and you would never have detected it.

She's a professional with 40+ years experience copywriting she wouldn't pass of garbage to a client. Using AI she estimates she has doubled her output and that's the hook for her. It makes her job easier.

I was surprised I had though she would be dead against AI. But she has embraced it as just another tool in her arsenal.

This is how it's used in mainstream (non RPG publishing) and has been for a couple of years now without the same uproar its caused in the art/illustration context. I am not sure why that is.

It almost seems like people are far more willing to see it as a tool, as a means to an end rather than an end in itself in the copywritng business.

 

[jamie_graham]

[information]
Welcome to The Pub J jamie_graham

Unfortunately do to the controversial nature of AI art in online discussions at the moment we've had a longstanding rule that any discussion of AI art is confined to a single ongoing thread: https://www.rpgpub.com/threads/a-i-art.8317/ (as described in the Beginner's Guide to The Pub: https://www.rpgpub.com/threads/beginners-guide-to-the-rpgpub.8516/)

However, since this thread has a number of active replies what I've decided to do is lock the original ongoing thread and this will now be the continuing thread. As such we will be changing the designation to Mod+, meaning simply that there will be stricter Moderating of personal or group attacks.
[/information]

Thanks and noted!

 

[Bunch]

Yeah I think any text made up whole cloth in AI is pretty much garbage without a huge amount of time invested in refining the prompts at least. Probably not even then!

How AI is actually used in a professional context for text content however is quite different though to what you describe above.

For example I have a copywriter friend who writes copy for a consulting business. She will generate about six or so 'seed' content for tweets for example on a specific subject using a number of prompts. She will run those by her client.

If the client chooses one or two he likes she will then work them up, rewriting them using her preferred grammar/tense for example and any jargon and house style that makes sense for the brand/audience.

By the time she has finished it's often quite different to the original 'seeds' she used the AI for. It's also undetectable as AI generated.

Or if its a longer piece she will write up the whole section of text herself, she will then feed that into the AI with a number of prompts that will 'improve' it. She will get back about 4-6 variations from which she will choose one, do a bit of further editing and send it on to the client for approval.

This is how AI is in practice used in a professional context for text content. She confimred to me that is how it is being used in journalism and mainstream publishing. You would be unable to detect that AI had any input in it at all. I'm pretty confident you will have read text content on a news website, brochure, tweet, blog post etc. that has had AI input and you would never have detected it.

She's a professional with 40+ years experience copywriting she wouldn't pass of garbage to a client. Using AI she estimates she has doubled her output and that's the hook for her. It makes her job easier.

I was surprised I had though she would be dead against AI. But she has embraced it as just another tool in her arsenal.

This is how it's used in mainstream (non RPG publishing) and has been for a couple of years now without the same uproar its caused in the art/illustration context. I am not sure why that is.

It almost seems like people are far more willing to see it as a tool, as a means to an end rather than an end in itself in the copywritng business.

That's how a lawyer friend of mine uses it. He compared what it put out for a contract vs what he would do and it's about 95% good. So now he just spends 15mins reviewing it's output vs 45 mins writing it from a boilerplate. It's a force multiplier. Client loves it because they get functionally the same work for 1/3 the price.

 

[Stumpydave]

I use it for lesson planning - give the AI the learning spec and request it expand on the topic. I need to double check the output for accuracy but it's saved me hours in planning.

Its a tool - like a more effective search engine.

 

[Art's Puzzlebox]

By the way I am a working designer/illustrator and I have friends directly affected by this who suspect they are already losing business. I'm not advocating for it, or personally happy with it. However I also don't share your optimism about the future. I think the momentum behind AI generated works is so massive, the money savings that comerecial entities could save is so enticing to those businesses, the lobbying so intensive it's almost inevitable in my view, as depressing as that might be.

Well... I really don't know what will eventually happen, its all up in the air at the moment. This could potentially be a "Napster" moment, where its working as is now, but will eventually fall to the wayside for something different. I wish I knew but I don't.

2. I don't wan t to embrace it. However I suspect eventually I will be forced to whether I like it or not. I am hoping for the best, preparing for the worst. Part of that preparation is discussing it and trying to work out how it could feasibly be used in a workflow with a more minimal impact or as a tool for the process rather than the end result.

I can't really comment on this. This is a choice you have to make yourself as an artist. But you don't have to feel pressured into something just because its popular either.

The entire reason I became an artist was to pursue what aesthetic i wanted to create because I didn't see it out there. The last thing i wanted to do was to follow what was popular as that is the death of art itself. I know an artist in the UK who is hands down the best handstyler i have ever seen in any country, the best. Now they must make works of commercialized street art to make ends meet. Literally they deleted their entire volume of handstyles from the internet and only paint copy-paste mickey-mouse with three eyes over and over and over, and people love it.

If you really want commercial success, its out there, if you are willing to forego your own aesthetic, and your own identity as an artist. Those artists who do find success using their own style have done so through a lifetime of hard work and dedication, like you understand being an artist yourself.

All AI generated images are essentially Public Domain under US law.

This is still being contested. What is a fact is that in the US, only humans can register their works with the US copyright office. Public domain art is not the same, because you face no risks whatsoever for using it. Ai-generated images still pose a potential risk if the laws change in the future.

The problem I see, is that when big business gets involved when huge savings can be made by using something like this then you and I are just steamrollered. I hope that's not the case, and I am involved in trying to prevent it.

Several AI developers are saying that AI companies will have run out of data to train AI models on by 2026. That means human input ("data") is essential to its future survival. If all the artists get sick of the current situation, and start poisoning their artwork and opting out of training models, there wont be any new data to train from. Either way its dead the way I see it, if they don't start to treat artists of all kinds (writers, artists, actors, etc) fairly.

Not to mention that models are prone to collapse and decay when they attempt to train on ai-generated images. They need us more than we need them.

 

[Bunch]

Well... I really don't know what will eventually happen, its all up in the air at the moment. This could potentially be a "Napster" moment, where its working as is now, but will eventually fall to the wayside for something different. I wish I knew but I don't.

I can't really comment on this. This is a choice you have to make yourself as an artist. But you don't have to feel pressured into something just because its popular either.

The entire reason I became an artist was to pursue what aesthetic i wanted to create because I didn't see it out there. The last thing i wanted to do was to follow what was popular as that is the death of art itself. I know an artist in the UK who is hands down the best handstyler i have ever seen in any country, the best. Now they must make works of commercialized street art to make ends meet. Literally they deleted their entire volume of handstyles from the internet and only paint copy-paste mickey-mouse with three eyes over and over and over, and people love it.

If you really want commercial success, its out there, if you are willing to forego your own aesthetic, and your own identity as an artist. Those artists who do find success using their own style have done so through a lifetime of hard work and dedication, like you understand being an artist yourself.


This is still being contested. What is a fact is that in the US a machine (or an animal) cannot author works, only a human can, and only humans can register their works with the US copyright office. Public domain art is not the same, because you face no risks whatsoever for using it. Ai-generated images still pose a potential risk if the laws change in the future.


Several AI developers are saying that AI companies will have run out of data to train AI models on by 2026. That means human input ("data") is essential to its future survival. If all the artists get sick of the current situation, and start poisoning their artwork and opting out of training models, there wont be any new data to train from. Either way its dead the way I see it, if they don't start to treat artists of all kinds (writers, artists, actors, etc) fairly.

Not to mention that models are prone to collapse and decay when they attempt to train on ai-generated images. They need us more than we need them.

I'm not sure that's entirely true. A machine can run a lot more experimental to see what works than a human can.

I watched my brother go through this as a photographer. With a few exceptions the public doesn't really care about the details nearly as much as the artists and professionals do. Most just want a picture of X +/- a bit. An amateur or a machine could do what most wanted and people would be happy with it. They weren't buying the art really. They were buying a picture of their kid or a building or something else they were attached to and quality only had to pass a certain bar

 

[Art's Puzzlebox]

I'm not sure that's entirely true. A machine can run a lot more experimental to see what works than a human can.

I do agree that a machine can create outputs faster than humans. That they are more experimental, or better or worse is I think left up to personal preference as it is very difficult to argue what "experimental" even is in art, let alone define what a machine does when it regurgitates algorithms to achieve a similar result as "experimental" as its just RNG'ing a prompt to death.

I also posted several points in that post, so i think there may be something getting lost due to it being long-ish? This seems a rather short response is all (please forgive me if I misread).

I watched my brother go through this as a photographer. With a few exceptions the public doesn't really care about the details nearly as much as the artists and professionals do. Most just want a picture of X +/- a bit. An amateur or a machine could do what most wanted and people would be happy with it. They weren't buying the art really. They were buying a picture of their kid or a building or something else they were attached to and quality only had to pass a certain bar

Yeah I do agree that the public don't care about the details. But if "quality" as you say only has to reach a certain bar, I don't really know what you are applying it to. Do you mean that in every industry? TTRPG's are a very different thing than commercial photography and I don't know if I could compare one to the other.

The problem with using these analogies is that this technology is unprecedented, there is nothing that one can compare it to without losing context. Using the photography example doesn't apply because ai is a thousand times faster, and not limited by the physical boundaries of reality like photographers are (ai can produce a thousand city portraits in seconds, which could take a photographer a lifetime to photograph every city and town on earth in person).

 

[robb]

Several AI developers are saying that AI companies will have run out of data to train AI models on by 2026. That means human input ("data") is essential to its future survival.

I don't think that is accurate as we are seeing a rise of data augmentation and synthetic data creation techniques (such as adversarial generative networks) that are changing AI workflows so that they will consume more AI generated content to get more variety from the data we already have. In some ways this may be good as it moves the vast data requirements away from being only accessible to those big entities already controlling a platform eg Google etc.

However I totally agree that artists should be compensated for their work that has been used on training models and that it takes time for laws to catch up (but progress is being made on that).

 

[Art's Puzzlebox]

I don't think that is accurate as we are seeing a rise of data augmentation and synthetic data creation techniques (such as adversarial generative networks) that are changing AI workflows so that they will consume more AI generated content to get more variety from the data we already have. In some ways this may be good as it moves the vast data requirements away from being only accessible to those big entities already controlling a platform eg Google etc.

Hmm, well I guess you are looking at different data than I am, as what I see tells me that retraining on ai-generated data eventually results in a decay in performance, and potential collapse. If that's true then it is an issue in regards to the longevity of the tech. In the meantime, human input is still essential to its existence.

However I totally agree that artists should be compensated for their work that has been used on training models and that it takes time for laws to catch up (but progress is being made on that).

Yes indeed.

In the meantime, who knows what will happen.

I will mention Nightshade again, as it is an emerging technology that evolved from the current methods used for data scraping. I am hopeful that a middle ground can be found between companies and artists, as the end result of a Nightshade situation can only be scorched earth for ai programmers.

 

[Bunch]

I do agree that a machine can create outputs faster than humans. That they are more experimental, or better or worse is I think left up to personal preference as it is very difficult to argue what "experimental" even is in art, let alone define what a machine does when it regurgitates algorithms to achieve a similar result as "experimental" as its just RNG'ing a prompt to death.

I also posted several points in that post, so i think there may be something getting lost due to it being long-ish? This seems a rather short response is all (please forgive me if I misread).


Yeah I do agree that the public don't care about the details. But if "quality" as you say only has to reach a certain bar, I don't really know what you are applying it to. Do you mean that in every industry? TTRPG's are a very different thing than commercial photography and I don't know if I could compare one to the other.

The problem with using these analogies is that this technology is unprecedented, there is nothing that one can compare it to without losing context. Using the photography example doesn't apply because ai is a thousand times faster, and not limited by the physical boundaries of reality like photographers are (ai can produce a thousand city portraits in seconds, which could take a photographer a lifetime to photograph every city and town on earth in person).

Qualify to the bar of that industry and I suspect in almost all but publications aiming to professional artist the bar is far lower than professional artists would like to admit.

I'm of the opinion a) training on art is not theft as it is essentially what all people do to learn skills that exist. I looked at other folks code to learn how to program. I looked at other folks management styles to learn what works for me.
A computer doing the same isn't theft in my mind so we're not likely to agree here.

Mostly though having gone down this road with other professions disrupted by Internet or computer advancement the computers win. They do more work in less time with good enough results for most. I've heard most of the argument why this time it will be different but usually I hear from the say person a decade or two later saying they wish they didn't fight it and just adapted in X, Y or Z way that still let them do most of what they wanted. I'm jaded in this area. The car isn't perfect but I'm not going back to the horse and buggy. I could never draw worth crap but I can describe what I want and get what for me is art. That's awesome.

 

[Dammit Victor]

Ethically, I'm not entirely sure where I stand. Every human artist ever learned by imitating the works of other artists, and no matter how perfectly a human artist imitates another human artist, it doesn't become a crime unless the plagiarist attempts to pass off a whole copied work as their own, or the forger attempts to pass off a simulation or an imitation as the product of the imitated artist.

We're talking about copyrights and pretending they're patents.

Also, every technological advance (basically ever) has threatened someone's livelihood, and historically... the technology wins, and the artisans have to find other work. We usually call this 'progress', but I think it is obviously not a good thing when we're talking about the Humanities. Is it because the Humanities are special, or is it because I'm a writer who can also feel Skynet breathing down my neck?

Professionally, I use AI programs to improve my work. I use ChatGPT to generate ideas, and I use DALL-E to provide visual references for the human artist(s) I hire. I would never dream of copying ChatGPT's output directly into a production document, or using DALL-E's output-- raw, or human polished-- as an actual piece of art in something I publish. There are still quality issues to be taken into account, but beyond that... the human touch is something the customer is paying for. It's part of the purchase price, it's part of the agreement I have with the customer about the provenance of my wares. And I am always going to prefer the custom of people who care about that distinction to that of people who don't.