[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/ic/ - Artwork/Critique

Search:


View post   

>> No.7001671 [View]
File: 370 KB, 1312x1226, 1691959090398359.jpg [View same] [iqdb] [saucenao] [google]
7001671

>>7001650
i just mean it in the sense that i won't put as much effort as i used to.
to me, it really only comes down to understanding a few things, and if you don't get those, you won't get it.
and as it happens, most people refuse to understand it even when explained to.

the best course of action is just showing you exactly how AI can be used and what i mean by all this. and then you can ask your lying eyes whether i'm telling the truth or not.


>>7001645
again, it is not recreating the training data.
do you know what "curve fitting" actually means? it means that it is trying to create something that statistically falls into place AMONG the existing data.

once again, your language suggests that you think that AI is copying shit.
how do you reconcile that point of view with the reality that AI is creating images that are not and cannot be copies of existing art?
contextual understanding of lkighting, reflections colors, none of these are static things that can be just copied. in your mind, how is this supposed to work?

>> No.6956263 [View]
File: 370 KB, 1312x1226, 1670675469785713.jpg [View same] [iqdb] [saucenao] [google]
6956263

>If it understood what a cat was, it would know there is no chance for it to be two-headed or have two tails, as it's not a trait that IRL cats have.
you say "if it understood a cat" but realize that the image AI has a limited, VISUAL understanding only. you're saying that because it makes mistakes, there can be no understanding. and that is a simplistic, black and white view of what is actually happening.

take the example of a cat's surface features.
>it's general shape, texture, elements like ear, nose eye and the proportions they are arranged in.
and now take in its deeper unseen features
>its skeletal structure, anatomy the resulting range of motion.
the former is visually easy to grasp, the latter is not. it would take much more data to see the patterns that tell it about the DETAILED anatomy of cats.
all this is to say that just because it makes errors, it doesn't imply non-understanding.

the other example i always use is reflections. SD's reflections are rarely accurate, but the fact that it always gives certain surfaces reflections at all is already a sign of generalized understanding.

cont.

Navigation
View posts[+24][+48][+96]