Originally published at: System Shock publisher posts AI-generated art, faces backlash | Boing Boing
…
2 Likes
But this was never about using AI to create artwork instead of using real people. This was about using AI to imagine what AI imagines itself to look like
Come on, we’ve been over this. AI doesn’t “imagine itself” to be anything. If you ask it to depict itself it just pulls together ideas from what real humans have imagined it to look like. Including prior System Shock art, I’m sure.
7 Likes
If an illustrator can only make a living via community backlash against the machines, I’m afraid that their commercial prospects are quite limited. It didn’t work for cameras, for steam shovels, for lawn mowers or for spellcheck either.
We’ve seen this movie many times before. The only useful path is to get the machines to work for you, not to insist that the machines stop being used.
4 Likes
Replace the word “illustrator” in this sentence with “any job” and you’re almost to the point.
Ain’t that capitalism in a nutshell. The problem is that the machines aren’t going to work for you and me. They’re going to work for the people who had money to begin with. And they’re going to have a lot more money now. You and me? Not so much.
5 Likes
I think people would be less upset if this System Shock remake hadn’t already been delayed 7 times. These shenanigans are just the icing on the cake for a game that was originally supposed to be released over 5 and a half years ago. It’s very likely given the game’s history that the only thing coming out later this month will be the announcement that it will be coming out in November, and that that will also not be true.
5 Likes
at least currently, for an something like this, you still need an illustrator.
like is it possible that the image is the raw output of an algorithm, selected at random by the person who does their twitter feed? yeah, i suppose so. but much much more likely is that an illustrator themselves generated a bunch of images, worked with other people to select a good one, cleaned it up, and then handed it over.
i’d bet there are probably studios producing internal concept art using ai, but also people sometimes use reference art from other games, movies, etc. for that purpose too. so i’m not sure it’s at the level of taking jobs… yet.
2 Likes
I can’t get upset about this, as it’s pretty much the only acceptable use for this kind of generated imagery, as far as I’m concerned - doing it for a lark, where the whole point is that it’s a contextually appropriate novelty and no actual illustrators are being replaced.
There’s definitely a layer of “let’s pretend” in this case, though (unlike so many cases of “AI art” use, where people have simply deluded themselves).
Given the cultural significance of Shodan in fan art and illustrator portfolios, I wouldn’t be surprised if that wasn’t the primary source for the training art, even assuming they used generic keywords - but it’s also extremely possible they straight-out used “Shodan” as a keyword and it’s 100% based on existing fan art.
Not exactly the same thing - people didn’t use cameras to replace illustrators by taking pictures of existing illustrations and saying, “Oh this isn’t your illustration, it’s an entirely different photograph! I don’t owe you nothing!”
Oh, it definitely is. The Chinese game industry, for example, has lost a significant percentage of illustration jobs to “AI” in the last year (70%, by some counts). That the output isn’t great (and there’s little control over it) doesn’t matter, because there’s not a lot of concern for quality. Remaining workers have ended up with far crappier jobs cleaning up generated images, and working even more hours (because they feel like they have to, to keep their jobs).
Of course, this works particularly well in the game industry (and especially the Chinese game industry), because there’s a whole sector that’s hungry for lots of images in highly conventional styles (for which there’s a lot of prior art on which to train the systems). Doing more original work (or even work in less popular styles) is currently safe, as is most game asset work (modeling, texturing, even creating 2D sprites). But some of that is soon to be infringed upon by automated systems. (And there’s some interesting applications of “AI”-assisted tools that could help alleviate certain types of drudgework for artists, but the attention - and money - isn’t focused there.)
2 Likes
6 Likes
It’s good for single images but if you want accurate concept images from different angles, usable storyboards or consistent animation, “AI” still sucks ass.
It’s like that weird kid in class that could do that one amazing character but only that drawing, over and over.
4 Likes
Aside from the ‘dispatches from late capitalism’ angle; I can’t say that them posting a vaporwave spriggan and calling it SHODAN fills me with confidence in their artistic direction.
The bots’ ability to go all surimi on the training corpus and produce a content-homogenate that can be molded and dyed to suit a wide variety of purposes is not to be underestimated; but it’s not a great fit for characters that are supposed to be iconic and distinctive.
3 Likes
Oh absolutely. Great for this tweet or a BB comment, but actual prod? Still seems quite a ways off. And as @fuzzyfungus mentioned neither theirs or mine looks the slightest bit like SHODAN, IMHO.
2 Likes
Band Name!!
1 Like
It wasn’t the first time and I don’t think it will be the last.
That seems like a shockingly foolish unforced error on their part.
When you’re in the business of being against things that the perpetrators are quick to insist didn’t actually happen doing anything that makes the “who are you going to believe: me or your lying eyes?” a more credible position seems like a terrible mistake.
On the game studio side it’s more of a dispute between labor and management among fictionmongers over how the fiction should be produced; so at least there aren’t really any epistemic stakes. That’s…less so…when your work includes a lot of having to insist that those people who were disappeared are in fact not alive and well and that that bad thing that was done was in fact committed; which seems like a bad place to get flexible about the distinction between actual and plausible photos.
4 Likes
In the case of this kind of AI (which really isn’t but that’s a whole other conversation) the work that the AI is doing has to have the work of actual humans as input, and the people who run the AI aren’t compensating the humans whose work they are using to train the AI. At this point, that’s probably not illegal or even copyright infringement, but it’s certainly ethically questionable. And even if it isn’t, it’s not a sustainable model. These AI’s aren’t coming up with anything original. They aren’t doing what human artists do when they create new art. Humans, of course, also use existing art as inspiration and for ideas and as a springboard to new art. But AI isn’t doing that. It’s literally just approximating existing art. Because it isn’t actually sentient, it can’t add anything new. So…if this AI puts the human artists out of business, the AI art will stagnate at that point. You are probably correct that the machines aren’t going anywhere. But we all, collectively, need to find a way to make this work in a way that allows human artists to keep doing what they do and be compensated fairly for it.
2 Likes
Uh oh. Soon we’re gonna need a list like we had for Duke Nukem Forever. It was a great running gag because the studio had endless excuses for how making a great game takes time and it isn’t their fault blah blah. Oh really? More time than these dozen entire franchises? More time than space ships and aircraft carriers? 
The game did eventually ship, but for those who haven’t seen it, here is the classic “Things That Have Happened Since Duke Nukem Forever Was Announced”:
https://gamefaqs.gamespot.com/boards/448848-duke-nukem-forever/50079522
My favourite:
The two Mars rovers, Spirit and Opportunity were proposed, authorized, announced, designed, launched and successfully landed upon Mars where they have been exploring the surface for over 2.5 years.
1 Like
I’m starting to think that we’re going to have Citadel ISS-2 run by SHODAN GPT-10 first.
1 Like
Oh, I remember those days. I wasn’t even anticipating the game but it was still a well-known joke that everybody seemed to be aware of. System Shock has a ways to go before it reaches that level… but it’s also a remake of a game that already exists, so that’s a whole other layer of pathetic. On the one hand, that makes the delays even more egregious, on the other, hopefully it will prevent the hype from becoming unreasonable, because no matter how good hopefuls expect it to be, they’ll still be expecting System Shock with updated graphics, at most.
1 Like
The DNF debacle was extra funny because they were trying so hard to make it sound like they were a studio actively building the game all that time, when we all knew it was at least 10 years of that one dude sitting on the IP doing nothing and trying to get funding.
