cross-posted from: https://awful.systems/post/5962853
science shows as true what you thought was only 99% true
https://www.youtube.com/watch?v=uVf7VUX_iUk&list=UU9rJrMVgcXTfa8xuMnbhAEA - video
https://pivottoai.libsyn.com/20251015-ai-is-not-popular-and-ai-users-are-unpleasant-asshats - podcasttime: 5 min 57 sec
I’m being required to use AI at work, and the first thing that jumped out at me was how often it says, “You’re right!” That is like a drug for a narcissist.
Fire your boss who made AI a requirement.
The only use for AI is generation of trash. What’s your job?
Cockatiel looking into a mirror.
It’s so clearly not a finished or functioning product. In terms of art, I tried to use it as a tool, and it’s useless. You can’t make AI assisted videos or art, because it can’t produce duplicates of anything. It can’t stay on model, it can’t use a set design to make anything. You can’t plug in frames and ask for inbetweens, you can’t give it a script and get comic panels, none of that works. You get random sloppy bullshit.
When the goal is getting people to use it to jack up company investments then getting people to use it, even if forced, does mean it is finished and functioning in the context of the goals.
It certainly isn’t functional for what it is advertised as though, as you pointed out.
Everything you mentioned can be done. I feel like a lot of the “AI can’t do anything” boils down to “I can’t figure out how to get it to do this”. Not that it’s easy or anything, comfy-ui is kind of difficult to get a hang of for example.
show me four panels made by AI that contain one character who’s design doesn’t change in any way in a location that doesn’t change in any way, in any art style it doesn’t randomly change. I know you will, with in a few hours if you understand the software, be able to produce these images and you may THINK they will be similar enough. go ahead and show me, I’ll explain to you why it’s not useable at all.
I have an entire fucking degree in computer animation. I’m not rubbing sticks together and going “boo hoo this is hard” I was trying to used AI at any point in existing and established production pipe lines. IT CAN’T DO IT. It’s a randomizer. it CAN NOT REPODUCE AN IMAGE OF A CHARATER THE SAME WAY MORE THAN ONCE. I don’t need almost the same images if you don’t look closely. I need EVERYTHING ON MODEL or it breaks the animation illusion. I’m not talking about the shirt changing styles. I mean if you draw round eyes it will give you an inbetween with oval eyes. which breaks the animation. yes. you can get “video” from AI as long as you need a video that’s less than 10 seconds long and isn’t required to match up to any other existing video. because it CANT DO IT!!! it randomizes everything every time. you can build models for it to work inside of that will MAYBE allow you to get random images that are all in the same general style. but the software can’t handle any higher level of specificity or accurate reproduction. I can write prompts down to the post production style, the lighting set up, the textures and dirt maps, I know how to tell it to give me an image with incredible specificity. it STILL RANDOMIZES EVERYTHING EVERYTIME. It’s not a tool for art work, it’s a random art generator that only works if you want one single image one time because it can’t duplicate it’s own results. if you want the same image from 10 camera angles. you’re going to be disappointed.
I have nothing against you personally, it’s just really not able to do real art work. it can’t. it takes a bunch of random images and it gives you a random output based on those images. it’s not functioning beyond that. it’s not capable. if you want to use the artwork to make a comic, or animation, it can’t handle the most basic of tasks needed to do so. because that means on model. matching designs and locations. it can’t. it’s a gimmick. no one is trying to use it in production at all anywhere I’m aware of in the actual animation world. it can maybe give you design ideas, because that task is ment to be variations on a theme and not reproducible results. but that’s the end of it’s usefulness, and even then it isn’t useful because design is the hardest job in the game because you CAN’T just keep pumping out knock off’s of other peoples work. you have to have a style that works for you and the production as a whole and a computer can’t figure that out. it’s just random image generation. that’s all it can do.
The parallel between how little MBA-types understand about processes and how incapable their AI pet projects are of performing said processes seems so obvious in hindsight.
That’s adding a lot of constraint to the simple list you had in your first comment. It can do all that, just not perfectly like you start asking for in the reply.
What I find crazy is that if you’re an animator, why are you paralysed when it comes to fixing mistakes? There’s a spot between it can’t do anything and it does everything for you, and that’s where it is right now.
It’s a bit how people say it can’t code. “It makes mistakes and then I have to fix them”. That’s part of the task! It aounds like when someone types in something into to the prompt box, they feel like they lose all ability to modify it after.
And ya, it can’t code a huge app all on its own in one go, but just like how it can animate a scene here and add some special effects there, it can code a function here and a helper script there, etc.
I feel like you are wrapping it up in your own expectations and ignoring the things it can do because of that.
I don’t mess with the animation part all that much, but I have been using Sora and I’m also part of a discord that talks about it a lot, I see what people are doing with the open source tools and it’s a lot farther along then it’s made out to be in this sub.
That’s adding a lot of constraint to the simple list you had in your first comment.
They explained the actual task in more detail, because you clearly don’t understand it. When someone says that image generators can produce animation that doesn’t mean ‘spitting out something that looks like shit that actual humans have to spend time cleaning up,’ and you know they’re not sold or advertised that way. It’s a useless additional step that jams up work flows of the actual artists because it’s giving you shit to fix with zero thought put into it. Human animators might consistently give you ugly shit with zero thought put into it that other people have to clean up as well, but they tend to lose their jobs.
I feel like you are wrapping it up in your own expectations and ignoring the things it can do because of that.
No, gen-AI pushers are lying about what it can do. Animation is not ‘produce a thing that looks okay if you don’t look at it too hard for 3 seconds,’ it’s the entire process. If a human being was hired with the expectation that they could animate and they were completely unable to link scenes together coherently, change perspectives, stay on model, and instead all they could do was short scenes with lens flare or some shit that other people still had to touch up, their ass would get fired. I’m not aware of any animators–as in, people paid to animate for their job, not people who play with AI to make 5 second generated videos–who actually want this in their job. It’s like telling someone they can only do their animation with a really shitty photocopier instead of pencil and paper.
I wasn’t talking about end to end animation and neither was he in the first comment.
Animations are complex multi step processes. I’m mainly saying AI can help and outright do some of these steps. It’s not currently perfect and I don’t think anyone should be forced to use this but I do think it will become common.
Each individual step is part of that whole process, and throwing trash into that process at any point renders the end product as trash. There are no steps it can do without humans wasting their time fixing its output when they could have just had a real artist do it right the first time instead.
Revisions are a huge time sink in animation, and this is basically a problem generating machine. It’s not a fixable part of gen AI, because it doesn’t think and can’t make decisions like an artist can. It’s only going to give you a probabilistic image based on your input. That is basically the complete opposite of what animators need.
This stupid ass bubble of gen AI getting shoved where it’s unwanted and useless is long overdue to pop.
Show me any image generator produced—and I mean produced as in no human input except text, not humans touching it up and editing the output—animation that doesn’t look like shit and I’ll consider believing it. Show me someone doing it on camera, stating their intended vision and specific scenes/shots and character designs that they want before they start, and they’re able to come up with something that matches what they set out to do and doesn’t look like shit or have absurd mistakes in it, and then I’ll believe it.
Not once have I seen anything suggesting the people using image generators are engaging in anything except constantly readjusting their expectations based on whatever slop pops out rather than seriously engaging with anything being generated on an artistic level, or having a set vision they want to create.
It’s also easier to just program a computer to do these things if you have the knowledge base, in most cases.
We need to recognize what “AI” or at least what is colloquially referred to as AI is: advanced predictive text using machine-learning.
It’s a Large Language Model and therefore it excels at language.
I’ve been having a blast just making random shit with it.
Edit: ComfyUI that is…
That cyborg vs synthetic battle was decent but I had my hopes higher after the build up
I’m so pissed that they had every plot line ready to wrap up but then just fucking didn’t.
It’s so frustrating because any device to keep these plots going is just contrivance.
I’m so incredibly disappointed that they decided against an anthology and are continuing the Marcy plot.
If there’s a season 2 I doubt it’ll be good, I’m super disappointed.
I use a local deepseek-r1 instance for simple code refactorings and boilerplate (e.g. to generate obvious tests). Am I an asshat now?
That depends… do you say obnoxious stuff like “AI let me 10x my productivity” or “You just need to right prompt, bro” or “AI will be good as soon as we build another trillion dollars in data centers”?
Nah, actually I kind of hate most genAI because it’s mostly a copypaste machine that steals other peoples code and art and stuff. But I do use it still because it speeds up menial tasks a lot and I don’t feel like I’m stealing shit from others that way.
i think it would do all of us a service to separate between generative AI and LLM assisted prompting.
Keep the human element. I’ve seen some use cases where AI can help solve a human problem with a human need. I usually see slop intended to boost other numbers, like social media engagement stats and SEO (need 3 pages of content for your spaghetti recipe? Good news! AI can help, and it can make the recipe too. Now one robot pleases another and the human need is an afterthought)
According to this sub? Yup, most definitely! According to my own opinion? As long as you use it responsibly and limited to the scope of LLM (scope, as in: whatever one would usually use a Random Text Generator for, aka. inspiration, non-critical parsing, and mass spam attacks), then go ahead.
This sub appears to be barking up the wrong tree most of the time anyway (just look at the name), so I wouldn’t value anything that is posted here too high, ironically.
I use chat5-reasoning in cursor for code generation. I am also an asshat. But it’s not because I use ai.
That’s the point the article made.
The most consistent predictors of AI use across studies were aversive personality traits (e.g., Machiavellianism, narcissism, psychopathy).
That’s a gross over generalization but yes.
I mean, I agree, but the anti-AI guys are also asshats.
