We here at PC Gamer can’t stop admiring the balls of Blizzard’s VFX artists (opens in new tab), but it turns out that suits elsewhere at the developer have some pretty big kahunas too: And have applied for a patent that will see an AI generate artwork for use in-game (opens in new tab), based on a specific style fed to it by Blizzard (spotted by GameRant (opens in new tab)).
Ruh-oh. Firstly, the patent itself is not necessarily an indication that Blizzard is or will be doing this anytime soon. But it is treading into some murky waters here, with the reputation of AI art among the public at something of a low: Only yesterday, people were getting upset at the System Shock remaster (opens in new tab) for creating a promotional image of its own evil AI using the AI software Midjourney.
But these technologies are rapidly spreading, to the extent that even modders are incorporating them into their work, and unlike for example NFTs (another much-maligned piece of contemporary tech) there are some compelling use cases for AI, especially when it comes to fleshing out the edges of worlds.
The patent is a technical document and is absolutely no fun to read, but essentially it explains an AI system that is fed a specific art style and emulates these images to create new art. The AI takes these initial inputs and, essentially, begins to construct a 3D model: Building a skeleton, texturing it, and adding detail. I suppose the crucial point here is that initial human input, which in this case would presumably be coming from Blizzard’s internal art teams (one of the biggest controversies over AI art is its use of copyrighted imagery without credit to the human creators of it, to the extent Getty Images is suing one of the largest tools (opens in new tab)).
It’s a tricky one. You can see why the idea of this system would appeal to some of those on the production side of videogames, especially on the scale Blizzard works at with some of its titles, and it’s important to note that (for now at least) the focus seems to be on how AI can be used for things like NPC conversations and generating lots of minor variants from a single enemy design.
There are obviously hardliners who don’t think a piece of AI-generated art should ever find its way into a videogame, but there are more nuanced worries about the potential impact of tsuch technology. Industry artists might hold the reasonable concern that, even if this thing needs to be fed Blizzard art to work, that could still mean a reduction in headcount for Blizzard’s art department, and we’re not even touching on how it could change the production process: Do we really want a future where the human artists are focused on touching up AI output?
These questions are going to become ever more pressing because, as this patent shows, companies like Blizzard are not going to wait around for permission to test these technologies, and we may see the outcomes sooner than we think. Whether companies will be upfront about using such technologies, and whether we’ll even be able to tell, may end up as the bigger question.
