Why are game publishers forcing staff to use AI wrong?
We're on a path to even more generic games.
Game publishers, who seem to go in the wrong direction and sticking with it, every time new tech surfaces (I’m looking at you Ubisoft, with your recently published NFT’s), are at it again with AI.
It seems to be perfectly clear that ChatGPT and the likes can code almost perfectly, or even optimise code to run better. Guru’s like Github, seem to think coders will be out of a job pretty soon, although, not saying it with as many words. You could imagine that coding on games, which is quite a big portion of the creation of the things, would be improved by AI. Or at least would make the process easier, quicker and less faulty than it was before. But game publishers don’t seem to see it that way. They don’t like creatives within their companies and would like to AI to do their jobs.
Blizzard for one, just announced Blizzard Diffusion, a reference to Stable Diffusion, the AI that creates images based on plagiarised images from artists all over the world. Blizzard said, in a leaked e-mail, that Blizzard Diffusion should be used for items like clothing on characters and more. It would speed up the creative process on design elements that are of lesser importance.
But that is just the thing, the design of a game, the look and feel if you will, is arguably the most important part. And Blizzard has historically been at the very top of creating amazing characters and generally unique looking games that make the games what they are.
It’s just that I can’t get my head around why you would use AI on the creative part of the games, and not on the technical parts. Even now, when the technical problems seem to be at such a point that they stifle the success of a possibly great game because of the poor performance of the damned thing.
So what Blizzard (and others) seem to be doing is removing the soul of the game /and/ make it run badly on release. Seems like a good move, right?
And when I say soul, I mean the craft, feeling and intent the artists working on great games put into their work. Great design made with great care and amazing intent leads to something the gamer can tap into. Maybe the gamer can even feel. It’s an emotion being conveyed through the art of games. A robot will never be able to do that, not even Stable Diffusion. It’s just not programmed to work like that.
Let me explain why not: AI takes prompts (input) and generates a response based on a dataset created of millions, or even billions of images and texts. It then looks for correlation in the dataset and determines the ones that fit best. From that it creates an output which is by definition a summary of the works it’s been trained on. It’s a generalisation of the dataset, a middle of the road image or text without any feeling, intent or emotion behind it. Those parts have been stripped out when the images or texts got demoted to ‘training data’.
In other words: while at first glance it looks like AI can mimic human creativity, it lacks in everything when you take a longer, closer look at it. It creates empty images and generalised texts. It’s not art anymore, it’s just generally created stuff which is great for mass use or social media posts.
Games are an art form, especially the ones that get a lot of praise for the emotional response they trigger in gamers. And getting to that point is creating something that is more than the sum of its parts. It needs a human touch, and removing that and letting a robot do it for you, will get you a general touch.
But more to the point of the weird usage of AI: the all important emotional response can be ruined by a janky running game, and design that does not resonate or communicate any form of human creativity. Which is where we seem to be going, if publishers get their way.
So now we wait until game publishers see the error in their ways, or maybe even see that AI isn’t the bandwagon you should be jumping onto without due care. Maybe the developers and creatives can get loud enough that management will hear them, but that has proven to at least be very hard in the past.