ya know i dont know too much about ai but to me it almost sounds like finding pics online with poses or backrounds artist do like me as well sometimes or taking photos of yourself to help. funny thing a beastars one when looking up japanese clothing i was like.. hold up wait a minute. saw that. but dang was good edit. one thing i dont agree with big time is doing that stuff for commissions. and also i would hope as well people still learn to make thier own talent even with references find thier own style ya know? a mix of both if ya need it to learn if that makes sense ^^;

er to edit i wanted to ask what your opinion is there ^^

I’ve fiddled with the AI image gen just to see what it was like, and it’s kind of like herding cats in the prompt writing and playing slot machines with the results. I see why it’s popular because it’s addicting in that you don’t really lose per-say. In the time spent finding the best result, though, I could’ve painted what I wanted without all the extra steps. People who haven’t learned the arts still see it as a means to do something they thought was impossible, which isn’t true in my opinion, talents can still be earned to some extent. It kinda made me jealous more than anything when I played with it. It was the magic wand, so to speak, but it wasn’t my magic it produced. It is an incredibly limited thing, however, even these supposed grand models have a fixed form and style thats still behind the curve compared to actual creators.

thats really fair, thinking about it, ai is way different than just like finding references,
when you mentioned its like a magic wand and taking time to do that vs learning. understandable big time. and yes the last part a fixed form, creators have more of like a for lack of a better word and using my name there, a flow
ya know i think i like traditional art too alot cause of that and sketches

on food packages? watcha mean?

ohh i gotcha, how is it ya can even tell with logos? i dunno if id be able to

oh! with a character that would make more sense for sure

I would say always be skeptical on budget shortcuts on food items, even if it’s just packaging. I know we shouldn’t judge a jerky by it’s bag but one would sure as heck remember their jerky expeirence after the fact and what kind of bag it was in good or bad, thats just cave critter logic.

I mean we’re already in August, we may just see more of it next year. It’s the weird rendering artifacts that make it the most jarring.

Greed corrupted it more than anything, I do see it’s benefits in aerospace technologies, for orbital infrastructure pioneering. I’m in the boat that AI is a double edged sword. I think there was a choice on how the models were developed and fed information but i think we missed an opportunity for a better result due the mentality of the AI servant. A lot of the Digimon series archs comes to mind in many cases on that.

I really don't like seeing ai being used as a source for inspiration or reference for anything, such as writing and drawing. I'm even afraid that at some point in the future such practice become so common to the point it will be inevitable for me to not get inspired by ideas of clanker origin, because the artists I don't know yet, whose art I might like and consequently get inspired by, will likely be using ai to get ideas and inspiration from. That'd completely ruin my trust in people's creativity, like, "Wow, this artwork of yours is so awesome! How'd you come with that?!!" and the boring answer is "Oh, I use ChatGPT".

Yaay, I'll have something to be boomer about in the future, how wonderful, how wonderful... dang it.

First thing I thought when I saw some older traditional artists using AI as reference for ideas was the cognitive atrophy it induces. Relying on AI for concreting any nebulous idea takes a large portion out of the conception process mentally. From an inspiration point of view I’ve been half impressed by some of the compositions generated from AI, but I then remind myself “well that looks like a Monet” and I try seeking out people artists and the old masters as a reference and research them. Pretty sure I’m already there in being boomer about something but I try keep my eyes open on different perspectives which is a challenge because cynicism and skepticism come naturally with age. It’s such a weird time to be alive.

I think I’ve noticed corrective algorithms being used on phone cameras now, especially in low light conditions it adjusts the final result based on several images taken. It especially became noticeable on lengthy exposures through my telescope for stars and nebula (some of which I used as stock for some of my doodles here)

The most dubious part of the AI Image gen was them scraping from sites like shutter stock and the like (as well as the internet at large). I was actually able to find a way to backtrack where the images were from on a model and found a lot of it came from those sketchy 3rd party wallpaper sites, which had already scraped from artists webpages so it’s like two fold scraping shenanigans.

AI has become so pervasive that policies protecting artists need to not be knee-jerk reactions making vague or far reaching rules without digging deeper into details. It’s a difficult argument to make, as well, that it’s only art if a person makes it because people can argue they still make the prompts and therefore the art. (people who are pro ai have likened that to it’s own art which in my opinion is childish and arrogant.)

Overall I wish the ai models were learned exclusively in the open source and copy left side of the internet, within the licensing rules of course, this whole content theft debacle might have been avoided and it would’ve been that quirky automation tool people were skeptical about.

I mildly missed the plot due to rushing to work today, oof! Concerning derivative work from ai yea that kinda feeds into the concern of everyone being inundated with it now days and people unconsciously drawing inspiration from memory of it.

I imagine the concern is whether the personally tailored model is perfectly hermetic from a moderation perspective. People can say a model is completely trained off data they truthfully own but whether FA has a team or council to verify it really is takes resources i doubt they have. That and there’s the community backlash part, people could say that model still contains x y or z data that the poster doesn’t own or have rights to. I think i’ve seen there even might be a trend starting that people just use the ai gen claim for defamatory purposes. It’s sadly very muddy and murky waters.

I suppose it depends on whether you want your art to be a mediation of the world, or of a mediation of a mediation; pastiche is an artform in and of itself, but I think artists always risk committing it by accident, which, tbf, we furries are all already a little guilty of. Regarding the rule, most AI stuff I've seen in furry has been fundamentally expedient, and rules like these legislate against that expedience more than anything, even if they're fairly blunt force. It at least stymies some of the spam you see on sites like deviantart.

There's is no obligation to research or understand anything, which is what I'm saying! Furry is wonderfully syncretic, whether we realise it or not. We've all produced art for many decades now that mixes the high and low with very little contradiction, and this is only enhanced by the range of skill and professionalism you see. I've run into some artists on her who partake in the behaviour you're describing, and yet their work is genuinely novel and in many ways beautiful. Even the majority of it which we might consider amateurish forms the background against which we place other art, and it's not something I wish to denigrate.

My point about expedience is not to support the marketisation of furry art either! Anecdotal, but the majority of furry AI stuff I've seen has essentially been reproducing the model you're arguing against: I've mostly only seen AI furry prompters adopt popular, commercially-appealing styles, and I've seen a fair few accounts which sell prompted commissions in a similar fashion. Ofc, this kind of kitsch derivation is already a thing with human artists, so it's less about whether AI allows people to freely express themselves, than it is about whether it conversely forces them into a certain mode, both of thought and style, especially if this mode becomes the de facto means for anyone looking to become an artist, or a writer, or anything else. Do we want to encourage art for the fursona's sake, or for its mass appeal? Would we have had a Henri Rousseau if he had simply decided to use midjourney?

But I appreciate the discussion! :> It's something I've thought about a lot, and I apologise if this is a bit all over the place. You bring up some good points, and I think there's a lot of ways things could go.

I gotta be honest I didn’t expect contructive discussion for how touchy the whole ai art conversation is now days.

Sticky’s got a point on the pallet swap recolors, I remember living through that era too. The ai art stuff just feels like an iteration of that. People get a pass on the base recolors and stuff either because its just been on the scene long enough or it’s within the terms of the original creator, which is it’s own market. You make a good point too, Disco, and I bet in another decade we’ll all look back on this period like it was the kitschy nuance that comes with every era.

The one hurdle that’ll need crossing, though, is these specific artist style models that have been cropping up more recently now that training methods are getting more public and useable. An artist on youtube I follow by the name Only Jerry(@DaHeckSaJerry) already put out a post expressing their discontent (they were as polite as possible in the post) that their style was directly being mimicked by a specially trained ai model. I think thats were the line absolutely needs to be drawn. At that point thats not only a brand and identity risk but a wound to the soul when someone, while in their mind they just want to see more art from that person, basically circumvents the artist just to get that art or style for their own viewing pleasure.

Sticky raises a valid point of artistic expression. AI is just a tool, and like a tool one can express creative habits such as building a house or destructive habits such as demolishing a house. The kicker is it takes seconds to break something and days to make something. The second kicker, of course, whether the materials for building were honestly sourced. I think the addage a good artist copies, while a great artist steals is an incredibly slippery slope even for people who paint from their own mind and hand.

To elaborate on the paragraph relating to specialized ai models. The issues to be hurdled is basically who owns and has rights to what ai models, based on their training source, and who has access to those models. Is it the source artist and by the source artist’s discression that the model is used or access be granted, or is it the public? That alone can go both copy-left and copy-right, there are ethical solutions for both outcomes.

Yeah, coming from an art theft angle, LoRAs do feel like the scummiest manifestation of this market-chasing and malicious behaviour. However, I have a lot of difficulties with the whole debate surrounding IP and copyright wrt to generative AI, because it frankly concerns issues of ethics and economics that exist outside the scope of just AI, and I've seen people jump to the defence of corporate IP law which was once loathed. If you've got any stuff to read on your latter point, I'd love to see!

However, wrt the argument about your point on copying and theft: these models are all basically statistical distributions, but importantly they're not a set of weights of *everything*, but of their underlying corpora. Someone can choose to use the existing data, but they can only use that data, and controlling the biases of this data through prompting alone is actually pretty tough, requiring specially trained models and workarounds, to say nothing of models' odd quirks (take, for example, ChatGPT's higher-than-average use of Em dashes in prose). Stacking models upon models can only make this issue more complex.

These biases are subtle, but they compound over time, and across iterations. I think this really fundamentally disrupts the narrative of these tools being general-purpose tools, as opposed to a means of imitation. The furry models I've observed are dictated not by the choices of an artist working consciously (and unconsciously) to steal or borrow, but by the structure reflected in e621 tags. This might be a negligible point from the perspective of someone generating AI porn lol, but models like Midjourney, Stable Diffusion, etc. don't just spit out anything, but rather fairly specific things.

My point I think was to suggest that this can have a stultifying effect for all art, and for beginner artists: not to just be limited by a predetermined set of data, but also to not even understand this fact, and to not have developed in such a way as to shape your own advancement and understanding. Sure, you can start generating lovely faux Rembrandts from day one, but where does that ultimately take you? It's a cool party trick, sure, but what are you even attempting to talk about, or mediate? What did you even achieve here that anyone else with the same prompt and a GPU couldn't? Sure you can curate, but models can be trained to curate too, possibly better than you. It all feels to me not just an example of the destruction that comes with mechanical reproduction in the sense of Benjamin, but of the destruction of the aura of creativity itself.

The flipside to this, however, is whether it matters: arguably part of the success of current models rests not on their capabilities to generate any image, but of their ability to create specifically kitsch, and to do so on an industrial scale (again, working on a predetermined outcome). Kitsch (capital-K) is not a new phenomenon; it's had a foothold for a long time, and provides a very easy cultural niche into which a large amount of generative imagery fits. To take that question about Rousseau I asked above: we create a diffuser through some process that grants it a unique and novel output, not trained on the work of any specific painter, but on an unseen and undetermined corpus. This artstyle ends up, to our sensibilities, crass, difficult, and undesirable, and we discard the project as a failure. If Rousseau's work had been the ouput of an AI, would it have had a chance at all?

These thoughts are all a bit half-formed though, so apologies for the ramble. I think I agree with both your points broadly, and I'm glad for the discussion!