Before we kick off this blog, a quick disclaimer: I work as a software engineer. I have been in the field for almost a couple of decades, and I am currently focused on seeing what this whole AI charade is about and what to do with it. Maybe this entitles me to an informed opinion or two; maybe not.
Recently, I had this game almost done, a huge amount of illustrations paid for. And I decided to give those away, scratch everything, redo it using genAI, and be public about it. To be honest, I thought about doing it using old adverts for a moment, but where’s the fun in that? You may have already heard about The Idles.
When you do something like that, you expect it to become a bit of a honeypot for angry dogmatism; that is kind of the point. What is interesting about this experiment is that it’s mostly coming in two flavors: arguments claiming that AI equals IP theft, and those with environmental concerns.
The plagiarism
If we talk about IP theft in the board game industry, we are mostly talking about artwork and therefore stable diffusion genAI models. I’m also assuming in order for them to be considered evil, they have to be used to produce major pieces of artwork as opposed to just as an aid—like in many Photoshop features. But maybe I’m being too picky already.
There are many models that do not even use copyrighted materials. There are even recent rulings vs. Stability AI in the UK that you may want to read that make this debate way more nuanced. But let’s go back to black and white, shall we?
Midjourney is arguably the most popular one and is known for having been trained using the LAION-5B dataset, which includes some illegally sourced materials. What is a dataset and how is it trained? Basically, they scrape everything under the sun, they curate it, and they serve it up under a creative commons license. It’s all very brute force and impersonal really, nothing like a good old burglar breaking in the middle of the night. Then people go and train their models on these datasets because it’s convenient.
The two largest lawsuits against Midjourney include one put forth by Dreamworks, Disney, and a couple of other big Hollywood hitters. They basically claim that Midjourney is being used to reproduce their Mickey Mouse IP. Midjourney claims that the process is transformative and therefore it’s fair use. If you understand how stable diffusion works (just Google it), they have a point. To be honest, I have little interest in corporate lawsuits and Mickey Mouse. The second is a class action lawsuit by a bunch of independent artists claiming that their works have been illegally sourced—that one I have way more sympathy for, and I really, really hope they get some money out of it. The rulings will most likely come out mid-2026 and may involve some large settlements. Or they may not. So for all of those taking action against AI companies training on illegally sourced materials, kudos to you, keep on whistleblowing. Regardless of the outcome, say Midjourney v8 was trained using synthetic data to avoid being dragged across courtrooms in the future. Or say Adobe Firefly which uses their own legal stock data. Will we be able to make the sweeping generalization that using genAI equals IP theft when the model in question is ‘clean’? Maybe not. So, does it mean it will become alright to use?
About destroying the climate
Onto the environmental concerns. There’re ways and ways of doing things. The current way is about a bunch of K-sniffing billionaires racing for AGI (artificial general intelligence), creating the largest bubble in the history of mankind, and racing to scale up the whole thing by sheer computational power and large models—building more and more data centers in the process. Or at least saying they’ll build them.
It is basically the brute force approach, again. Even in this worst - case scenario, the environmental impact is orders of magnitude lower than many other issues we have on the plate. A lot of the papers in circulation are already a year old and project the AI environmental impact to grow proportionally to what has happened so far. That is, simply put, wrong. Let’s look at the gold standard of measuring the environmental impact. Currently, data centers are responsible for 0.5–1% of global CO2 emissions (only a fraction of that is AI). Transportation is about 16%, a lot of it avoidable if we de-globalize the planet. Just giving ideas.
Eventually, the bubble will pop, as they always do. Hard to tell when, but probably within the next 1–2 years. Eventually, this gold rush will consolidate and we will probably look at a different way of doing things - a.k.a. the Chinese way. Which consists of a different philosophy where we produce smaller, more specialized models that consume fractions of that energy. Nobody wants to pay large energy bills, especially if your investors don’t front them. Kind of what they did with DeepSeek. So if the energy consumption becomes sustainable, will it make it alright to use AI? More importantly, what if I use a stable diffusion model locally on my PC instead of playing GTA VI? Is that okay?
Wait, but I’m still pissed off
I gotcha mate, so am I. Try waking up every day to the news that some coding agent will take your job for the last two years and seeing a good chunk of your industry being fired because of that illusion. And rehired shortly after.
It’s hard to put a finger on what feels so profoundly off about AI, but it just does. It’s also a very complex reality that refuses simple childish mantras such as ‘AI is bad’. More importantly, it’s just a tool, and it’s not going away.
So here we are just pissing everybody off a little, which is something I absolutely adore to do. But it’s done for a good reason, just wait and see.
[To be continued]
Bonus track: Midjourney is one of the only AI companies that is actually hugely profitable, so I wouldn’t hold your breath and wait for the bubble to burst and wash them away. Maybe it’s time for a plan B.