“you cannot resist the machines, protest is futile”
Okay Skynet lmao
2 Likes
Then when a human being publishes a painting, he or she should be given the same credit.
You should give credit for all the pictures you have used as references, all the textbooks you have studied, and so on.
7 Likes
I think DrawAI will do so within a few years at the latest.
AI will learn from the pictures AI draws.
This is because the preceding ShougiAI and others have already done so.
These game AIs already do not need human game records.
The AI will learn from the AI’s game records it fights the AI and loses itself.
It will only be a few more years before we can argue that AI is stealing human achievements.
This is similar to what some people said about ShogiAI.
AI is just imitating the game record found by humans.
However, as a result of technological development, human game records are no longer needed, and we have already passed that stage.
DrawAI will do the same in time.
Perhaps it will happen within a few years.
stable diffusion was released in August, NovelAI appeared in September, and as of October, more than 10% of the art that humans have submitted to pixiv so far has been drawn by AI.
Probably, now, more drawings have been drawn by AI than by humans so far.
Unlike game AI, DrawAI has a hard time making it clear that it is a lost game.
It would be an indicator of more shared drawings or less shared drawings.
Or a picture that was ordered to be retaken might be said to be a picture that was not good, and a picture that was saved by a human might be a good picture.
Good and bad paintings would be quantified based on these indicators, and the next generation of AI would learn from them.
Or, an AI will be developed that will generate that picture by trial and error as a target.
4 Likes
My stance on this whole thing:
If you use other people’s work without explicit permission to fuel the AI, it’s wrong. Even if the AI doesn’t explicitly store the actual pictures.
However, if the program ever becomes publicly available, I belive it could be a fun tool to use. So I propose that whenever (if) it’s uploaded, we use a “raw” version, so to say. One that doesn’t have any data in it, and we then feed it only with vanilla GBA portraits. And official sprites and artwork from other games, if possible.
This way, no one’s work is being missused. And everybody wins.
8 Likes
I think this discussion has lost meaning and we gain little from continuing like this. We don’t know IF and WHEN the tool comes out, it’s just an anon on 4chan. (I’m not requesting a mod to lock this, I’m just stating my opinion that at this point the discussion seems fruitless)
3 Likes
still, it is necesary to discuss this things in this moment to be prepare if the tool is free to use, i mean being agree or not is not important at the end if we are not prepare for it, discussing this kind of things ignoring the personals opinion aleast help to know what will the repo do in the moment the tool is here and also help for people to decide what they will do, if somebody doesnt like it being used, discussing it aleast let it clear if they should leave or if the way the repo will manage it is good enough for them to stay
2 Likes
For now I’m leaning towards “Only F2E stuff can be used to train AI” and “Everything AI generated must be tagged as so”, but feel free to propose other solutions.
10 Likes
Given that a class action lawsuit was just filed for infringement of artists’ copyrighted works on several StableDiffusion-based AI sites, this would be where my stance would be until legal matters are sorted out to determine precedent. It would have to be used on images that the artists effectively have said “you can go wild and do whatever you want with this image”, even if regular artists might study the works of another artist to incorporate techniques into their style.
There’s a difference between entering a prompt of “portrait of X by Y” (or “portrait of X in Y’s style”, etc.) and learning the nuances of why an artist does what they do to achieve the effect that they’re going for and then applying those learned principles on your own - the former is only making a facsimile of an artist’s style with whatever the prompter wanted, whereas with the latter, an artist can learn how someone shades the hair on their portraits or how they do meticulous trim work or unique, non-symmetrical armor components, etc.
(Personally (which doesn’t necessarily translate to the discussion at hand here), given that many artists rely on commissions as a means of providing for themselves, I’m against AI generation on the grounds that, if the networks are trained on works that were not given permission, then you are effectively inflicting a loss of income on those artists’ businesses, and introducing copycats as well as adding “confusion” into the “market” and devaluing their own work, much like knockoff clothing/jewelry/handbags/etc., even if the AI output is only being used for personal use. (Things get much sketchier when commercial usage enters the picture.) Honestly, even if an artist does give permission for networks to be trained on certain arts, it’s still going to cause this effect, which is why I think you’ve seen a staggering number of artists in the full landscape be so against AI-generated art. Pixel art is a little bit different since it’s much more specialized and, as a modding community, many artists’ contributions here were made to be shared and such from the beginning in mind.)
(Besides, why would I want to generate a portrait in, say, Glac’s style, when I could gleefully pay him to make it himself and both feel good about it and support an artist at the same time?)
21 Likes
ohanothersorceforslicing,great
(Seeing this was oversized jpg)
…never mind ,that’s not useable
I think AI art uses such a massive sample size to base its art on that the end product is diluted enough to not be attributed to any individual artist.
The idea that AI art is going to completely kill the art industry is frankly untenable, however, so is the idea that AI art is going to itself be killed by the art industry. I think it’s fair to say that, if you know what you’re doing, AI art in its present state is faster and higher quality than a good portion of traditional art. That means it’s here to stay. However, a random, generative program will never be able to reach the heights of human art due to its purely technical, logical nature. That means there will still be human artists making art. What does this mean for the FE community, particularly as relates to GBA art? A few things.
-
The average quality of portraits will increase. This will happen for a few reasons.
A) AI tools will be developed (likely, anyway) to automate processes like making eye and mouth frames, minis (or chibis, or whatever they’re called, I forget), and confirming that the portrait fits within the hackbox. AI base drawing (There’s a better term, but I’m still new to the community, so I don’t know what it is. Basically I’m referring to the solid-color basic shapes portrait artists and animators use to build off of) will speed up initial drafting processes and/or give inspiration to those with ‘artist’s block’.
B) AI-generated portraits will set a decent bar for entry: The only non-sentimental reason to use a hand-made portrait is that it’s better than what the AI can do, which based on the initial post seems to be surprisingly high quality.
C) Editing what the AI has already done will become a skill unto itself. This is good for a myriad of reasons, including an increase in quality beyond what the AI can already do, as well as providing a good base by which aspiring artists can learn to fine-tune.
Edit: Every time I mention portraits, just assume the same applies to battle animations -
Copyright and/or training issues will become less of a concern. As relates to copyright, this will happen naturally as the sample size from which it draws increases, as @Jack_G mentioned, as well as, I suspect, legal battles will result in most high-end generators having anti-plagiarism routines. As for training, I think a large number of moderate AI-supporters, such as myself, will take to marking their work (not to imply that I have any work worth using, lol) as explicitly free to train the AI with, just to get people to stop arguing about it, if nothing else. Pointed stare intensifies
-
We’ll start to see a wider variety of art, such as menu art/pallets and map pallets, stemming from the extra time AI generation has given them for art, as well as just using AI to generate said pallets. I suspect, though AI isn’t at this point yet, that this will also mean an influx of custom battle animations, which excites me 'cause smooth, dramatic, and/or flashy attacks = fun.
Can AI art be used for evil? As much as anything else can. Will it change the way the visual arts work? Undoubtedly. But will it also result in plentiful good, both for artists and for the users of that art? I think so.
Edit: Tried to have a less condescending tone.
But what if an artist is… unavailable?
You wait for the artist to be available
2 Likes
Yall do realize that A.I art can actually coexist with hand-drawn/sprite art? By klok’s set standard these generated portraits are still gonna need human intervention. There will always be people who will choose to buy thought-out, well-made commissions, and those who would try their hand at ai for its ease and cost-effectiveness. Lets try to look at it like a tool and not a threat.
5 Likes
To be honest, I think there will come a point when AI art becomes as good and even exceeds the quality of man-made art. And I think that time will come sooner than most expect too, probably in the next 10-20 years. Every time technology becomes more advanced, people worry that they will be made redundant and it hasn’t happened so far.
there will always be people who like artist’s work more than an AI’s. You can also never get a precise image of what you want with AI unless we get to a point where it can literally read your mind lol.
You can have a conversation about what you want from an artist
You can’t with an AI
AI won’t be replacing real artists anytime soon because only real artists can create exactly what a person wants, AI will eventually be able to get real close sure, but it won’t be able to get as close as real artists can anytime soon.
There will for sure be a time where AI perfectly able to replicate an artist’s style and give you exactly what you want but by that point we’ll have a lot more problems with AI taking jobs then just artists so that’ll need to be worked out when we’re at that point in time.
4 Likes
People have been worrying about technology taking their jobs for centuries. First, most people worked in primary fields (majoratively farming), when machines made that unnecessary, they moved to secondary jobs (factory work), when machines automated that, they moved to tertiary jobs (service sector), where the majority of the workforce is now. Once technology automates that, the next logical step is for people to move to quaternary jobs.
And, even if ai starts taking people’s jobs, once the unemployment rate reached a certain point, governments would be forced to start handing out ubi. Attempting to stop technology progress is a futile endeavour. Where things can be made more efficient, there is money to be made, and where there is money to be made, that method will always be favoured.
1 Like
Well you can. You just cannot guarantee it will perfectly fit your prompt.
1 Like
This I couldn’t agree more with. It’s one of the things that makes humans superior. Because there will always be people who want really specific, and also REALLY freaky art. And out of those people, more than a few will be willing to spend their money
True in some way however with an AI you’re much more limited right now by having to almost perfectly describe what you want, real people are better at interpretation right now and will be for a while. AI definitely is getting better though