Forum - View topicINTEREST: Shinchosha Publishes AI-Drawn Manga
|
Goto page Previous 1, 2, 3, 4, 5, 6 Note: this is the discussion thread for this article |
| Author | Message | ||
|---|---|---|---|
Cutiebunny
 Posts: 1767 |
|
||
No, it's exactly the same with a different form of media and it's a great analogy. Notes are notes just as pixels are pixels. Men at Work were inspired/tutored to use the Kookaburra rift in their song and make that rift a large part of that one song. They transformed it into something new, which, judging from the description of current AI program with additional human input needed for editing purposes, is exactly what is going on here. And that seems obvious - You can't have a character appear one way in one panel and then, using a different style, completely different in another. There needs to be consistency and, until AI improves to a point where it can create a consistent style, it will need input. Same thing with sampling a song. And, as pointed out in my original post, Men at Work never credited the creator of Kookaburra for using their song until they were sued some 25 years later and were forced to pay royalties on a percentage of those that they received from 2002-present. Are musicians now crediting their sources? For the most part, yes, but again, only after it no longer became acceptable to sample existing songs without credit/royalties. This is what will likely happen once AI not only advances to a point where it can modify anything, consistently, into other styles, regardless of the source but when these modifications can be done easily, by anyone, in mere minutes. Sorry for those of you who went to school to learn how to draw, but this pattern of becoming obsolete is nothing new and one which everyone will need to get used to in one way or another. |
|||
|
|||
Gem-Bug
 Posts: 1307 |
|
||
This is such an unbelievably anti-human, sniveling, sad stance. Congratulations, I guess? Last edited by Gem-Bug on Tue Jan 10, 2023 9:08 am; edited 1 time in total |
|||
|
|||
NeverConvex
 Subscriber Posts: 2511 |
|
||
|
I think part of the problem is that this won't actually make anyone obsolete; ANNs can create impressive, novel pieces, but they can only do that given a sufficiently rich training corpus (and extensive, scientifically discovered training regimen). And, while in principle they could generate anything (being "universal function approximators"), the likelihood that they'll generate something radically distinct in style from any combination of their inputs is probably quite low. They also are clumsy to 'pilot' -- short of someone inventing legitimate artificial intelligence in the strong sense, our ability to tell an ANN what we'd like it to create is very limited (and maybe even with strong AI; describing art ahead of time is very difficult, even if you're telling a human artist what you want).
Given that, it seems likely we will lose something quite fundamental if we let the technological development massively displace practicing artists. I think it's unlikely that the market could anticipate what kinds of creative output we wouldn't get because of largely limiting ourselves to ANNs re-combining existing work, and I doubt the consequences (in loss of artistic output) of that would be understood for decades (if ever; this is a theory-light field). I don't think we want to form a feedback loop like: artists make work -> ANN trained on work -> artists lose jobs -> remaining artists make more work -> ANN trained on new work -> still more artists lose jobs. Eventually there's not really any new work being produced, and ANNs can't substitute for that part of the process. That sounds like a pretty sad thing to let happen, and building new IP that expects meaningful compensation when a person's work is used as a training image in an ANN seems like a really natural way to address it. Separately, it's also just a great deal more tedious and less creatively fulfilling to spend your time making up prompts for an AI generator than to generate art yourself. I think it makes sense to protect satisfying/fulfilling careers from replacement by tedious ones to a significant extent, where possible. (Though, that said, I imagine there will be a smaller but still significant trend of hybrid workflows, in which traditional artists use AI art as another tool for producing projects that an ANN by itself would likely struggle to achieve. In small ways, some of these are already mainstream -- e.g., the 3D art software Blender has a built-in AI-denoising tool, for allowing an ANN to automatically inferring what noisy pixels in your image should actually look like. That's a much less dramatic use than whole image production, but is a good example of AI fitting neatly into existing workflows.) |
|||
|
|||
 |
All times are GMT - 5 Hours |
|
|
|
Powered by phpBB © 2001, 2005 phpBB Group