MirceaKitsune
patreon
The AI hoax may be the latest attack on artists
🕑 Added 2022-11-21 15:24:48 +0000 UTCSince this has been popping up on our art sites it needs to be spoken of. Humanity's latest coping mechanism and attack on reality, known as AI, has been hoaxed into the world of artists and in the process of being infiltrated into our communities. I'm officially warning everyone to beware of either delusions or deliberate psychological experiments, involving people claiming that computer code can literally generate original art out of thin air. Incredibly enough we're now at this stage in the devolution of humanity into madness.
I was already done with the world and this further cements my complete detachment. I spent the past years explaining to a society brainwashed by Star Wars that conventional computer code cannot be made to generate art or recognize content from pixels or think like a person, those beliefs are on par with witchcraft and go against basic laws of physics logic and mathematics; There's no writable code formula for it, it doesn't matter if you call it machine learning nor how many resources you have, binary code has no way of approaching such complexity by design... you'd need a physical conscious brain which involved quantum mechanics and more, that may be possible someday but until then you can't botch the process using existing hardware. To anyone with basic knowledge of how computers work, that stuff is worthy of ushering someone to a psychiatric checkup... not for today's responsible adults who run giant corporations and entire nations, those can afford to live in a parallel existence we aren't even allowed to contradict (it's dangerous misinformation) and I'm sure nothing will ever go wrong in that.
Thus welcome to the latest escalation: Instead of accepting reality and its core limitations, people are now applying oil paint filters to their art and claiming it was generated by men made out of C++ code... they even created websites like that Dal-E for automating the process using keywords to mix sprite patterns. Of course that wasn't enough: Now they take over art sites to impose their epic story onto the rest of us, using the classic game of good cop / bad cop as the usual tactic to introduce every latest thing: First the protagonists come with their fictional danger or technology rabidly claiming its existence while shutting down anyone who gets in their way, then everyone else needs to make statements or enact rules regulating that thing as if it were real. We're now at the later stage where art sites are announcing policies on "AI generated art" ignoring the fact that, you know... such a thing is not physically possible!
Is it worth the effort humanity? Going to such length for what's literally a childish fantasy taken from your favorite 90's films? All for a coping mechanism to deny your reality and limitations? Are you managing to live in your favorite video game and escape reality yet? Because I hear the climate will soon go haywire and there's some war involving nuclear powers while a global economic depression is happily on its way: I'm not sure if the man trapped in the body of a computer can save you from that and take you over to binary land, you might need to close your eyes really hard and port the mechanics from Tron next.
Some people said it well: We live at a time where reality no longer exists at a fundamental level. Even things that physically and logically cannot exist are being simulated as the masses are playing along, without anyone of common sense being able to get in the way any longer, all to appease scammers or people who can't cope with the limitations of their own existence. What even is reality any more then? Why should I pretend I live in a real world, instead of acting like I'm some superhero character who can do whatever I want regardless even of the very laws of physics? Maybe I should take their lesson and abandon this boring mundane existence in favor of a fantasy that isn't there yet... after all, if I really want it, it must be true right now somehow! We're at the point where one can legitimately ask those questions.
Comments
Mircea Kitsune
What I concluded they're likely doing is they have a large database of sprites, either 2D and / or 3D resources meant to work together and be blended in a certain way: By using keywords they can automatically generate and colorize an unique character or object. The core assets must have been done manually of course, it's not a mind in the computer that magically learned to envision them in existence; Them not admitting this and saying what they're actually doing is what upsets me... the technology for such a thing requires no AI and is something I myself thought of doing if I had the time and knowledge to code it, basically using NPC's in a game engine with random objective to render an unique movie which would be an awesome technique to have.
HSTL
Well, having spent quite a few hours trying out the NovelAI's Image Generation, I kinda agree with you. AI can't really generate original art out of thin air. But within the limited parameters it has been trained with, the NAI is pretty damn good at pretending it can really make art. Ask it to give you a feral wolf with green fur, red eyes, buckteeth and dick but no testicles, sitting in front of a barn with trash around it and you'll end up with... Likely a wolf with five legs, red eyes, fur that's a mix of green, red and white, a somewhat twisted face that does have buckteeth, and a dick which may have balls or not, sitting or leaning on a building of some sort, and likely having flowers on the grass around it. But with some tweaking, like banning flowers, and generating a dozen images, you'll likely end up with one or two good ones. Unfortunately canid's of different sorts are the only species the NAI's Furry Beta 1.3 is really good with so far, dragons, bunnies, horses, goats and few others work well enough some of the time, but others like badgers, various pokemon species and aquatic creatures tend to be a disaster. And many, many other species, items, poses, etc. the AI has no idea about at all. For it is as you said, it can't create anything truly original at all. From what I understand, to train diffusion AI's you need at least 20 images of the thing you want it to learn about, from different angles and from different poses, to have any chance of the AI being able to make use of it. So if someone want's an AI to "paint" images of their OC, or an original species, well, better have a thick stack of references to feed to the AI. Also, NAI users can't submit material for the AI training yet, nor is it even being planned for now. Might have something to do with how time consuming the training is even for heavy duty hardware. As for the rest of the rant, well, escapism is what people tend to do precisely when things suck. I know I can't handle more than four or five hours of direct contact with reality a day at a steady basis these days, after that I either retreat in a protective cocoon where things are better or cry myself to sleep. Unfortunately that is reality too.
