So I was supposed to be writing about a province in the ChargeReal that’s about to see a whole lot of warfare, and finishing off a few Militia fighters from Combat Octopus. Instead, I get to step back from the creative side to talk about the current hot thing – Generative AI.
In the last few days, I’ve received emails from four separate people asking me to take a look at a reasonably popular system listed as a Modern Warfare Skirmish game. You may have seen it, the guys behind it have hammered out a pile of content for it, and lots of people seem very keen to recommend it. Putting aside my own issues with the core rules and their attempt to model modern combat (aka, it doesn’t), the bigger problem is that it is dripping with generative AI art, used throughout the books. A lot of it is very obvious, especially to someone with an interest in the specific details of Ultramodern and whose Spidey sense gets kicked off. Its use is also not disclosed, despite it being very obvious to even the untrained AI.
However, it’s not the only ruleset I’ve looked at that has the same issue. I’m in a strange position where, for some reason, people like to send me stuff to look at, either to give feedback on or as a request to feature on the blog. Up until now, every time one of these has arrived with obvious generative AI inclusions, I’ve quietly said that it’s policy not to cover it and left it at that. You may notice I didn’t cover a certain project from an STL producer last year and it was due to its use of AI-generated assets in its marketing – instead, I delayed and covered the items after they were released to their store. The unsaid policy worked.
However, with the sheer amount I’m receiving, I think it’s time to make it clear my position. If you include assets created using Generative AI in your Wargaming Product, ChargeBlog will not talk about it, be that in The News Stand or in specific features. If it is later discovered that something that WAS covered was using AI, coverage will cease and potentially have the existing article removed.
There are plenty of people who have written in great detail online about the ethics, moral, legal and environmental issues around AI (this post about the sustainability of LLMs was particularly concerning) but from perspective as a wargaming consumer, let me explain to you why I personally don’t like products that utilise Generated AI.
And no, I’m not going to use AI Artwork for this. Instead, have a few good examples of non-AI work to fill in the gaps.
From the first point, it looks bad. Distractingly bad. I know some people turn their nose up at stock photography for their inconsistencies and inaccuracy, but at least they are using actual real-world items and so are, for the most part, recognisable as a real item. No matter how hard you try, generated art hasn’t yet managed to understand relatively static items such as common gun designs, continually mixing and merging elements from every assault rifle it’s seen into something that makes South Park’s guns look incredibly accurate recreations. When you roll this out to the people, you see it across everything from poses and equipment down to such minor things as consistent hair. The visuals with a ruleset are meant to inspire you, meant to get you in the mood to play and put you in place, not present you with a generic mess that pulls your eyes from the rules to the artwork designed to accompany it. In addition, most AI-generated art is of massively varied art style, meaning each page clashes with the next rather than providing a consistent visual brand or sensation to perform its role. More importantly, most of these rule sets that go heavy on AI lack the most important part of a rulebook – useful diagrams to help identify and illustrate key rules concepts – when you play Skirmish Sangin, multiple pages are dedicated to showing you how things such as line of sight work. Merely showing two poorly detailed characters with lines next to each other is not a suitable replacement.
Building a well-made rulebook is a skill and an effort. It takes time to make all the pieces fit together, combining the written word with everything needed to both illustrate how the game works and prepare the player for the experience. When you read something like BPRE28’s rulebook and expansions, every page is laser-focused, crafted together, merging art, diagrams and words to prepare you for the game you are about to play. And it’s not just about being pretty – every hour of work you do when writing the book to help the player will save tens of hours down the line. It is a vital piece of effort that you, as a wargames creator, need to do.
The second point is more of an implication – if someone has chosen to generate their art using AI, to attempt to save time and money through giving up control to the machine, what else have they potentially also passed off to the box in the cloud? What other corners have been cut to get something that (on the surface) looks fine but on the inside might be rotten to the core? How much of the prose that is written to illustrate the game was written and checked over by a human to make sure it was both accurate and makes sense? We’ve seen plenty of rulesets made by human hand released that need a second or third iteration to make into something that stands the test of time, and that is after it’s gone through plenty of eyes to check. While you can prepare a ruleset in a very small time, it takes much, much longer to go through it all, to rip it apart, find the exploits and hacks players will find to break what, on a basic run through, seems to make sense. This all requires work and time and is unfortunately one of the first things to be dropped by the wayside when you push to get something out. Putting aside all of the mechanical issues with not checking and double-checking your system, there is a moral issue – publishing a ruleset that is untested and charging money for it is incredibly disrespectful to your customer base. It also poisons the well, as wargamers are vastly more likely to complain about a ruleset that has wasted their time and it’s hard to claw back your goodwill from such events.
So what should you, a wargames writer whose hand is hovering over the ChatGPT link, do instead? I’m going to be the first to admit, creation is hard work – it’s blood, sweat and tears, spent now so your players have a glorious time later. But there are plenty of options to help with it – for one, collaboration. Find other writers and artists to fill the gap in your knowledge and help push forward – yes, this does require a little give and take (and fair compensation for their work) but you also gain the benefit of everything have gone through two sets of eyes, two different understandings that will help clear up and refine the vision. If you’re worried that someone else might be using your AI, asking for work in progress element should help deal with that.
For visuals, the most obvious way to gain more imagery is literally sitting in front of you – the figures you use to play the game. It doesn’t require a massive number of figures or terrain to provide a mass of variants to fill your book. If you have a collection of STL files, then combining them with Blender and using the lightning engine and simple textures will give you even more options. Combined with simple Photoshop elements, these provide the perfect setup for assembling easy-to-read gameplay diagrams. For realistic images, there is a collection of stock photography websites that can provide pictures for commercial use for a selection of prices (ranging from 3 to dangerously expensive). Especially for an ultramodern player, many photographs from the last 20 years have entered into stock photography archives, so it is very possible to find a suitable piece to match your ruleset.
If you feel like there is nothing wrong with Generate AI, despite all the news and details that the world is providing to you, and you want to continue using it in your products and life, then go for it – I can’t stop you, and even I can see it’s limited uses for crunching data. But please, please, Please, stop sending it to me to look at.
Leave a Reply