Click here to see the poll and vote.
Over the past months, the subject of AI images (which are currently allowed and widely used on this wiki) has been brought up and discussed multiple times, by different users.
Some users and staff members have spoken up against the practice of allowing AI-generated pictures. Our admin, ManBearSmoog, has kindly asked me to start a discussion, so that we can weigh up the pros and cons, and I'm myself curious to know what others think about it.
The (potential) benefit of AI images[]
Let's start with the pros, then. First, I do genuinely believe that AI images have the potential to be a helpful tool for this wiki, for the following reasons.
- AI can be used to illustrate subjects that otherwise wouldn't be illustrated at all. Some subjects are so obscure that even fanart is extremely rare or nonexistent.
- AI allows us to create the images ourselves. This is better than googling for fanart and uploading it without the artist's permission (if you DO have the artist's permission, that's great, but it often happens that people neglect to ask for it).
- AI has the potential to create images that are more accurate to the source material, in cases where the existing ones aren't.
I will elaborate on the third point, but first, let's get the second one out of the way, because I know some people are asking "Why don't you just draw your own images? Or commission someone else to do it?". The answer to that is that not every user has the artistic skills required for the former, or the money required for the latter. So, although this is indeed an option for some, it's not for everyone.
Now, about the third point, any fan of practically any fictional universe that is primarily based on literature will know that sometimes, even images drawn by professional artists are not exactly accurate to how the subject is described in the text. Chaugnar Faugn is not supposed to look like a literal elephant. The Dweller in Darkness is not supposed to have a face.
We most certainly DO want to display these "inaccurate" images, as they still have artistic and historical value. ...However, we also want to display at least one image per page that is compatible with what is being described in the text, if possible.
I say that AI can help with that, because this has already been done on this very wiki. Laban Shrewsbury is a character who has no eyes. His opaque glasses conceal empty eye sockets. The former main image on his page showed visible eyes, so it was moved to the gallery and replaced by an AI image that is closer to the original description.
The problem with AI images[]
Now it's time for the cons. To be honest, it feels awkward to list down the problems people have with AI images. It feels like something everyone has already complained about, including me, but let's go.
- AI images are (currently) unreliable and more often than not generate strange characters with super-deformed hands or nonsensical clothes.
- The use of AI generators is also a form of art theft.
- For the above two reasons, the use of AI images can leave a negative impression on a significant number of visitors to this wiki.
I don't think many people will challenge the first point, but if you want examples, take a look at the images we currently use for Albert Wilmarth or Elihu Whipple. You'll notice that their hands, their eyeglasses, the objects surrounding them, all look like they came straight out of R'lyeh.
Depictions like these might be less of a problem when we deal with abhuman characters like, say, Wilbur Whateley or Richard Upton Pickman... But even then, I'm fairly sure Thurber would have mentioned if Pickman's hands were deformed like that, if nothing else to assure us that this condition has never interfered with his painting abilities!
And since I mentioned Pickman, I suspect he would unleash his Ghoul pack on anyone trying to steal his paintings. As most people know, real world artists aren't too thrilled about art theft either. And what else to call a tool that uses preexisting images to generate new ones?
Defenders of AI claim that what the generator does is no different from a human artist looking at online images for inspiration... This argument fails to convince me. Using art for something that the artist didn't consent to is pretty much the definition of art theft.
In other words, an artist who posts their work online is implicitly consenting to other people looking at their work, potentially for inspiration as well. But they are NOT necessarily consenting to their work being fed to an engine that will use it to make up new images.
Some thoughts about the future[]
No, I don't pretend to know what the future will be like, I'm just a member of a cosmically insignificant species that experiences time linearly. But still, I think it's relevant to ask ourselves: will the issues of AI continue to be issues for long? A few years from now, will AI still be drawing people with six-fingered hands and asymmetrical eyeglasses?
When Smoog first asked my opinion about this whole AI thing, I offered two reasons why I'd be hesitant, even reluctant, to outright ban AI art from the wiki.
- The first reason was that it might not be possible to prove whether an image was created by AI or not.
- The second was that we might consider the possibility that, at some point, publishers might release actual Mythos books and other media using AI imagery.
While I've been having second thoughts about these two points, I still ask myself if we might make a decision that we'll regret or be forced to revert later.
What if we opt to ban AI from this wiki, only to find that a few years from now it will be extremely difficult for a layman to even tell AI images apart from human-made ones? Or perhaps to find that, a few years from now, some artists will be using new combinations of traditional and AI-based techniques to create art that really can't be properly categorised as one or the other? Or to find that, at some point, AI will be widely used in all sorts of media?
But this is indeed all speculation, and it ultimately doesn't address the buopoth in the living room, which is the problem of art theft.
Whether we should continue to allow AI images on this wiki is not a matter for one person, or even just the staff to decide. This is something that should be decided by the community. Which is why I end this blog by inviting you again to go to the poll and vote. The poll will be open for two weeks (it closes on October 17th), so if you need time to reflect about the matter, no worries!