I Tried Replacing Myself with AI
I'm a Senior Concept Artist, I’ve worked with AAA and indie studios such as Sony Playstation (Housemarque), Embark Studios, Electronic Arts, and a few others. I tried to replace myself with AI.
The results follow below.
Step 1 - Generating the AI Concept Art
The AI version is modeled after what we call in the industry "style over substance", something that looks cool, but is impractical. It'll progressively break the more you use it.
We're reaching the point where most creators launch AI-generated content claiming it'll change the world and dethrone entire industries, but actually, most of them are mid. Rather than just stating that, I pointed out some areas where the AI-generated version falls short, and instead of just discussing them theoretically, I decided to demonstrate those flaws by painting over them.
You can argue there's a certain style in having a round hat in a war zone, but fashion is rarely a priority in a scenario like this; and we're growing increasingly exhausted by chasing trends (what's cool today will rarely look cool in a couple of years under this premise). This is why it's good to have an intention before you start working. It's good to understand your subject matter before you attempt to create it. (Yes, I'm flexing the process)
Why AI Got Wrong:
Writing this list of fixes felt akin to teaching an intern on the basics of the job, but even interns would have gotten a few of these things right without my help.
The Rat King Allegory - Why Most Gen AI Fails in Replacing Artists in Game Dev
In nature, there's the idea of the rat king, where many rat tails get tangled together, preventing the rats from moving independently. As a result, they rely on other rats to bring them food and even protection from predators. However, this entrapment eventually leads to their demise as they rot and die together. Gen AI and AI art is increasingly dependant on humans in the same way, not the other way around.
The parallel with an economy of ideas generated to tick boxes rather than a cultural vision is real. If you're in a project where nothing clicks, are you relying on a chaotic, entangled mess of what an algorithm is serving you? Eventually, finding new (and original) ideas will be almost impossible.
The Solution for AI Generated Generic Art
If you're stuck in your project right now, how are you researching? Are you accepting what gets served to you by the algorithm, or are you filtering content to your needs, and locked in an intention?
New ideas don't need pre-planning, and to be pre-served, they need distance, exploration, and childlike curiosity to take shape.
There are probably uses for AI, such as creating assets in mass, for unimportant NPCs or filler background assets in game development, though we run the risk of it stopping being efficient and becoming cutting corners when you're dealing with important work that relies on storytelling to thrive.
This process is simple:
1. Identify the objective; it's usually what most are ignoring. It's the difference between writing a creative brief stating "Combat medic character design" and "Combat medic character design, with XYZ specific look, to help marketing sell this character skin more than the others have sold yet. For this, we’ll try ABC and 123."
2. Build a story around it. People relate to people. They want to see their personality, backstory, and in what other ways they'd make the same decisions, choose the same gear for this battlezone.
3. Deliver it aesthetically awesome, still.
And most important of all: don’t forget to slow down and have fun; creative projects are supposed to be fueled by joy and growth; they’re not a race. The demand for personal work and personal perspectives has been on the rise against the “design by committee” for years; this is a market you can safely win as long as you focus on what matters: research and intention.
If you enjoyed this article, don’t forget to check my portfolio on the frontpage of this site, and read more over my blog.