at some point it's just like. do they even fucking like the thing they're asking AI to make? "oh we'll just use AI for all the scripts" "we'll just use AI for art" "no worries AI can write this book" "oh, AI could easily design this"
like... it's so clear they've never stood in the middle of an art museum and felt like crying, looking at a piece that somehow cuts into your marrow even though the artist and you are separated by space and time. they've never looked at a poem - once, twice, three times - just because the words feel like a fired gun, something too-close, clanging behind your eyes. they've never gotten to the end of the movie and had to arrive, blinking, back into their body, laughing a little because they were holding their breath without realizing.
"oh AI can mimic style" "AI can mimic emotion" "AI can mimic you and your job is almost gone, kid."
... how do i explain to you - you can make AI that does a perfect job of imitating me. you could disseminate it through the entire world and make so much money, using my works and my ideas and my everything.
and i'd still keep writing.
i don't know there's a word for it. in high school, we become aware that the way we feel about our artform is a cliche - it's like breathing. over and over, artists all feel the same thing. "i write because i need to" and "my music is how i speak" and "i make art because it's either that or i stop existing." it is such a common experience, the violence and immediacy we mean behind it is like breathing to me - comes out like a useless understatement. it's a cliche because we all feel it, not because the experience isn't actually persistent. so many of us have this ... fluttering urgency behind our ribs.
i'm not doing it for the money. for a star on the ground in some city i've never visited. i am doing it because when i was seven i started taking notebooks with me on walks. i am doing it because in second grade i wrote a poem and stood up in front of my whole class to read it out while i shook with nerves. i am doing it because i spent high school scribbling all my feelings down. i am doing it for the 16 year old me and the 18 year old me and the today-me, how we can never put the pen down. you can take me down to a subatomic layer, eviscerate me - and never find the source of it; it is of me. when i was 19 i named this blog inkskinned because i was dramatic and lonely and it felt like the only thing that was actually permanently-true about me was that this is what is inside of me, that the words come up over everything, coat everything, bloom their little twilight arias into every nook and corner and alley
"we're gonna replace you". that is okay. you think that i am writing to fill a space. that someone said JOB OPENING: Writer Needed, and i wrote to answer. you think one raindrop replaces another, and i think they're both just falling. you think art has a place, that is simply arrives on walls when it is needed, that is only ever on demand, perfect, easily requested. you see "audience spending" and "marketability" and "multi-line merch opportunity"
and i see a kid drowning. i am writing to make her a boat. i am writing because what used to be a river raft has long become a fully-rigged ship. i am writing because you can fucking rip this out of my cold dead clammy hands and i will still come back as a ghost and i will still be penning poems about it.
it isn't even love. the word we use the most i think is "passion". devotion, obsession, necessity. my favorite little fact about the magic of artists - "abracadabra" means i create as i speak. we make because it sluices out of us. because we look down and our hands are somehow already busy. because it was the first thing we knew and it is our backbone and heartbreak and everything. because we have given up well-paying jobs and a "real life" and the approval of our parents. we create because - the cliche again. it's like breathing. we create because we must.
you create because you're greedy.
18K notes
·
View notes
Thinking about image model generated art and gifmaking is giving me some weird vibrations about how there really is some weird association of the virtuesvirtues of a medium with the virtues of the people working in it. Gifmaking being associated with KPop fans doesn't make the concept of frame interpolation racist, and someone marketing themselves as a cheaper alternative to some other artist doesn't make the concept of generative art inherently class antagonistic.
It's somehow reminiscent of CJ the X's distinction between "stupid art" and "evil art", how a medium that has a low skill floor can produce things that are very stupid and easy to perceive as low-effort but how that's not the same as them having something wrong with them. If you look at my animation tag, most of it is motion graphics done with AfterEffects, and while it's probably wrong to call it a low skill floor program the way an AI art generator is... there is still a world where instead of programmatically telling shapes to whizz by on a screen, a different Van would have drawn those same animations frame by frame, producing exactly the same animation.
And I don't think the fact that I did them programmatically somehow invalidates the artistic intent that went into them, y'know? I could open AE right now and produce a 250x250 looping gif of clouds and while I know how to do that quick, to make it look good and to make me like it, I would have to spend time considering how the various elements, colours, timings and whatever the particle system/noise generator I use spits out fit together. I would have to fiddle with seeds and levels and timings to make it look good. I would have to spend a long time just staring and thinking about what I'm making before I could make it good.
I don't know enough about generative art tools to know how much fiddling goes into them once they're taught and ready to go, but I do know enough about deep learning to know it's a haphazard, frustrating process that you as the artist have only limited control over, which is why it doesn't appeal to me. But I have made gifs in the past, and I know how that process requires an eye for consistency and composition, framing and colour that a lot of other visual artists don't have because they're not working with time as one of the creative dimensions.
And like... who am I, from my high horse as someone in possession of these skills, to tell someone who is still developing these skills or who has a different aesthetic concept of what is good than me, what they're making is low-effort. That's not my judgement to make. I didn't make it. Only the artist themselves can say if somehing was low-effort or not. I don't see why I should have so little faith in other artists to assume they have no interest in putting in any effort.
6 notes
·
View notes
Sorry to bother you but I remember you making a tutorial on how to make your own ai with beta character ai? I can't seem to find it though. I'm I just imagining things?
You're not, I did make one, however...
Honestly?? I wouldn't even bother at this point. Character.ai has had so much drama with the devs being horrible human beings: most likely stealing/selling user data (you want to delete a bot off their website?? Too bad. You can't. Also they took a bunch of bots with high interaction counts and locked the creators out of editing, essentially stealing them from people), not listening to feedback whatsoever (actively making things much worse with every update), implementing a "nsfw" filter that blocks even the most sfw things like hugging or kissing (even with family/platonic characters), has a nasty habit of comparing anything romantic to children ("she looks at you like a daughter looks to her father" like ew what the actual fuck, and this happens literally all the fuckin time so you gotta ask how it made that association hmm), dumbing down the ai in favor of making it "safer" (now characters hardly ever act like their source material, often making up random shit), bots ignoring user input (often going on a tangent describing the environment and ignoring dialogue), bots being stupid in general (forgetting what you said literally 2-3 messages ago and giving horribly lackluster responses), devs shadowbanning people/deleting any posts respectfully calling out their shit to facilitate propaganda in their favor and make it seem like nothing's wrong, and so much more.
Drop it like a newborn giraffe.
Some other options: Pygmalion, which I'm still learning how to use and it ofc isn't going to be as advanced as Character.ai since it's a 6b model while the latter is like 100+b, meaning less data power so responses are less intelligent + slower, BUT it is free, it is open source, and you can get some pretty decent responses if you hold its hand a lot. There are some UIs you can use it through, like forks of TavernAI that give you the ability to use world info and get it to have a better memory, one for group chatting with different characters (tho I haven't tried that), a customized background, and so on. The devs are still working on a site but you can access it right now- I can make a smol tutorial for the basics of that if anyone wants it.
OpenAI is also an option, you can also use it through Tavern and it will give responses akin to if not much better than Character.ai- though you gotta pay. HOWEVER you can get a free trial with $5 to use... and there are ways of using it free after that, but I can't say it here cause if it's too widespread, it'll prolly get patched out, and we don't want that.
How to access Pygmalion - Go here and click this: (when it asks u for permission to access ur google acct, click yes)
It'll take a while to load, maybe like 5 mins or something,
Once you get to this, click the UI links (usually only one will work, use that)
Now, that will take u to the KoboldAI UI. Wanna use Tavern instead? (good choice, gives a lot more customization) Use this link.
Now scroll down till u get here to this drop down menu. Choose Pygmalion 6b or 6b Dev (to be quite honest I'm not sure what the difference is yet). Wait for that to load, and while you're waiting, go here to download the Tavern files:
boom (or some other fork)
Go here and click download ZIP.
Unzip and look for this:
Ignore everything else, just click this. (the file's safe, if windows gives you a warning, ignore it) Now you'll be here (I'm using a different fork so it may look different:)
click the 3 lil lines in the upper right corner to access settings and characters.
Once you've added/edited characters as you like (I suggest looking at the default ones if you're curious about how that works) you can come back over to the google drive and, if it's finished loading, you'll have two url links like before:
Paste one into the API key and click connect. I used the top one but the bottom might only work for you.
There ya go! A bit complicated, but that's a free substitute for character.ai.
If you instead wanna use OpenAi, (assuming you're either paying or using the free trial, or... something else ;), you'd go here on ur openai account page and go down here:
and create a secret key. Copy that and go into the UI of your choice... ima use TavernAI again. Go into settings, change your API to Openai, and paste that API key you just copied into the API link. Voila!
Here's a couple helpful links for further info: https://www.reddit.com/r/PygmalionAI/ (they have a discord, with some really helpful guides)
character creation/writing tips for Pyg
Soft prompts
General tips
I'm still a noob who knows jackshit about coding, so I can't offer much help, BUT I know the basics so ye. Hope this helps!
14 notes
·
View notes
it occurs to me that part of this "chatgpt will make up nonexistent articles and quotes and whatnot and then attribute them to real people" issue - I suspect it has something to do with "referencing and citing Actual Things In the World where appropriate" and "not plagiarizing existing writing" being two goals at cross-purposes for training a program on like, general writing I guess? - when it doesn't already have a background of understanding the difference between "real" and "fake".
like, you have something that understands "here are strings of text! I know a lot about how parts of these strings tend to get put together to form larger blocks of text. I also know that it's important to make sure I don't arrange those blocks or strings of text in ways that directly match existing blocks of highly unique text in my dataset."
so like... it completely makes sense why it'd have trouble contextually with stuff like "citing real scientific article titles"? unless you put it through a much more rigorous training designed specifically to learn what (contextually) makes something an "article title" as opposed to other text, and what distinguishes "citing something [e.g. putting an exact copy of existing unique text into what it generates]" from "plagiarizing".... that's probably not a task the computer is equipped to handle.
12 notes
·
View notes