I am not debating that it's more than just upscaling the picture, but it is still inferring graphical details that do not exist in the original picture based on machine learning. It's still not fundamentally different. The upscaling cannot guarantee that all the pixels look how the original artist would have done it if he could have done it at that resolution and frame rate.
And mind you even there, we already have to realize that the artist doesn't actually have a specific intent for every single frame and camera angle that might be possible, unless perhaps there is a specific cut scene with a fixed camera.
In the picture of Grace, DLSS5 isn't adding some random make-up out of nowhere, it's taking the make-up that is there and infers a more realistic-looking face with that make-up. The kind of face that the artist could only get with a tool like this - or by making the scene render unacceptably slow, which also wouldn't be their intent.
It's adding makeup because its database has people who would have makeup in this way and have a nose job in this way. It's taking what it can glean from places like Instagram and others and putting that effect on there. But that doesn't necessarily jive with the character of Grace, nor any of the characters it makes look like photo influencers. It changes the situation and the image just like interpreting all lighting as being blasted out. It's a face an artist could only get with a tool... but the artist would also have to
want that face, and we've seen a lot of devs bash at it using this sort of homologized ideas of what scenes and people look like to ruin them.
It cannot strip intent just as a loose hair in a painter's brush isn't stripping their intent - even if it leads to an extra line or a gap. If the painter is still happy with the result, the intent is achieved.
It can absolutely strip intent and plenty of the devs have outright complained as such. It modifies faces, models, lighting, all sorts of things. It's not a limitation of the artist, but something active that they have no control over once its on. To remove such things is to strip away the presentation, the intended idea or effect. You can try and dance all you want around it, but such massive changes it makes absolutely dilute what the artist intended, especially given what we've seen so far.
But there is no one else doing the actual work. It's just me, using a tool. In the relationship between a commissioner and artist, there are two seperate humans at work. That isn't true when I use AI, there isn't another person. There is just an AI tool. The AI tool has, as you say yourself, no intent. How could that strip away my intent? When I draw with my pencil, I don't really intent to spread graphite on a paper, but that's what is happening (I think, roughly, not really an expert on this matter.) But what I want is to create a particular imagine I have in my mind. But the tool is spreading graphite on paper, without any intent on its own, just as the physics happen to work out, and I use it until I am satisfied (or give up, to be honest, I am not an artist of any quality or skill.) Still somehow, art can be made with pencils. But AI is a special type of tool where this stops being possible to some metaphysical quality? I don't believe so.
There's not a person doing the work, but a machine attempting a facsimile of it. Like, I don't see how this is exactly hard to get: it's the same relationship between a commissioner and an artist, except that it lacks the thought of an artist and instead tries to recreate it through algorithmic data sets. You only have limited control of what this level of generative AI can do, and given that their idea of how to control it is largely by setting how transparent it is really keys in on how
little control there is over the tool itself. It's turn on, turn off, and either make it more or less difficult to see. There's no control over the actual result, no thought to direct it.
Also, the "artist intent" is kinda overrated or misplaced in games sometimes. Sure, games are also art. But it might not really matter to the "artistic intent" behind if if the characters were generated by AI - instead, the real focus was on telling the story they wanted to tell, and the exact looks of the characters and sets might never have been as important as the characterization. Or maybe it wasn't even that, maybe it where the specific mechanics and type of gameplay the game achieved, and in theory, it didn't matter if the game had a comic look or photo-realistic graphic or not, but they went with the eye-candy variant in hopes of getting more people excited for the game, and if DLSS 5 gets you there, it's the tool of choice. The "artistic intent" for a game might actually not be impacted at all by whether a character is wearing more or less make-up than the original model that it was based on.
I feel like the saying "maybe the looks weren't important" kind of proves my point: it lacks intent. Artists will create things all the time and while they won't have the most compelling reason for certain choices, they do
make them, even unconsciously. Just leaving such things up to AI because you don't care kind of strips that whole aspect away from the characters, doesn't it? It removes a whole idea and aspect that could have mattered, but now lacks something because we let something generate a random character and we wrote stuff for whatever it outputted. Now whatever is
written will have intent, but not the design of t he characters themselves, and I think you'd basically have to concede that with this sort of argument.
Like, you're basically say "Yeah, it strips away some of the artist's intent, but what if they didn't want that part?" and that's conceding my idea. And that's fine, but the second half of the argument I would say is that it seems like a whole lot of people in videogame land care about that and we shouldn't suddenly act like it doesn't matter.
I would cut out these metaphysics of "artistic intent" and the line where some tool can strip it out or not.
That's cool for you, but it seems like the people who actually make the games care about this sort of thing.
The problem with AI here isn't that you can't create art with it. It's that you might be creating it unethically, because for training its model, intellectual property was stolen by big corporation from original artists. In using it, you might also risk the livelihoods of people that used to create art (quite possibly the ones that was stolen from in the first place). That I can all understand.
It's kinda as if you were painting with blood. Sure you can create art with that. But if you steal the blood from blood banks (or even more nefarious means), less blood is available for patients and you violated people's constent, you're still doing something that is bad for socienty and we probably should not support your blood paintings, even if they look gorgeous and evoke strong emotions. I can get that.
I would say that you can't create art with it
and it's incredibly unethical. You are removing yourself from the actual creation of things, and instead just using a plagiarism machine that attempts to recreate art without really knowing what art is. Creating art with blood has meaning, intent, thought behind it; AI
doesn't. It's imitation, mimicry, the styling of something without the thought of it. It puts things in because it is weighing how often it sees something. That can be passable sometimes, but it's not really art and I don't think we should even humor the people who
want it to be art with such a concession.
