Obviously, AI writing has come a long way in the past few years. Still, mostly a fancy auto-complete task with a bigger database to scrape, but that database allows it to put more and more things together in a more complete way than ever.

With that said, we have seen a rise in AI written books and scripts on various platforms. Wargame Vault and similar places have very strict rules about declaring when AI has been used for writing, editing and artwork purposes. That way, you can screen AI generated content out of your searches.

However, as a rules writer; I am curious how many of you are ready to spend $$$ on AI written rules for games?

It's depends on rules itself and not who/what writes it. Like It doesn't matter who get money game designer or data scientist who train AI model.

No.

Word count requirement.

Also?

No.

 Mad Doc Grotsnik wrote:
No.

Word count requirement.

Also?

No.

Same, I don't need computer generated slop in my hobbies.

Current LLM's would probably be terribly bad at generating actually working rulesets. LLM's are concerned about putting words after one another, not whether rules outlined by such word salads would actually work. You'd need some specific machine learning model that was built from the ground up to come up with game rules. They'd probably do an OK job writing narrative events for RPGs, and character dialogue though, if prompted adequately..

Absolutely not. Never. Even if it weren’t guaranteed to be soulless trash.

I also avoid self checkout as much as possible, too. I do not want to make myself complicit in my own obsolescence.

It's obvious that no current LLMs could make proper rules, but designer for this task AI make them better. Also it's can play test own rules if designed so. But it's required insane amount of work

Why would I pay for something if nobody had to labour to produce it? It's like asking if I'm okay with being charged for sunshine.

 kabaakaba wrote:
It's depends on rules itself and not who/what writes it. Like It doesn't matter who get money game designer or data scientist who train AI model.

Writers (usually) get paid a tiny pittance of the worth of what they create. The data scientist will earn less.

Data scientist got solid salary for what he do. And creating, training and supporting AI model is a long term task.

I wouldn't even take the rules if they were free.

Just say no to the Abominable Intelligence.

I'd rather eat bricks.

In fact you never know if they written by AI or 'Umies. For example core rules made by humans but all digits are adjusted by AI

Fun real world fact?

AI gibberish is already impacting my work.

Apologies for keeping things vague? But stock BS wording, citation of foreign laws or (more amusing) non-existent laws. And when actual uk laws and regulations are cited, absolutely no contextual understanding is demonstrated.

Sound likes a laugh? It’s not. Not when I have to sift through page after page after page of AI Drivel to find out what the complaint is actually about.

Upside? I know the person submitting that nonsense hasn’t read or understood a lick of what their preferred AI platform threw up. And so a wee phone call and I start to get to the heart of the matter.

Also, coming up 14 years into my career, I’m not in the least bit hesitant to call out AI scruff for what it is, and point out when it’s cited irrelevant, foreign or entirely fictional legal things.

Right now? AI is ‘Man In The Pub’. I can only hope we can nip it’s reckless and bollocks use in the bud.

 Easy E wrote:
Obviously, AI writing has come a long way in the past few years. Still, mostly a fancy auto-complete task with a bigger database to scrape, but that database allows it to put more and more things together in a more complete way than ever.

With that said, we have seen a rise in AI written books and scripts on various platforms. Wargame Vault and similar places have very strict rules about declaring when AI has been used for writing, editing and artwork purposes. That way, you can screen AI generated content out of your searches.

However, as a rules writer; I am curious how many of you are ready to spend $$$ on AI written rules for games?

A few responses come to mind:

1) No thanks, I'll just pirate it.

2) WHY would I do that? I can get the same slop by entering the prompts myself.... "Siri: Write me a new 40k ruleset in the style of GW....

3) no.

Great! Looks like I am not out of a job yet!

I fed my RPG into two different LLMs...

and they had serious issues with rules that occurred earlier on the page.

Every paragraph was a new world... and they ignored the previous rulesets, and would comment on how I was missing rules, even though they were literally 2 paragraphs earlier!

Argh!

It was an exercise in frustration.

Anyone who completes an entire manuscript with the help of AI, deserves to be sainted.

Heck, I couldn't stick it out for one section... and I'm a certified Jedi Knight.

Absolutely not. I try to avoid the slop machines as much as I can.

No, never.

Yeah, even ignoring the ethical issues around AI use and the fact that current LLMs are patently unsuited to creating a functional ruleset... if an AI could generate rules, why would I pay for them when I could just get the AI to generate them myself?

AI-generated anything is not a marketable product. When you remove the need to invest anything in creating the product, you remove the need to pay for that investment.

No.

 insaniak wrote:
Yeah, even ignoring the ethical issues around AI use and the fact that current LLMs are patently unsuited to creating a functional ruleset... if an AI could generate rules, why would I pay for them when I could just get the AI to generate them myself?

AI-generated anything is not a marketable product. When you remove the need to invest anything in creating the product, you remove the need to pay for that investment.

Echoing the chorus of "Hell no," with emphasis on Insaniak's points.

Not a chance

I’m going to go with No.

No.

I'm a programmer by trade and I already have to deal with LLM slop making software worse. I don't need a glorified random text generator wrecking this hobby as well! :-)

 Easy E wrote:


However, as a rules writer; I am curious how many of you are ready to spend $$$ on AI written rules for games?

I'm not entirely sure the 3rd edition Horus Heresy ruleset wasn't passed through an AI generator. It reads like unnecessarily complicated stereo instructions.

I do enjoy how clumsy and verbose AI writing is.

If I can say it in three words, AI will say it in 10 words to 3 paragraphs.

Absolutely not. I also won't buy products with AI art in them if I know about it.

The whole thing makes me sick.

 kabaakaba wrote:
It's obvious that no current LLMs could make proper rules, but designer for this task AI make them better. Also it's can play test own rules if designed so. But it's required insane amount of work

Ai might actually be useful for playtesting. In theory you could give AI the rules and tell it to playtest and balance points values according to its testing. IT's the kind of thing AI should be very good at as it's basically how chess engines teach themselves to play. Wargames are obviously much more complex, but the principle is the same, I suspect the problem would be with creating useable outputs but if you gave the AI the structure you wanted the points to be in (integer values only, no conditional costs based on previous selections, etc) it could work.

As far as generating the rules from scratch? Absolutely not. Aside from the problems of Ai created anything, good rules are as much about creativity as they are mechanics.

Heresy does read as though it was passed through an AI filter as the final edit, as others have pointed out here. I've rarely seen such an impenetrable ruleset in my life. I'm sure it all makes perfect sense, but it does an awful job of communicating that.

Slipspace wrote:

 kabaakaba wrote:
It's obvious that no current LLMs could make proper rules, but designer for this task AI make them better. Also it's can play test own rules if designed so. But it's required insane amount of work

Ai might actually be useful for playtesting. In theory you could give AI the rules and tell it to playtest and balance points values according to its testing. IT's the kind of thing AI should be very good at as it's basically how chess engines teach themselves to play. Wargames are obviously much more complex, but the principle is the same, I suspect the problem would be with creating useable outputs but if you gave the AI the structure you wanted the points to be in (integer values only, no conditional costs based on previous selections, etc) it could work.

As far as generating the rules from scratch? Absolutely not. Aside from the problems of Ai created anything, good rules are as much about creativity as they are mechanics.

Heresy does read as though it was passed through an AI filter as the final edit, as others have pointed out here. I've rarely seen such an impenetrable ruleset in my life. I'm sure it all makes perfect sense, but it does an awful job of communicating that.

ChatGPT and Copilot cannot do this.

I have tried.

When my human playtesters were unavailable, I thought I would feed my rules into these two LLMs and see how different effects would affect combat.

They could not remember more than a single paragraphs worth of rules and would insert their own made up rules, come up with numbers that are impossible (ie. Roll 3d6 and give me the total. OK, the total is 2. )

It was an infuriating waste of time. I spent way too long arguing with a system that... Argh. It pisses me off thinking about it.

The only cool thing Chatbots do successfully is splice in really cool backgrounds to your warhammer miniatures.

They are a waste of time and energy to do anything else.

 Easy E wrote:
Obviously, AI writing has come a long way in the past few years. Still, mostly a fancy auto-complete task with a bigger database to scrape, but that database allows it to put more and more things together in a more complete way than ever.

With that said, we have seen a rise in AI written books and scripts on various platforms. Wargame Vault and similar places have very strict rules about declaring when AI has been used for writing, editing and artwork purposes. That way, you can screen AI generated content out of your searches.

However, as a rules writer; I am curious how many of you are ready to spend $$$ on AI written rules for games?

Short answer? Feth no.

While we all heavily criticize GW's rules writing, I think we can all at least agree that they both give a crap and try (even if they miss the mark). Much like AI-generated slop in other art forms, rules would just be beyond an AI's ability to create in any meaningful way.

Also, if it's worth doing, it's worth paying a person to do it right. [insert joke about GW rules here]

 Lathe Biosas wrote:

They could not remember more than a single paragraphs worth of rules and would insert their own made up rules, come up with numbers that are impossible (ie. Roll 3d6 and give me the total. OK, the total is 2. )

There was a 40K player like that in a gaming club I used to go to...

1) I feel like this thread could age pretty fast with the development of AI.

2) Right now AI is useless for writing coherent rules.

3) I could see AI in the future being useful for basic scenario and story creation in campaigns. OPR Starquest has a basic system of creating missions, where you just tell the app which models you have available. I could see AI being helpful for stuff like that.

 kronk wrote:

I'm not entirely sure the 3rd edition Horus Heresy ruleset wasn't passed through an AI generator. It reads like unnecessarily complicated stereo instructions.

I have a new theory about this.

Since release, we've learned that the game is much improved and flows smoother than 2.0

I think that the team made a slick new book, that cut down on page flipping by culling USRs, nested rules, wargear referencing, etc... then some "bean counter" (the archetypal enemy in GW office culture stories) came in and said "oh, no no no, the book can't be shorter than the last one, do it over" and the devs were fed up and just fed it to a LLM and asked to double the page count without altering meaning.

Sgt. Cortez wrote:
1) I feel like this thread could age pretty fast with the development of AI.

AI getting better won't change the fact that there's no point paying someone else to use it for you. If anything, as AI improves and provides better output with less need for revising prompts and filtering through failed outputs, there will be even less reason to pay someone else to do it for you than there is now...

 insaniak wrote:

Sgt. Cortez wrote:
1) I feel like this thread could age pretty fast with the development of AI.

AI getting better won't change the fact that there's no point paying someone else to use it for you. If anything, as AI improves and provides better output with less need for revising prompts and filtering through failed outputs, there will be even less reason to pay someone else to do it for you than there is now...

Well... there might be some disagreement with that.

I looked up AI created RPGs and there are a bunch of them. Most are rules-less text responses.

But this means that lonely people are in the market for AI driven drivel.

 insaniak wrote:

Sgt. Cortez wrote:
1) I feel like this thread could age pretty fast with the development of AI.

AI getting better won't change the fact that there's no point paying someone else to use it for you. If anything, as AI improves and provides better output with less need for revising prompts and filtering through failed outputs, there will be even less reason to pay someone else to do it for you than there is now...

You are probably right about the original question of directly paying AI to write rules. I was rather thinking about posts dismissing anything AI outright or describing it as too dumb to write rules. Give it a couple of years.
Also, chances are many rules we pay for in a couple of years will be written with the help of AI.

The more I find out about how AI is being used, the more I understand where the Butlerian folks were coming from. "In some cases it's actually the best tool for-" Don't care. On the pyre, burn it all. The only winning move is not to play.

 Lathe Biosas wrote:
Well... there might be some disagreement with that.

I looked up AI created RPGs and there are a bunch of them. Most are rules-less text responses.

But this means that lonely people are in the market for AI driven drivel.

Or it just means that people are making it. Doesn't mean anyone is actually buying it.

When you make it easier to create something, more people will create that thing. See the proliferation of DIY T-shirt sites filled with product that people made because they could, and because they were sold the idea that it would make them money... and that never actually sell, because the market is oversaturated due to the ease with which the product can be made.

As AI becomes more accessible, more people will use it to create content that they think other people will want... but that very accessibility means that there becomes increasingly less reason to not do it yourself. Returning to the T-shirt comparison, if there's a site that lets you buy custom T-shirts that also contains an AI image tool that lets you create whatever T-shirt you want, there is no reason to ever pay other people for their designs... if you see a design you like, you just ask the AI to create it for you.

Of course, in time, that means fewer people creating new content, which gives the AI less material to plagiarise, which makes everything worse for everyone. It all only works in the long term if AI becomes properly regulated and creatives are properly reimbursed for their work being used to train AI.

AI should be kept out of the creative sphere..

I don’t want AI “art”. I don’t want AI movies. I don’t want AI music. I don’t want AI telly. I don’t want AI stories of any length.

Art and creativity are such a huge part of what it is to be human. When I read a book, I’m invited into the mind of the author. They’re communicating thoughts and views to me. When I see human made art, it can provoke strong emotional reactions, and not always what the artist intended.

These things take skill and care. Someone has put thought, care and passion into them, even if I end up not liking or appreciating the end product.

AI cannot create, only regurgitate. It cannot offer a novel view or perspective. It can’t speculate. It will never create or reinvent a genre.

Finally? Humans are perfectly capable of churning out absolute crap. We don’t need a fancy programme to do that. I mean, Tracey Emin managed to convince art world goons that a manky bed was art.

AI generated rules? No.

AI playtesting? That sounds theoretically helpful to a small operation like us, but also sounds like it's not there yet.

Buut...

AI art? Neither Jim or I can draw for gak. Neither Jim or I can Photoshop for gak without putting the year aside. We can't afford to pay an actual human artist. So we either don't do Apocalypse: Earth with art at all, or we do it with some AI art, and hope that it sells enough that we can one day do a 3rd edition that allows us to hire human illustrators.

This is tech that hasn't really shown its hand yet, and will probably take at least a decade to bed in. Meanwhile, it is what it is, and what it is has its place as long as we all recognise its not the cause of or solution to all of humanity's problems

Do Apocalypse Earth with no art at all, then. I’m not the only person posting here who will refuse to buy a product loaded with AI art.


Automatically Appended Next Post:
If you can’t draw and can’t afford art, can you take pictures of miniatures or something?

 BobtheInquisitor wrote:
Do Apocalypse Earth with no art at all, then. I’m not the only person posting here who will refuse to buy a product loaded with AI art.

Same feeling here.

Skinflint Games wrote:AI art? Neither Jim or I can draw for gak. Neither Jim or I can Photoshop for gak without putting the year aside. We can't afford to pay an actual human artist. So we either don't do Apocalypse: Earth with art at all, or we do it with some AI art, and hope that it sells enough that we can one day do a 3rd edition that allows us to hire human illustrators.

Then you use no art until you can actually afford to recompense someone who can, or learn it yourselves. I would refuse to buy a product that has used any generative AI.

How do you copyright your work with AI art in it?

I was contemplating using an AI altered map of 1987 Berlin, but I was concerned by the ownership issues of using AI art in my work.

I ended up playing cut and paste with scissors and a scanner.

I am also making a game but I can't afford a rules writer, so I'm just going to publish Apocalypse:Earth under my own name with the words moved around a bit

Lathe Biosas wrote:

Slipspace wrote:

 kabaakaba wrote:
It's obvious that no current LLMs could make proper rules, but designer for this task AI make them better. Also it's can play test own rules if designed so. But it's required insane amount of work

Ai might actually be useful for playtesting. In theory you could give AI the rules and tell it to playtest and balance points values according to its testing. IT's the kind of thing AI should be very good at as it's basically how chess engines teach themselves to play. Wargames are obviously much more complex, but the principle is the same, I suspect the problem would be with creating useable outputs but if you gave the AI the structure you wanted the points to be in (integer values only, no conditional costs based on previous selections, etc) it could work.

As far as generating the rules from scratch? Absolutely not. Aside from the problems of Ai created anything, good rules are as much about creativity as they are mechanics.

Heresy does read as though it was passed through an AI filter as the final edit, as others have pointed out here. I've rarely seen such an impenetrable ruleset in my life. I'm sure it all makes perfect sense, but it does an awful job of communicating that.

ChatGPT and Copilot cannot do this.

I have tried.

When my human playtesters were unavailable, I thought I would feed my rules into these two LLMs and see how different effects would affect combat.

They could not remember more than a single paragraphs worth of rules and would insert their own made up rules, come up with numbers that are impossible (ie. Roll 3d6 and give me the total. OK, the total is 2. )

It was an infuriating waste of time. I spent way too long arguing with a system that... Argh. It pisses me off thinking about it.

The only cool thing Chatbots do successfully is splice in really cool backgrounds to your warhammer miniatures.

They are a waste of time and energy to do anything else.

I think part of the problem with playtesting is likely that the actual rules for most games contain enough errors and require just enough human interpretation that an AI can't actually play the game properly because it can't parse the rules. That's why AI has been so successful at games like Chess, Go and Poker. The rules can be accurately described with no ambiguity so the AI can just get on with optimising the best routes to victory. In theory, if you could actually teach an AI how to play 40k I'm pretty sure it would solve a lot of the balance issues in a day. It would probably take longer to go through that process than just getting humans to do it, though.

Skinflint Games wrote:AI generated rules? No.

AI playtesting? That sounds theoretically helpful to a small operation like us, but also sounds like it's not there yet.

Buut...

AI art? Neither Jim or I can draw for gak. Neither Jim or I can Photoshop for gak without putting the year aside. We can't afford to pay an actual human artist. So we either don't do Apocalypse: Earth with art at all, or we do it with some AI art, and hope that it sells enough that we can one day do a 3rd edition that allows us to hire human illustrators.

Do it without art. AI art in a commercial product is currently on very questionable ethical grounds, for a start. Aside from that, if I see a game with AI art everywhere I'm also going to question how much of the written content is human generated and how much is AI slop.

original question: No, I would not pay money for machine generative game rules.
machines can't grasp human experiences, like what make games desirable or bad.
Even people who can do that struggle with making consistent/meaningful rule systems.
We ask a program questions, it spits out a return regardless of whether it actually solves our problem, or has meaningful outcomes to our lives. ai don't care, it can't.

 Lathe Biosas wrote:

How do you copyright your work with AI art in it?

I was contemplating using an AI altered map of 1987 Berlin, but I was concerned by the ownership issues of using AI art in my work.

I ended up playing cut and paste with scissors and a scanner.

The same way you do with non-ai art. Depends on how much additional transformative work you put into the finished product. Standalone generative images or text cannot be Copyrighted by itself, but if it's featured on a page or within a larger work (like a pamphlet, pdf or book), where you've done human-directed editing: meaningful composition placement of assets, scaling, edited written text with chosen fonts or word emphasis, etc. That arrangement, that combination as a fixed work can be Copyrighted, because it's more than just the image as an image or text as text. that image/text is now a part of larger thing.
So not just throwing an ai map in a word document and calling it a day, that won't do.
And you might still run afoul if the original source map is under a copyright and not public domain.

 Lathe Biosas wrote:
But this means that lonely people are in the market for AI driven drivel.

This includes all the AI girlfriends out there.

Yeah, I thankfully live near a smal; college with an Art program, and I have found it is a good source of low-cost art. These students need stuff for their portfolios and/or classes. Therefore, getting a small amount of pay is a bonus!

However, I find pictures of miniatures on table works very well too. If they are from your own collection you just need to attribute the maker, the painter, and the photographer where necessary. Many times, the painter and photographer is you! That makes it much easier. A great example of how to do it is in Osprey Games original Dragon Rampant game.

That said, I see the appeal of AI artwork; but it is ultimately built on an illusion and a lie.

Re: thread title:

Hell no. And it's not just about quality, ethical concerns (such as plagiarism), and so on. I simply can't imagine getting invested in a game that someone just tossed out there. It's not as if there's any shortage of amazing games that were made with passion by designers, artists and writers who believe in their artistic vision and respect their customers/players.

AI slop makers think they can put a number of individually good-tasting and nutritious ingredients in a blender, turn it to puree, serve it to me in a bowl, and have me compliment them on their cuisine. If I'm not interested in even tasting it, they'll act baffled and say: "But it's very nutritious! And you like all the ingredients I put in it." Well, maybe I'm being difficult, but I do expect more of food than nutrition. The purpose of eating is not only to get nutrients, in the same way that reading novels, watching movies, playing games, etc. is not only about passing the 8 hours out of every 24 that are mine to enjoy.

Here's a quote I think about often, regarding AI-generated literature: "Why should I bother to read a book that no one could be bothered to write?" And to me, the same applies to all genAI media. Why should I bother to play a game that no one could be bothered to make? And don't tell me: "Oh, they still needed the idea, in order to enter the prompts". Screw that. Having an idea is suuuuper easy. I have plenty, myself, yet I lack the discipline and drive to make them a reality. But I do respect those who do have both the ideas and the drive. If any of my ideas ever sees the light of day, it'll be because I got off my ass (or sat my ass down, depending) and put in the work.

There are already more games than any single person could ever play in a lifetime. At the end of the day, genAI isn't about fulfilling unmet demand with supply, because supply is already more than enough. It's really about cutting the middleman of "actually making the thing". And then be paid for it.

.

 BorderCountess wrote:

 Lathe Biosas wrote:
But this means that lonely people are in the market for AI driven drivel.

This includes all the AI girlfriends out there.

AI... girlfriends?

 Lathe Biosas wrote:

 BorderCountess wrote:

 Lathe Biosas wrote:
But this means that lonely people are in the market for AI driven drivel.

This includes all the AI girlfriends out there.

AI... girlfriends?

Stay unknowing, it’s better for your sanity !

There could be scope to do something that makes playtesting and stuff like point calculation more reasonable. I'm not against using AI to *help* with stuff, but AI output as the main product just leaves me feeling cold.


Games need to be fun, dramatic, cause some emotional response, none of which a machine can do. If I'm paying for entertainment I want to actually be paying a human to do something with it.

 kabaakaba wrote:
Data scientist got solid salary for what he do. And creating, training and supporting AI model is a long term task.

I can't see anyone creating and training models to generate games (though I guess it's possible), so I assume most AI games, literature etc will almost all just be prompt engineering.

 Skinflint Games wrote:

AI art? Neither Jim or I can draw for gak. Neither Jim or I can Photoshop for gak without putting the year aside. We can't afford to pay an actual human artist. So we either don't do Apocalypse: Earth with art at all, or we do it with some AI art, and hope that it sells enough that we can one day do a 3rd edition that allows us to hire human illustrators.

AI art is a bit icky, and whenever I've seen anything with AI art I usually dodge it because I don't have much confidence that it's only the art that's artificial. There's just too much of an association with AI art slop covers to AI book slop.

I'd much rather have no art, stock art or photos of miniatures than AI art, and feel like the author values their work enough to not degrade it like that.

Nope. I wouldn't even try it for free.

-Guardsman- wrote:

Here's a quote I think about often, regarding AI-generated literature: "Why should I bother to read a book that no one could be bothered to write?" And to me, the same applies to all genAI media. Why should I bother to play a game that no one could be bothered to make? And don't tell me: "Oh, they still needed the idea, in order to enter the prompts". Screw that. Having an idea is suuuuper easy. I have plenty, myself, yet I lack the discipline and drive to make them a reality. But I do respect those who do have both the ideas and the drive. If any of my ideas ever sees the light of day, it'll be because I got off my ass (or sat my ass down, depending) and put in the work.

This doesn't just apply to AI-generated literature. Fred Knudsen of Down The Rabbit Hole went into it in one of his past streams, I believe the one about Empress Teresa(because the author essentially dedicated the book's entire site to "answering" criticism by essentially talking about how great the idea of Teresa is and just how much he's written about that one idea) and he's talking about all literary endeavors. Essentially boils down to "An idea isn't enough." A lot of writers(especially self-published ones) are of the belief that their idea is good enough for their work to be lauded because the idea is just so good. Without the dedication and time devoted to developing the skill and the art of presenting and expounding on that idea in a skilled manner, people aren't going to take to the idea. As you said, it's easy to have an idea and along with that, it's super easy for someone to go "that's a great idea" then never think on it again because the idea itself isn't actually noteworthy or memorable.

Ignoring all the other issues inherent in generative AI for a moment, the ease at which AI allows people to go "I have a great idea!" then spit out a soulless end result with none of the passion, effort, skill, etc. of actually presenting that idea in a way the average person will care about and how many people have taken to doing it is worrying.

Yeah, for me I have pages and pages of ideas.

The difference between an idea and a finished work is turning the idea into a reality.

It is super easy to be a wargame designer. You need two things:

1. A finished game
2. People who play the game

Most aspiring wargame designers get really tripped up on step 1. A idea is not enough for people to play.

I think a better question is how many people even pay for rules when there are so many sources of free rules on the internet.

When you can find even GW rules hosted on a Russian site, what is the point of paying for rules?

 Tyran wrote:
I think a better question is how many people even pay for rules when there are so many sources of free rules on the internet.

When you can find even GW rules hosted on a Russian site, what is the point of paying for rules?

Pretty pictures and cool fluff to read.

Heck, my RPG has essentially a novella wedged into it so far.

I pay for rules for games like Horizon Wars or Urban Manhunt because they usually have some art, some unique lore, and a lot of heart. I also buy rulebooks for long dead games for the lore. Even the rule book that came with Technolog/Robogear minis back in the day was a hell of a read.

I’ve bought rule books for smaller games before and been absolutely burned by the lack of lore or flavor. I’d probably have a huge Conquest: LAOK army if their stupid starter set rule book had included anything other than the driest crunch.

Well, that was about the reaction I expected from the community after reading the title.

I would also have to say no, I'd not be interested in an AI generated ruleset, even if AI tech got better at all.

That being said, I don't think it, or at least potentially future AI. has no place in Wargaming at all.

What I mean is, think of how massive in scale 40k is, how many models per army, how many kinds of permutations of lists you could possible have, how many varied matchups there are, the possibilities are nearly infinite.

While I don't think AI has a place in game design, I think it could be used to aid in game balance, running tens of thousands of simulated games, something it is completely impossible to have actual humans do, and determine the best strategies, units, and play patters and use that data to adjust the rules of your human designed game appropriately. So many people complain about army imbalance (even though I think GW does, in general, a pretty decent job with it), having an AI mathematically determine how strong everything is against the entire field and point them appropriately could finally end that issue.

The interesting bit would be what factors that might matter more to an AI than a human, like perhaps how fast an army would play, or an army being good only if you made the best possible move in any given scenario, which just wouldn't be possible for a human.

It reminds me a bit of the Starcraft II AI bots that long ago determined that mass blink stalker wins every game hands down because and AI has such fast reaction time that they can use the ability far better than any human could.

Anyhow, those are my thoughts on the matter.

 Tawnis wrote:
... having an AI mathematically determine how strong everything is against the entire field and point them appropriately could finally end that issue.
.

It won't, unless it comes up with a system that assigns points costs based on the actual forces on the table rather than just the contents of the army book.

Set points costs can never be even close to accurate, because how effective a unit is depends so much on what else is in the same army with it, what army they are facing, and the objective of the mission in play. The points cost can never realistically represent a units' true value unless it changes as needed to take those things into account... it will only ever be a rough average, and having AI assign that average won't change players' perception of that unit's value in their personal army.


Automatically Appended Next Post:

 Tyran wrote:

When you can find even GW rules hosted on a Russian site, what is the point of paying for rules?

Ignoring the legal aspect for a moment, if you like a product then there's value in supporting the entity that made it so they can continue to make product that you like.

 insaniak wrote:

 Tawnis wrote:
... having an AI mathematically determine how strong everything is against the entire field and point them appropriately could finally end that issue.
.

It won't, unless it comes up with a system that assigns points costs based on the actual forces on the table rather than just the contents of the army book.

Set points costs can never be even close to accurate, because how effective a unit is depends so much on what else is in the same army with it, what army they are facing, and the objective of the mission in play. The points cost can never realistically represent a units' true value unless it changes as needed to take those things into account... it will only ever be a rough average, and having AI assign that average won't change players' perception of that unit's value in their personal army.

That's partially true. However, with much more data accessibility you could point units, at the very least, differently for each detachment, which would help with that somewhat. If something has an incredibly high variance between being very powerful in some situations and terrible in others, that's a point for the actual people to come in and make judgement calls about what to do with it or perhaps it needs a rule adjustment? The point would be to save the human resources for those difficult discissions and use AI as a support system to handle the data collection/crunching and meaningless busywork.

Well, virtually properly configured AI could make billions of playtests in short time and adjust points pretty neat. But rules different things. And should be written by humans

 Tawnis wrote:
If something has an incredibly high variance between being very powerful in some situations and terrible in others, that's a point for the actual people to come in and make judgement calls about what to do with it or perhaps it needs a rule adjustment?

Nope, that's just how things work in these sorts of games. For example, a high rate of fire but low penetration weapon is super effective against weak, unarmoured targets, and next to useless against heavily armoured targets. If you point that weapon against the former, it's going to be overcosted against the latter. If you point it against the latter, it's too cheap against the former. Point it in the middle and it's 'wrong' all of the time. Complicate that further by having more players using one of those armies than the other... and the more granular the points system is, the worse that gets because the variance makes a bigger difference.

AI won't fix that. Yes, a computer program built correctly could potentially analyse game data faster and more efficiently than people can... but that doesn't change the inherent imbalances that are an unavoidable side effect of having variable army lists fighting against other variable army lists with variable objectives on tables with a variable amount, type and layout of terrain. The best you can ever do is find a workable average... and regardless of how much analysis is used to arrive at that average, players will see it as wrong based on their own specific experience on the table.

^^^^ What Insaniak said.

Balance is a Unicron and AI will not change that.

Would I buy AI written rules? Nope. Not only nope, actually, but frak nope.

Despite working in the software industry, I have a very strong dislike for any generative AI (and I've actually worked on some AI/ML backends, so I understand how those things work).

Are there spots where AI might be helpful in game design/development? Probably, but not with a general model LLM. And the time it would take to properly train a model is probably not something that most tabletop game companies are gonna have the resources to do.

Regarding balance... 100% agree that is a cryptid. So long as each side can have different builds of their armies, and those builds are open to multiple permutations, it's a fools' errand to try to seek "perfect balance."

I think it's important to make that point about LLM "AI" which is more like a very, very fancy autocorrect.

Like the idea that a LLM could do anything mathematical is wrong, they're quite bad at maths and it's because they're not built to do maths, they're built to simulate conversation.

It's always very obvious to me when a kid uses a LLM to do their physics homework because the LLM doesn't really understand anything about physics or mathematics, it just can create something that looks like what someone might write about physics, or what a maths solution might look like. But it's usually nonsensical, perhaps mixing terms from related but different fields and so on.

AI is a marketing term for these programs and I'm really against using it for them. I'm also fairly confident when they start charging what these things actually cost, people are not going to find them worth it. There's probably some use cases but nothing that justifies the insane costs of the current way it works, and I'm skeptical that they'll be able to get around the hardware issues they have to improve that end of things.

 insaniak wrote:

 Tawnis wrote:
If something has an incredibly high variance between being very powerful in some situations and terrible in others, that's a point for the actual people to come in and make judgement calls about what to do with it or perhaps it needs a rule adjustment?

Nope, that's just how things work in these sorts of games. For example, a high rate of fire but low penetration weapon is super effective against weak, unarmoured targets, and next to useless against heavily armoured targets. If you point that weapon against the former, it's going to be overcosted against the latter. If you point it against the latter, it's too cheap against the former. Point it in the middle and it's 'wrong' all of the time. Complicate that further by having more players using one of those armies than the other... and the more granular the points system is, the worse that gets because the variance makes a bigger difference.

AI won't fix that. Yes, a computer program built correctly could potentially analyse game data faster and more efficiently than people can... but that doesn't change the inherent imbalances that are an unavoidable side effect of having variable army lists fighting against other variable army lists with variable objectives on tables with a variable amount, type and layout of terrain. The best you can ever do is find a workable average... and regardless of how much analysis is used to arrive at that average, players will see it as wrong based on their own specific experience on the table.

You're correct, but also missing the point.

There will always be edge cases, but when you have an AI that can run millions of game simulations, averages will start to show up based on a given army's performance vs the entire field. It wouldn't just spit out win/loss and VP, you'd get average damage outputs, survivability, contributions to VP, the whole nine yards.

The system will never be perfect, but it would give GW FAR more data to work with on how to balance the game. That's all it is for, data collection. Maybe they work out a math algorithm that assigns a point cost to it, and maybe they do that all by people looking at the data, it doesn't really matter since we don't know how they come up with points for every new edition that launches as is.

My point is that more information can only help with game balance, but the final calls on how to use that information should still be done by people, and that a workable average is really what you want if you are trying to get into that magic 50% win rate that GW strives for. If half the time your a high rate of fire but low penetration weapon feels like its pulling a lot of weight in a match up, and half the time it doesn't, isn't that what they're going for?

No, I get the point, I'm just questioning the actual need.

And ultimately, given the sheer number of variables in 40K, I would question the likelihood that an engine could actually be developed to consider them all in a way that would produce measurably better results than the current system. It's nice in theory, but seems like a lot of work to set up, for minimal actual benefit.

LLMs aren't simulation engines, so moot point.

It may be possible to build such an AI, but it would have nothing to do with the current crop of "AI" everyone is talking about (and would be far more expensive).

I mean, LLMs aren't a viable option for creating a functional set of games rules, either, (at best, you would get something that looks like a set of game rules, but internal consistency and actual function would be unlikely) so the premise of the thread is already relying on new technology.

Me talking about specific ML algorithm and not a chat - bot most here talking about. Why should chat bot do balance game.also talking about complexity of game, it's not really that complex. Algorithm which balancing data center resources we use in company iirc use 700+ parameters.

 Tyran wrote:
I think a better question is how many people even pay for rules when there are so many sources of free rules on the internet.

When you can find even GW rules hosted on a Russian site, what is the point of paying for rules?

Because I'm not a pirating oxygen thief, and don't believe I'm entitled to the content of a book without buying the damned thing.

Quite interesting to read some of the replies here, and as expected it is mostly negative.

Here is an extra (I will say very facetious) question to add to that:
Are you definitely not OK with AI-written rules, but in the same breath are you also OK with GW not crediting its rules writers and developers?

I have read some accusations that some of the recently released rules (notably the new Horus Heresy) are at least in part AI-written. I personally don't think this is the case (in that case it just needed an editor), but it does now open the door to this sort of accusation.

I'm absolutely not okay with them not crediting writers.

 insaniak wrote:

 Tawnis wrote:
If something has an incredibly high variance between being very powerful in some situations and terrible in others, that's a point for the actual people to come in and make judgement calls about what to do with it or perhaps it needs a rule adjustment?

Nope, that's just how things work in these sorts of games. For example, a high rate of fire but low penetration weapon is super effective against weak, unarmoured targets, and next to useless against heavily armoured targets. If you point that weapon against the former, it's going to be overcosted against the latter. If you point it against the latter, it's too cheap against the former. Point it in the middle and it's 'wrong' all of the time. Complicate that further by having more players using one of those armies than the other... and the more granular the points system is, the worse that gets because the variance makes a bigger difference.

AI won't fix that. Yes, a computer program built correctly could potentially analyse game data faster and more efficiently than people can... but that doesn't change the inherent imbalances that are an unavoidable side effect of having variable army lists fighting against other variable army lists with variable objectives on tables with a variable amount, type and layout of terrain. The best you can ever do is find a workable average... and regardless of how much analysis is used to arrive at that average, players will see it as wrong based on their own specific experience on the table.

I don't think that's as good an example, because you'd presumably average it out over thousands of games with random combinations of armies. An anti-infantry unit will presumably have a fairly average scoring against a random mix of infantry and armor.


Where I think it'll be more useful is catching unit buffs which are often easily broken. For instance a character that gives re-rolls to any unit within x range, will have a completely different outcome if the unit is Grots or Grey Knights. It'll also be easier for an algorithm (AI or not) to catch really obscure game breaking combinations of special rules and units.

Being able to calculate playtests in a huge data set would be really useful: 1000pts of X beats 1000pts of Y 50% of the time is perfect, whereas 900pts of X beats 1100pts of Y 60% of the time is broken.

 Da Boss wrote:
I'm absolutely not okay with them not crediting writers.

When you look back at some of the bile and vitriol directed at certain rules writers? It becomes more reasonable.

 Mad Doc Grotsnik wrote:

 Da Boss wrote:
I'm absolutely not okay with them not crediting writers.

When you look back at some of the bile and vitriol directed at certain rules writers? It becomes more reasonable.

I bet Mat Ward wished his name wasn't mentioned on a certain codex or two.

 Mad Doc Grotsnik wrote:

 Da Boss wrote:
I'm absolutely not okay with them not crediting writers.

When you look back at some of the bile and vitriol directed at certain rules writers? It becomes more reasonable.

This. When you know why they stopped attributing, it makes more sense.

 Valander wrote:

Despite working in the software industry, I have a very strong dislike for any generative AI (and I've actually worked on some AI/ML backends, so I understand how those things work).

Same, I actually work with AI, using Machine Learning for diagnostics, but I'm really uneasy about how much generative / LLM AI is being shoehorned in as a solution for everything given as pointed out it's just a giant autocompletion and is still famously but confidently incorrect often.

 Tyran wrote:
LLMs aren't simulation engines, so moot point.

It may be possible to build such an AI, but it would have nothing to do with the current crop of "AI" everyone is talking about (and would be far more expensive).

I agree here too, though I think AI as a term is in such a bubble now pretty much any software is called an AI. A proper simulation engine, possibly even based off of something like Total War, would be able to do all of the number crunching.

An LLM may be useful for more language based tasks like checking for ambiguous wording or words which have confusing/bad meanings in other languages.



It's also worth highlighting that 40K isn't all of wargaming, and that whilst it's a total mess where balance is impossible, lots of other games have closer balance and less variation to grind though.
For example a historical skirmish game probably doesn't have the ability for an infantryman to fist-fight a Titan, or for a unit to have multiple stacked saves, re-rolls etc. The difference between a guy with an axe and a guy with a spear is minimal.


Automatically Appended Next Post:

 kabaakaba wrote:
Me talking about specific ML algorithm and not a chat - bot most here talking about. Why should chat bot do balance game.also talking about complexity of game, it's not really that complex. Algorithm which balancing data center resources we use in company iirc use 700+ parameters.

To most people, AI means a LLM.

But I don't think an ML (or at least a Neural Network, which is what I assume you mean) is the right tool for the job here either. Partly because in order to train it you'd need to have another algorithm generate enough inputs and good outputs, and then you may as well use that algothim.
NN's are great for something small in scope for detecting if an image contains a cat, and it's pretty easy to train given enough data, but where would you get the data for units in a wargame?

Unless you mean some other form of ML.

Not okay with no writing credits.

Also, further up we seem to be talking about using a computer Model to help with balance rather than a LLM. However, there is a limit to what computer modelling can do, and there is a reason no one has done it yet.

Absolutely not. No AI art, no AI writing. It's Theft, and it's scummy as hell.

What about using AI to alter photographs?

Like removing a vehicle from a photo, or adding a street lamp?

Herzlos wrote:

 insaniak wrote:

 Tawnis wrote:
If something has an incredibly high variance between being very powerful in some situations and terrible in others, that's a point for the actual people to come in and make judgement calls about what to do with it or perhaps it needs a rule adjustment?

Nope, that's just how things work in these sorts of games. For example, a high rate of fire but low penetration weapon is super effective against weak, unarmoured targets, and next to useless against heavily armoured targets. If you point that weapon against the former, it's going to be overcosted against the latter. If you point it against the latter, it's too cheap against the former. Point it in the middle and it's 'wrong' all of the time. Complicate that further by having more players using one of those armies than the other... and the more granular the points system is, the worse that gets because the variance makes a bigger difference.

AI won't fix that. Yes, a computer program built correctly could potentially analyse game data faster and more efficiently than people can... but that doesn't change the inherent imbalances that are an unavoidable side effect of having variable army lists fighting against other variable army lists with variable objectives on tables with a variable amount, type and layout of terrain. The best you can ever do is find a workable average... and regardless of how much analysis is used to arrive at that average, players will see it as wrong based on their own specific experience on the table.

I don't think that's as good an example, because you'd presumably average it out over thousands of games with random combinations of armies. An anti-infantry unit will presumably have a fairly average scoring against a random mix of infantry and armor.


Where I think it'll be more useful is catching unit buffs which are often easily broken. For instance a character that gives re-rolls to any unit within x range, will have a completely different outcome if the unit is Grots or Grey Knights. It'll also be easier for an algorithm (AI or not) to catch really obscure game breaking combinations of special rules and units.

Being able to calculate playtests in a huge data set would be really useful: 1000pts of X beats 1000pts of Y 50% of the time is perfect, whereas 900pts of X beats 1100pts of Y 60% of the time is broken.

I design for a similarly complex game (Magic the Gathering). We can collect those type of win rates from our online clients. Data like “decks containing this card had a 54% win rate.” The data is fairly accurate because of the large number of games played. In theory, a simulation engine could generate that data before we print the cards, and we could make balance adjustments. Such an engine would be really resource intensive to create, and Magic is already a game that can be played digitally. I have to assume that a Warhammer simulation engine would be a lot more difficult to create. I can’t imagine that the investment would be remotely worthwhile.

 Platuan4th wrote:

 Mad Doc Grotsnik wrote:

 Da Boss wrote:
I'm absolutely not okay with them not crediting writers.

When you look back at some of the bile and vitriol directed at certain rules writers? It becomes more reasonable.

This. When you know why they stopped attributing, it makes more sense.

Ideally there would the be some way of acknowledging all of their writers and developers without assigning blame for any particular codex flaws.


Automatically Appended Next Post:

 Lathe Biosas wrote:
What about using AI to alter photographs?

Like removing a vehicle from a photo, or adding a street lamp?

I know a couple of photographers who consider this to be more like an advanced photoshop tool. One of them with use the AI and then go back and clean up with regular methods. The other will use the AI and then sell the client a janky-ass image that looks more off the closer you look at it. But for part of a slideshow, it works fine.

I do want to make one thing clear. It is extremely likely GW is already using AI for its writing. GW already owns a massive amount of text on its own lore, rules and art, so scrapping the internet for content isn't needed.

It will be the same for any large company that owns large amounts of text and data on which to train AI.

It is the smaller independent writers and developers that are stuck as either being unable to compete or being forced to use AI. But for corporations? AI is here to stay, specially corporations in games and entertainment industries.

 Tyran wrote:
I do want to make one thing clear. It is extremely likely GW is already using AI for its writing. GW already owns a massive amount of text on its own lore, rules and art, so scrapping the internet for content isn't needed.

Given the sharp drop in quality of both technical writing and lore, that is obvious

Right now I'm playing with AI to make my photos I took look like grainy "1970's black and white spy photos."

 Tyran wrote:
I do want to make one thing clear. It is extremely likely GW is already using AI for its writing. GW already owns a massive amount of text on its own lore, rules and art, so scrapping the internet for content isn't needed.

You mean in its published writing? That’s a pretty serious accusation for you to toss around without presenting any evidence.

 Platuan4th wrote:

 Mad Doc Grotsnik wrote:

 Da Boss wrote:
I'm absolutely not okay with them not crediting writers.

When you look back at some of the bile and vitriol directed at certain rules writers? It becomes more reasonable.

This. When you know why they stopped attributing, it makes more sense.

Do we think that GW writers, even Matt Ward at his most ridiculed, get even 10% of the excrement thrown their way compared to the likes of the Star Wars or Marvel writers? And a lot of those guys are based in countries where it goes beyond to death threats, and the people making those threats have easily obtainable means to back them up - and unfortunately every now and again a lunatic goes through with it. And yet we still see names and script-writing and production credits.

GW have said that is the reason, I think its for the far more prosaic reason that it stops developers making a name for themselves and then moving on, as has happened many times previously. The culture in the company is such that you are absorbed, hive mind-like, and your creative output is wholly owned by GW. So for this reason I can see why they would have absolutely no compunction at all about using AI to write the rules, background text etc, even if they are not doing so already.

If an individual writer judges that they would prefer not to be credited because of despicable losers giving them abuse, then that's fine and good. But they don't give writers or artists the choice to be credited. So I think the Matt Ward excuse is just that, and extra disgraceful for using what happened as cover to treat others poorly.

Wow, this theory really makes sense if you start with the premise that GW is full of lying, unethical morons.

To be fair, it doesn't need to be a lot of them, just the guy in charge.

 Pariah Press wrote:
Wow, this theory really makes sense if you start with the premise that GW is full of lying, unethical morons.

Not the furthest stretch though

Well, you know what they say, it takes one to know one. Bye!

 Lathe Biosas wrote:
Right now I'm playing with AI to make my photos I took look like grainy "1970's black and white spy photos."

Surely that technology predates the AI buzzword bubble? I've done it in, like, 2012, for my Necromunda gameplay shots, with a mobile filter app.

 Pariah Press wrote:
Wow, this theory really makes sense if you start with the premise that GW is full of lying, unethical morons.

Have you never worked in a large organisation before? Lying unethical morons are usually pretty good at weaseling into positions of power. I can't think of any reason why you wouldn't leave it up to the artists to be credited or not that is actually about protecting and valuing the artists, and I also find it extremely unlikely that all of the artists and writers and GW unanimously decided that they don't want to be credited for their work any more.

So what's your explanation? Do you have a better one than me?

To be honest most of GW fluff pieces nowadays feel like AI writing.

 lord_blackfang wrote:

 Lathe Biosas wrote:
Right now I'm playing with AI to make my photos I took look like grainy "1970's black and white spy photos."

Surely that technology predates the AI buzzword bubble? I've done it in, like, 2012, for my Necromunda gameplay shots, with a mobile filter app.

Yeah, my previous, dumber phone had editing software that could do that.

 BobtheInquisitor wrote:

 lord_blackfang wrote:

 Lathe Biosas wrote:
Right now I'm playing with AI to make my photos I took look like grainy "1970's black and white spy photos."

Surely that technology predates the AI buzzword bubble? I've done it in, like, 2012, for my Necromunda gameplay shots, with a mobile filter app.

Yeah, my previous, dumber phone had editing software that could do that.

Everything I read nowadays says that all photo Manipulation that I don't do personally, but use a computer program to make the adjustments at my command is labeled AI.

Am I wrong?

AI is a massive term that refers to any decision making algorithm.

AI has been a thing for over a century.

The current "AI buzzword bubble" are Large Language Models, which are a specific type of AI.

 Lathe Biosas wrote:

 BobtheInquisitor wrote:

 lord_blackfang wrote:

 Lathe Biosas wrote:
Right now I'm playing with AI to make my photos I took look like grainy "1970's black and white spy photos."

Surely that technology predates the AI buzzword bubble? I've done it in, like, 2012, for my Necromunda gameplay shots, with a mobile filter app.

Yeah, my previous, dumber phone had editing software that could do that.

Everything I read nowadays says that all photo Manipulation that I don't do personally, but use a computer program to make the adjustments at my command is labeled AI.

Am I wrong?

Yes. There used to be preset filters (I don’t know the actual name) that changed a cluster of settings. I’m pretty sure the software that comes on an iPhone direct from the manufacturer has at least an “automatic” setting to get a basic job done.

Am I understanding you saying anything that takes “adjust three slider bars” and turns it into a single click is AI?