Hi David. Have you tried to block those blokes by a filter?
You might even use the phrase to make it more precise.
My recomondation would be to put a filter in place with the web-app (https://framapiaf.org/filters hopefully gets you there directly). In my experiece this is the easiest way to create and manage them.
Hope that helps. Enjoy the day and please do not let people like her/him get to you (easier said then done, I know. My whish and hope is to be able to cheer you up). 😘
@madiko Looking at the screenshot makes me believe it's from the birdsite.
@davidrevoy you know that you have a loyal community over here. Since the owner switched there, the ratio of trolls increased. I'm aware that the network has value for marketing. But don't let that drag you down.
@madiko Yes, thank you for the advice , but it was on the birdsite as mentioned by @RyunoKi.
True Ryuno-Ki, Thanks. Yes, I archived my Twitter data, removed all people from "my follow" that I could find here on Fediverse, and just kept my Tw account as a (RSS) relay to post the main news now (yes, for the visibility of my project).
True, I should be more prepared to receive comments from Posthumanism/AI Take Over, NFTs, Crypto, and all the baggage that come with that.
It's not built-in and not supported by official team working at Krita Foundation. You can find Stable Diffusion as a plugin made by someone externally to the team. https://github.com/w4ffl35/krita_stable_diffusion
It requires downloading the model of the AI separately, and you'll need an account.
Stable Diffusion/LAION model is not Free/Libre and Open-Source compatible. It's a form of Open-Source with many rules on the top to prevent "AI Take Over" and unethical usage (how ironic).
@Joseph_of_Earth I think you are right about it. The same than when a street portraitist with charcoal finishes a piece nowadays and people around them tell "it looks like a photo" as a compliment. I guess when photo appeared, it wasn't well received to compare a painting to a photo.
I think it's a difference in kind. If an artist is going for photo-realism, I can imagine it being a compliment to have your work be compared to a photo. In a material sense you can argue that photos are the ultimate way to capture a picture of the world.
To be compared to AI as a compliment implies that AI has a high level of skill, and you are good when measured up against that level of skill. I will be sad if AI becomes the ruler we use to measure quality in art.
I think most commenters are overreacting here. They likely never heard of Krita so didn't have the clue. AI-generated images are hot right now and it looks like many others; they likely saw it in passing and liked it enough to comment.
To address some other comments, AI image generation isn't "stealing" just because it's trained on your work. It's making new works inspired by it. Would you hate people for making works inspired by your work? The AI is doing the same thing.
I find it problematic that you put on the same level:
1. Creators who love, hate, and interact with input within the zeitgeist and then get inspired (consciously or unconsciously) by that.
AND
2. An A.I. software that received a massive list of URL of images inputs+descriptions, without the consent of their authors, and curated by a team (with their own political choices).
I wouldn't call 2 inspiration but just massive stealing of data for analysis.
that's not my intention. I'm not equating those. People are conscious, emotional beings, while AI currently are not. There may be ethical issues w/scraping publicly posted data but that's a separate issue from what I was saying. That said, people can view the images, do analysis and use it for inspiration, so why can't AI?
The real problem as I see it is that it lowers the value of such work. That is an unfortunate downside but it's often the cost of progress from new tech.
phrasing like ‘inspired’ anthropomorphise a heuristic that does a complex version of rote reproduction. it creates nothing new, its output is based wholly and entirely on reproduction of pre-existing imagery. overfitting examples to expose how the sausage gets made show this beyond a shadow of a doubt. experts on ‘AI’ agree with this.
and when the creators and owners of the material that is being reproduced did not license their material for that use, it can indeed be called ‘stealing.’
@Tanath My 0.02€: I don't see a fundamental difference between a human artist's brain receiving inputs from what the person can see while looking at other artist's *published* creations, and an AI being fed the same creations although the AI will need a lot more inputs to learn than the human can do (AIs are still very inefficient at learning).
The morality issue starts when the user asks the AI “make a picture like David Revoy's”. Maybe an “AI-not-welcome” licence may help?
Oh, I did it, but it was rather polite and first degree, I'm sure it's disappointing compare to all the talent and humor I read from the replies to my post.
I wrote:
" Thanks, but I don't use any A.I. It's a painting, from scratch.
I'm against the way the databases for training the model were built without the artists' consent. "
I really wonder how we, as human beings, will handle this AI augmented future. I expect it will often make us think about how we evaluate certain things. For example: Do we like a painting because its pretty? because it tells a story? because it inspires our imagination? or maybe because of its history? how it was created or by whom? Humans have always built tools to extend their abilities and to overcome their limitations but how long will it take to get used to these new tools?
@sepia These questions are in loop in the history of modern art and contemporary art, and it is a fascinating philosophical adventure to read about it and forge a philosophy. More than 100 years of artists tried to answer that.
Attaché : 1 image
Experiment: Stable Diffusion Ai (left) and Krita paint-over (right).
In short, this process is tedious. Fixing anatomy issues of the Ai takes a lot of time.
I know, but it's not the A.I with its little legs who have decided by a beautiful morning to scrape the web for art alone. 😅
No: it's a team of dev − eg. working at LAION https://laion.ai/faq/ − who decided to scrape all URL and ALT text, enrich the database with their description, curate the result (eg. remove the watermark, keep the bias of stereotype work/gender/race). Political choices here.
LAION, Large-scale Artificial Intelligence Open Network, is a non-profit organization making machine learning resources available to the general public.
So according stable diffusion AIs trained on web-scraped images, ALL doctors are white men wearing blue shirts with neckties, lab coats, and stethoscopes.
AI can solve problems ONLY IF you have good data to train it, and acquiring is very costly. My professional experience working with AI is that there is hardly ever good data. But companies cut costs, use bad data (garbage in, garbage out) and systemic biases can end up being reinforced.
@sepia "Register an account to opt-out your images from Stable Diffusion V3"
This is the most fucking upside-down logic I've ever seen, and definitely 100% illegal basically anywhere on the planet. However this is the internet - it will be used anyway, and with the modern economy also using data in immoral amounts things won't change in the near future. 😔
@Natanox@sepia I’m reminded of how Internet Archive operates on an Opt-Out model. Do y’all have a problem with Internet Archive? After all, it scrapes stuff too.
@Liberonscien@sepia Valid argument. However the Internet Archive does not process or manipulate them in any capacity other than to make sure they stay accessible. Neither do they profit from them nor replicate anything to profit from.
From a legal standpoint it might be comparable. Not from a moral one though.
@sepia I used to draw with stroke-based rendering that i saw from some dude somewhat like 5 or 6 years ago (even can't remember his nickname now), and i never asked his consent. Should i now kill myself for the sake of justice?
@lonelyowl First, you haven't made any commercial benefit out of it, nor have you pretended to replace the job of the said "dude". 🤣
And also, please don't compare your own process of learning and practicing with your affection for a style or an artist at a cultural time to a software feed with a long list of URL and description batch downloaded and curated by a team of worker (eg. no Mature, no Watermark, and more other choices). It's not intelligent, it's a tool.
@sepia > you haven't made any commercial benefit out of it
Well, in general, I tried to draw for money, but the process of creating a single simple picture was incredibly long, one time i spent four days without sleep and get 15 bux for it. I assume that even african kids in cobalt mines get paid more. Now i draw only for the lulz and because of boredom.
However, other people were much more lucky in this regard. We have enough of clones of popular artists like kuvshin, sakimichan and others who make a lot of money by patreon and comissions, and no one soul saw a problem within it.
> And also, please don't compare your own process of learning
I'm lack of technical knowledge to have a strong opinion about it, but i have an observation how a system that never saw drawings with Taylor Swift successfully adopted real life Taylor's appearance to generic art rendering style. (I merged checkpoints of novelai and Taylor model from someone from reddit)
Claiming that this relatively primitive system can reason would be too naive, but it seems to be capable of more than just brainless copying.
Anyway, complaining that other people's pictures were used in training without consent looks like absolutely fantastic greed, comparable to that demonstrated by Disney and other media corporations, because even DMCA leaves ways to use other people's work to create derivative works, as long as they are not completely brainless plagiarism and/or publishing without changes. If you ignore this fact, then it turns out that that dude has the right to burn part of my brain so that I forget about his drawings.
The only thing that is really sad about this whole story is that it looks like another profession will die as a result of automation, and there is still nothing like UBI anywhere in the world. In this regard, I am extremely interested to know what is the point of educating economists if we going ignore them anyway.
@lonelyowl In general I have to agree. A fundamental technique of our species is to learn by copying and no copyright law can, will or should ever prevent that. Now the way these AI systems are trained is obviously not ok either, especially if companies make an effort to remove watermarks, deliberately ignore the rights of creators and sell the result. The question is where do we draw the line? Would it be ok to train a system on photos of protected art? When is the result unique?
@sepia If you stop thinking in terms of intelligence chauvinism, implying that machine learning is fundamentally (not technically) different from human learning, you may find that countless words have already been written about the originality of art and the nature of plagiarism, and much of it is already built into our copyright laws.
If the machine itself is able to separate the watermark from the drawing during the learning process, then there seems to be no reasonable way to prevent it from doing so, since I can do the same.
Artists frequently participating in various holywars about who stole what from who involving humans only (i dislike them btw, but now it's not about that), applying the same logic to ai generated art will be much less inadequate than trying to prevent fair use.
@sepia Also, would be interesting to ask if activision blizzard legally able to take down sakimichan's (or other same kind of artist) patreon for using their protected by copyright characters to accrue money 🤔
@lonelyowl It's a topic I know a bit as a free culture advocate, I wrote a full conference about the problems with fan-art years ago (for LibrePlanet https://www.youtube.com/watch?v=AvtcHAMsQp4 ). Yes, commercial fan-arts are all under the fair use of companies. They could ask for removal at any moment and engage law-suit too. But it would be very unpopular with the fan base. So, it's a balance the companies learned (or not) to find. Commercial AI will face same challenge, eg. https://lexica.art/?q=star+war
"Why I'm speaking at the FSF libreplanet conference? because this is the origin of my work for the libre culture in general. I saw a very big parallel betwee...
Probably the best solution would be to apply copyright to final result of generation process based on how generated picture is going to be used (commercially or not) instead of moderating dataset.
Hi, David. You can no opt-out your images in https://haveibeentrained.com/. This will prevent, at least, Stable Diffusion 3 to be trained with your works.
@marnic Merci. Je pense que j'ai posté cette copie d'écran et réaction sincèrement ressentie car c'est la goutte d'eau qui a fait un peu déborder mon vase... J'étais resté bien silencieux sur ce sujet jusque là. Sans doute trop.
Alors que oui, ça peut être une blague, ou un troll commun de l'air du temps, ou une question légitime des temps qui arrivent et j'ai pris ça a coeur...
Bon après, si il y aurait eu un smiley, ça serait 100x mieux passé en blague.😉
J'avais posté ça il y a quelques jours https://framapiaf.org/@marnic/109489021372748775 Car en tant qu'amateur d'art et d'art numérique le déversement de toutes ces images généré en 2 clics m'énerve de plus en plus.
Je vois plein de belles images sur Masto ou ailleurs mais aujourd'hui je ne sais plus …
Je ne sais plus si c'est l'oeuvre d'un artiste qu'il faudrait féliciter pour son superbe boulot ou juste quelqu'un qui a demandé un truc vite fait à une IA.
@marnic Ha oui, en effet. Oui, Bien vue. Je pense que ça sera le groupe le plus enclin à des comportements protecteurs, sincère et/ou éthique de préciser par un hashtag la nature des oeuvres. Je pense que "humanart" ou similaire risque de prendre de la traction, car les autres imposeront leur aiart comme normalité et feront tout pour noyer ça ni vue ni connue. Comme de la fausse monnaie qui vient dévaluer un marché. Il faut que je me fasse à ce mot clef + partager des vidéos/process.
I thought A.I. ripping off our software (via github auto pilot) was bad enough. But it looks like it is even worse for artists. At least nobody was asking programmers: ”which A.I. engine wrote your program”.
@Gavy@Vilrax Ha, ben ça m'interesse aussi pour le coup si j'ai un concepteur. J'ai une assez grande liste de bug-report sur mes biais et ce *#@! de corps. Assez pour me demander si j'étais pas un projet fait le vendredi soir sur un coin table après un pot de départ un peu trop arrosé et publié à la hâte avant de partir en vacances. 😆
@wekeys Autant le "ton métier ne sert plus à rien.", je comprends, et comme c'est mon métier, je prend ça au sens large. Mais la seconde partie est assez violente et plus pour ma pomme: "va te vendre", "sur Twitch", "react cringe" ... Je t'ai fait du mal par le passé ou je me fais une idée?
@fartkart Oh, I did. It's just not on the screenshot. But you know, you'll be disappointed by what I really answered because I was polite and all. I wrote: " Thanks, but I don't use any A.I. It's a painting, from scratch.
I'm against the way the databases for training the model were built without the artists' consent. "
Attaché : 4 images
Process:
Digital painting from scratch (no A.I.) using Krita 5.1.3 on Fedora 37 KDE spin. Tablet: an Intuos Pro Large. A description of each step is written on the alt text of pictures.
📦Source and 4K wallpaper: https://www.
It's not a guess, LAION trained an AI to detect watermarks, the code is here: https://github.com/LAION-AI/watermark-detection , on their website, they also mention often the quality of their asset for cleaning this part of the dataset.
A repository containing datasets and tools to train a watermark classifier. - GitHub - LAION-AI/watermark-detection: A repository containing datasets and tools to train a watermark classifier.
@chengdulittlea@sepia oh actually the large one does have a detector for watermarked images, and for I believe porn, so that any downstream user could avoid training on such images.
@xarvos Oh yes, in September I was very excited by it and enthusiast, you remember it well. (src https://framapiaf.org/@davidrevoy/108969346357981937 ) but that was before tool like https://haveibeentrained.com/ confirmed the model was trained with the work of artists (mine included, fully). It changed my appreciation of the tool a lot. I wish a model existed with only CC-0 and Public Domain photo/art/clipart...
Attaché : 1 image
Experiment: Stable Diffusion Ai (left) and Krita paint-over (right).
In short, this process is tedious. Fixing anatomy issues of the Ai takes a lot of time.
@wekeys Hello Perso je lis ce post comme du cynisme de situation, une illustration d'une tendance globale qui, comme un personnage, dirait cela et avec cette violence. Donc rien de personnel. Le personnel on le trouve dans le fait que l'auteur choisi ton post initial, le personnel est donc une belle 'embrassade' d'empathie qui partage le "tired" de ton post initial. 😉
Attached: 1 image
Oh, that's so good.
Actual artists on Artstation, fed up with AI generators scraping their work, started posting "no AI" logos (the classic circle with a line through it) in protest.
@kookie Hey, I run an open-source webcomic where all material are downloadable in free access and under an open license. You'll find the 4K wallpaper in a post I published later: https://framapiaf.org/@davidrevoy/109513537083560213 (and if you want a print, I can upload the design on my Redbubble shop).
My main income is made with patronage and online donation. If you want and can support my quest to continue to make art this way , you'll find on https://www.peppercarrot.com/en/support/ all the options (Patreon/Liberapay/Bank/Paypal)🎁 .
Attaché : 4 images
Process:
Digital painting from scratch (no A.I.) using Krita 5.1.3 on Fedora 37 KDE spin. Tablet: an Intuos Pro Large. A description of each step is written on the alt text of pictures.
📦Source and 4K wallpaper: https://www.
@kookie Thank you! No it doesn't punish me. On Patreon, I'm not using it 'per month', but 'per webcomic creation', and I post rarely a new episode (one every three or four months). So, I'll not impact much your budget. In best case, the new episode should come at the end of February.
C'est triste… Mais en prenant du recul, c'est aussi un drôle de compliment.
D'un point de vue historique, cela me fait penser au remplacement des peintres par la photographie, puis les photographes par la photo numérique, etc. Il y en a toujours, mais moins 😋
@gigantos@Tanath Theft of image data would deprive you of it, this is not the case, despite what the propaganda from music labels and film studios has been hammering for years.
Regarding creation, I believe (until proven wrong or told that the problem is ill-defined) that one does not know whether the processus of creation of humans or AI is similar or different. I try to refrain from claiming it's either, until I have more information.
Seeing in that a new form of life that need same privilege than human about copyright is −Imo− pure ideology ( called posthumanism/AI TakeOver I think ).
It's not because the trick is complex that this machine 'think'. If it would be, the problem of 6 fingers rendering would be not one. 🤣
But your remark about the copyright privilege is interesting: who should be the legal entity held responsible w.r.t. copyright? The human training the AI? The AI (probably not)? The user of the AI? I guess the question is similar for online meme generators (pictures taken from wherever with an overlay of text created by the user of the tool). Is there any legal decision on those?
@matthieu You didn't wrote that. Sorry 😉 That was a caricature. Just one I couldn't resist after reading too many comments, I think I got just too many feedback over the last 24h and I don't think well now.
Yes, the responsibility of AI copyright is an interesting case. I'll see what people harvest from the recent provocation of Disney copyright ( like https://vmst.io/@selzero/109512557990367884 )
Attached: 4 images
Well this has come to a head. In order to protest AI image generators stealing artists work to train AI models, the artists are deliberately generating AI art based on the IP of corporations that are most sensitive to protecting i…
@gigantos@Tanath I thought more about the copyright and AI. If you consider that the training of an AI is like looking at published pictures (i.e. meant to be looked at), then it's the person who uses the AI as a tool to generate a new, derivative, inspired-by-what-it-has-seen picture who is responsible for respecting copyright law. Just like a person who is inspired by another artist is responsible if it produces something that looks too much like an original work.
@Timkongart Yes, it looks like I'll probably get more replies like that in the future. I wrote a ...
🚫 No NFTs 🚫 No A.I.
... in my Bio, maybe it will help the commenters to know it is now highly unrealistic I used any A.I. engine in my art. I'll see if it prevents anything.
@matthieu Except in case of a person, you can't prove the source of inspiration, you can't open the brain and see the picture stored somewhere to legally prove it was used. It might a coïncidence, a re-creation, etc... In case of a computer and “an advanced photo mixer software generating potential derivations based on statistical probability", it's possible to prove legally. The Concept Art Association is starting a fundraiser to start a lawsuit, btw https://www.gofundme.com/f/protecting-artists-from-ai-technologies
@gigantos@Tanath So you claim that it's possible to take a trained AI and find the original images in it? The way I understand neural nets, I doubt that very much.
The lawsuit seems to want to discriminate between humans who watch the source images and get inspired by it and software doing the same (but on an industrial scale). So it fights against the industrialization of the process, right?
@matthieu Yes. The image are dataset provided by LAION. LAION makes triage (remove watermarks, pr0n, infringing picture). You can get the set here https://laion.ai/blog/laion-5b/ , it's the one that train model for Midjourney/Dall-E and Stable Diffusion. Website like https://haveibeentrained.com/ are new services returning if your art is part of the Laion dataset (or other). It's just a big database of URL + description.
We present a dataset of 5,85 billion CLIP-filtered image-text pairs, 14x bigger than LAION-400M, previously the biggest openly accessible image-text datas...
@gigantos@Tanath Ah. The dataset is not the neural net. The dataset is a list of annotated images to look at. Conceptually not much different (except in size) from e.g. a Mastodon user who boosts others' pictures for everyone to see.
But as the dataset effectively contains images that LAION probably has no permission to redistribute, it would indeed breaks the law. But let's be accurate: the point of contention is the redistribution of the training set, not the AI.
@matthieu Yes, totally.I'm all ok with AI if they were trained on Public Domain material. Many tools (already in Photoshop, a smart lasso tool, a sky replacer, a background extractor) already use trained model to do some job faster and more accurately, and with a dataset that Adobe research had permission.It's OK.
The real issue is with getting −without my consent− 10 times my Alice in Wonderland CG Award illustration on LAION-5B, training all AI 😔
"I've been training this neuronal model for seven years straight but it is not generating the expected output" sounds so much more exciting than "my kid is now seven and acts up"
I could never get my head around these people who just cannot seem to realize that there's actually people with skills who make stuff…
It's just too far removed from baseline reality for my brain to acknowledge people can actually think like that. Like, who do these people imagine is actually creating all this stuff?
In hindsight, it seems only natural that these people would now start to assume everything is done by an AI just because they can't do that particular thing. :/
Attaché : 4 images
Process:
Digital painting from scratch (no A.I.) using Krita 5.1.3 on Fedora 37 KDE spin. Tablet: an Intuos Pro Large. A description of each step is written on the alt text of pictures.
📦Source and 4K wallpaper: https://www.
@Homebrewandhacking Sorry, I'll not help to identify the account and that's why I blured their name and replaced the avatar with a default... I try to prevent doxing as much as I can. I just wanted to show that for us, artists, we are now in this new era where this type of question is common. If you want to discuss with AIArt enthusiasts, you'll find many after exploring the right hashtags here.
I'd love a better tool to help artists but, in my limited understanding, it draws on artists copyrighted material to create things.
I'd love a tool that was trained on public domain art because I'd like stock art that wasn't pale, male and male gazey because in my niche, bootstrapping is a real problem.
hacknorris
•RMPArquimago
•Fay Corazón de Aceite
•David Revoy
•Finnley Dolfin
•florian :flan_hacker:
•R. L. Dane :debian: :openbsd:
•Mike Risley
•Nyandzette
•raphael
•Spin D'Accord
•Björn
•Nizar Kerkeni 🇹🇳 نزار القرقني
•Xavi
•Jectoons :krita: :debian:
•Jookia
•Arne Babenhauserheide
•(better: »One to whom you can actually explain your wishes. For 300€¹ I’ll show you.«)
¹: guess. Insert the actual cost of your image when paid fairly.
David Revoy
•Franziska Köppe | madiko
•You might even use the phrase to make it more precise.
My recomondation would be to put a filter in place with the web-app (https://framapiaf.org/filters hopefully gets you there directly). In my experiece this is the easiest way to create and manage them.
Hope that helps. Enjoy the day and please do not let people like her/him get to you (easier said then done, I know. My whish and hope is to be able to cheer you up). 😘
Framapiaf
Serveur Mastodon hébergé sur framapiaf.orgRyuno-Ki
•Looking at the screenshot makes me believe it's from the birdsite.
@davidrevoy you know that you have a loyal community over here. Since the owner switched there, the ratio of trolls increased. I'm aware that the network has value for marketing. But don't let that drag you down.
David Revoy
•True Ryuno-Ki, Thanks. Yes, I archived my Twitter data, removed all people from "my follow" that I could find here on Fediverse, and just kept my Tw account as a (RSS) relay to post the main news now (yes, for the visibility of my project).
True, I should be more prepared to receive comments from Posthumanism/AI Take Over, NFTs, Crypto, and all the baggage that come with that.
Franziska Köppe | madiko
•Yes, you are right. I should have read a little more carefully.
Anyway: Good success with "Pepper&Carrot". I really love them and wish you all the best.
David Revoy
•@RyunoKi
Geotechland
•Natasha Nox 🇺🇦🇵🇸
•I'd probably react really pissed to these people answering like "it's called skill and talent, b**ch". Would change nothing though...
嘎！我鹅鹅又回来啦 :goose_walk:
•schratze
•Emmet O'Neill
•Jeko
•David Revoy
•monkee :fairydust:
•Roelf Renkema :nafo: :admin2:
•@theropologist
David Revoy
•It's not built-in and not supported by official team working at Krita Foundation. You can find Stable Diffusion as a plugin made by someone externally to the team.
https://github.com/w4ffl35/krita_stable_diffusion
It requires downloading the model of the AI separately, and you'll need an account.
Stable Diffusion/LAION model is not Free/Libre and Open-Source compatible. It's a form of Open-Source with many rules on the top to prevent "AI Take Over" and unethical usage (how ironic).
GitHub - w4ffl35/krita_stable_diffusion: A Stable Diffusion plugin for Krita
GitHubJoseph of Earth :fedora:
•"Wow, you made this? It looks like it came out of an AI!"
David Revoy
•Joseph of Earth :fedora:
•To be compared to AI as a compliment implies that AI has a high level of skill, and you are good when measured up against that level of skill. I will be sad if AI becomes the ruler we use to measure quality in art.
Joseph of Earth :fedora:
•David Revoy
•El Jefe ":verified:" :donor:
•David Revoy
•El Jefe ":verified:" :donor:
•Efi (nap pet) 🦊💤
•SammyGalen
•Hikarii
•Hgmarty
•Chilly Dee Williams
•David Revoy
•ghostdancer
•Lau 🔞
•David Revoy
•Marnic
•😁
ploum
•Konato
•Nicol'Ours Chasseur2Lueurs B.
•spacebarbarian
•Dylan Sanchez
•LilaTequila
•David Revoy
•GothicAnon
•🌈 Dana
•Thomas Frans 🇺🇦
•Nevral
•Tanath
•To address some other comments, AI image generation isn't "stealing" just because it's trained on your work. It's making new works inspired by it. Would you hate people for making works inspired by your work? The AI is doing the same thing.
David Revoy
•Content warning: Aiart / opinion
1. Creators who love, hate, and interact with input within the zeitgeist and then get inspired (consciously or unconsciously) by that.
AND
2. An A.I. software that received a massive list of URL of images inputs+descriptions, without the consent of their authors, and curated by a team (with their own political choices).
I wouldn't call 2 inspiration but just massive stealing of data for analysis.
(edited for CW)
David Revoy reshared this.
Tanath
•The real problem as I see it is that it lowers the value of such work. That is an unfortunate downside but it's often the cost of progress from new tech.
raphael
•Content warning: re: Aiart / opinion
phrasing like ‘inspired’ anthropomorphise a heuristic that does a complex version of rote reproduction. it creates nothing new, its output is based wholly and entirely on reproduction of pre-existing imagery. overfitting examples to expose how the sausage gets made show this beyond a shadow of a doubt. experts on ‘AI’ agree with this.
and when the creators and owners of the material that is being reproduced did not license their material for that use, it can indeed be called ‘stealing.’
Matthieu Weber 🇫🇷/🇫🇮
•The morality issue starts when the user asks the AI “make a picture like David Revoy's”.
Maybe an “AI-not-welcome” licence may help?
farfadet46
•Marcel Costa
•Sorry for the joke, but here we know you too well to make this comments.
Hugs!
David Revoy
•Felix
•Steven
•David Revoy
•kirby
•kirby
•David Revoy
•I wrote:
"
Thanks, but I don't use any A.I. It's a painting, from scratch.
I'm against the way the databases for training the model were built without the artists' consent.
"
kirby
•SEPIA Open Assistant
•David Revoy
•Content warning: Aiart / opinion (2)
About getting used to AI, I embraced it in September ( read https://framapiaf.org/@davidrevoy/108969346357981937 ) but that was before all artists − thanks to website like https://haveibeentrained.com/ − saw their artworks being used without their consent.
*That* changed everything.
David Revoy (@davidrevoy@framapiaf.org)
FramapiafDavid Revoy reshared this.
Yiming Wu ✅ Use OurPaint
•Just do the thing you do, ai doesnt have the personality and the life experiences to make the same thing you make
David Revoy
•Content warning: Aiart / opinion (3)
No: it's a team of dev − eg. working at LAION https://laion.ai/faq/ − who decided to scrape all URL and ALT text, enrich the database with their description, curate the result (eg. remove the watermark, keep the bias of stereotype work/gender/race). Political choices here.
I just hate that.
(Edit: I removed a Ctrl+V error😅 )
@sepia
FAQ | LAION
laion.aiDavid Revoy reshared this.
Ramin Honary
•Content warning: Aiart / opinion (3)
So according stable diffusion AIs trained on web-scraped images, ALL doctors are white men wearing blue shirts with neckties, lab coats, and stethoscopes.
AI can solve problems ONLY IF you have good data to train it, and acquiring is very costly. My professional experience working with AI is that there is hardly ever good data. But companies cut costs, use bad data (garbage in, garbage out) and systemic biases can end up being reinforced.
Natasha Nox 🇺🇦🇵🇸
•Content warning: Aiart / opinion (2)
This is the most fucking upside-down logic I've ever seen, and definitely 100% illegal basically anywhere on the planet. However this is the internet - it will be used anyway, and with the modern economy also using data in immoral amounts things won't change in the near future. 😔
Liberonscien :VeryQueer:
•Content warning: Aiart / opinion (2)
Natasha Nox 🇺🇦🇵🇸
•Content warning: Aiart / opinion (2)
From a legal standpoint it might be comparable. Not from a moral one though.
David Revoy
•Content warning: Aiart / opinion (2)
Yes, very interesting point about Internet Archive, and fair reply from Natasha about the purpose, finality and moral.
@sepia
Owl!🦉
•Content warning: Aiart / opinion (2)
I used to draw with stroke-based rendering that i saw from some dude somewhat like 5 or 6 years ago (even can't remember his nickname now), and i never asked his consent. Should i now kill myself for the sake of justice?
David Revoy
•Content warning: Aiart / opinion (2)
And also, please don't compare your own process of learning and practicing with your affection for a style or an artist at a cultural time to a software feed with a long list of URL and description batch downloaded and curated by a team of worker (eg. no Mature, no Watermark, and more other choices). It's not intelligent, it's a tool.
@sepia
Owl!🦉
•Content warning: Aiart / opinion (2)
> you haven't made any commercial benefit out of it
Well, in general, I tried to draw for money, but the process of creating a single simple picture was incredibly long, one time i spent four days without sleep and get 15 bux for it. I assume that even african kids in cobalt mines get paid more. Now i draw only for the lulz and because of boredom.
However, other people were much more lucky in this regard. We have enough of clones of popular artists like kuvshin, sakimichan and others who make a lot of money by patreon and comissions, and no one soul saw a problem within it.
> And also, please don't compare your own process of learning
I'm lack of technical knowledge to have a strong opinion about it, but i have an observation how a system that never saw drawings with Taylor Swift successfully adopted real life Taylor's appearance to generic art rendering style.
(I merged checkpoints of novelai and Taylor model from someone from reddit)
Claiming that this relatively primitive system can reason would be too naive, but it seems to be capable of more than just brainless copying.
Anyway, complaining that other people's pictures were used in training without consent looks like absolutely fantastic greed, comparable to that demonstrated by Disney and other media corporations, because even DMCA leaves ways to use other people's work to create derivative works, as long as they are not completely brainless plagiarism and/or publishing without changes. If you ignore this fact, then it turns out that that dude has the right to burn part of my brain so that I forget about his drawings.
The only thing that is really sad about this whole story is that it looks like another profession will die as a result of automation, and there is still nothing like UBI anywhere in the world. In this regard, I am extremely interested to know what is the point of educating economists if we going ignore them anyway.
SEPIA Open Assistant
•Now the way these AI systems are trained is obviously not ok either, especially if companies make an effort to remove watermarks, deliberately ignore the rights of creators and sell the result.
The question is where do we draw the line? Would it be ok to train a system on photos of protected art? When is the result unique?
Owl!🦉
•If you stop thinking in terms of intelligence chauvinism, implying that machine learning is fundamentally (not technically) different from human learning, you may find that countless words have already been written about the originality of art and the nature of plagiarism, and much of it is already built into our copyright laws.
If the machine itself is able to separate the watermark from the drawing during the learning process, then there seems to be no reasonable way to prevent it from doing so, since I can do the same.
Artists frequently participating in various holywars about who stole what from who involving humans only (i dislike them btw, but now it's not about that), applying the same logic to ai generated art will be much less inadequate than trying to prevent fair use.
Owl!🦉
•Also, would be interesting to ask if activision blizzard legally able to take down sakimichan's (or other same kind of artist) patreon for using their protected by copyright characters to accrue money 🤔
David Revoy
•Commercial AI will face same challenge, eg. https://lexica.art/?q=star+war
@sepia
How to free the imagination
YouTubeSEPIA Open Assistant
•Owl!🦉
•Probably the best solution would be to apply copyright to final result of generation process based on how generated picture is going to be used (commercially or not) instead of moderating dataset.
David Revoy
•@sepia
Néstor Arellano
•Content warning: Aiart / opinion (2)
Hi, David. You can no opt-out your images in https://haveibeentrained.com/. This will prevent, at least, Stable Diffusion 3 to be trained with your works.
Cheers. 😀
Have I Been Trained?
haveibeentrained.comMorgan
•Marnic
•Remarque j'ai faillit la faire sur celle du dragon, mais je me suis ravisé, c'est une mauvaise blague.
David Revoy
•Je pense que j'ai posté cette copie d'écran et réaction sincèrement ressentie car c'est la goutte d'eau qui a fait un peu déborder mon vase... J'étais resté bien silencieux sur ce sujet jusque là. Sans doute trop.
Alors que oui, ça peut être une blague, ou un troll commun de l'air du temps, ou une question légitime des temps qui arrivent et j'ai pris ça a coeur...
Bon après, si il y aurait eu un smiley, ça serait 100x mieux passé en blague.😉
Marnic
•https://framapiaf.org/@marnic/109489021372748775
Car en tant qu'amateur d'art et d'art numérique le déversement de toutes ces images généré en 2 clics m'énerve de plus en plus.
Marnic :unverified: (@marnic@framapiaf.org)
FramapiafDavid Revoy
•Je pense que ça sera le groupe le plus enclin à des comportements protecteurs, sincère et/ou éthique de préciser par un hashtag la nature des oeuvres.
Je pense que "humanart" ou similaire risque de prendre de la traction, car les autres imposeront leur aiart comme normalité et feront tout pour noyer ça ni vue ni connue.
Comme de la fausse monnaie qui vient dévaluer un marché.
Il faut que je me fasse à ce mot clef + partager des vidéos/process.
Hera
•der.hans
•"Cool witch! What did you use for this?"
If they followed you, they'd know it was most likely Krita
David Revoy
•Lyons
•Andrius Štikonas
•David Revoy
•@chengdulittlea@sepia
David Revoy
•Vilrax
•David Revoy
•David Revoy
•montag
•David Revoy
•:duck-shaker: fartkart :duck-shaker:
•David Revoy
•I wrote:
"
Thanks, but I don't use any A.I. It's a painting, from scratch.
I'm against the way the databases for training the model were built without the artists' consent.
"
:duck-shaker: fartkart :duck-shaker:
•David Revoy
•David Revoy (@davidrevoy@framapiaf.org)
FramapiafDavid Revoy
•Content warning: Aiart / opinion (3)
It's not a guess, LAION trained an AI to detect watermarks, the code is here: https://github.com/LAION-AI/watermark-detection , on their website, they also mention often the quality of their asset for cleaning this part of the dataset.
@chengdulittlea@sepia
GitHub - LAION-AI/watermark-detection: A repository containing datasets and tools to train a watermark classifier.
GitHubDavid Revoy
•Content warning: Aiart / opinion (3)
Ha, right. Please ignore my previous reply, I read this one of your a bit later after sending it. Yes, a Mature+Watermark filtering.
@chengdulittlea@sepia
Tak
•Content warning: Aiart / opinion (3)
⏣ (nut)
•David Revoy
•David Revoy (@davidrevoy@framapiaf.org)
Framapiafbubar
•Hello
Perso je lis ce post comme du cynisme de situation, une illustration d'une tendance globale qui, comme un personnage, dirait cela et avec cette violence.
Donc rien de personnel.
Le personnel on le trouve dans le fait que l'auteur choisi ton post initial, le personnel est donc une belle 'embrassade' d'empathie qui partage le "tired" de ton post initial.
😉
Fred en bois
•The Book of Kels (@Nezchan@wandering.shop)
The Wandering ShopDavid Revoy
•Grincheux 🚲
•David Revoy
•Benjamin Bouvier 🥐
•🌅 Aurore maxi-mum to be 💜
•Sylvhem
•Purple
•|Gon0S| ⏚
•Stupeflo
•David Revoy
•David Revoy
•David Revoy
•My main income is made with patronage and online donation. If you want and can support my quest to continue to make art this way , you'll find on https://www.peppercarrot.com/en/support/ all the options (Patreon/Liberapay/Bank/Paypal)🎁 .
David Revoy (@davidrevoy@framapiaf.org)
FramapiafDavid Revoy
•Agnes A :verifvelo:
•Adrien Plazas
•David Revoy
•David Revoy
•Norihiori
•D'un point de vue historique, cela me fait penser au remplacement des peintres par la photographie, puis les photographes par la photo numérique, etc. Il y en a toujours, mais moins 😋
Courage !
Monsieur B.
•Rémi
•David Revoy
•- Religious: "This is a holy book that gods have written".
- Economist: "There is an invisible hand that balance the free market".
- Youtube: "It's not our moderation decision, but the one of the algorithm"
- AI "It's not us who stole the img data; but the AI who gets inspired and think by itself"
The bow to our invisible 'superior entity almighty'.
Not for me. 😆
@gigantos@Tanath
Nabil-Fareed Alikhan
•David Revoy
•Matthieu Weber 🇫🇷/🇫🇮
•Theft of image data would deprive you of it, this is not the case, despite what the propaganda from music labels and film studios has been hammering for years.
Regarding creation, I believe (until proven wrong or told that the problem is ill-defined) that one does not know whether the processus of creation of humans or AI is similar or different. I try to refrain from claiming it's either, until I have more information.
David Revoy
•if you browse the Stable Diffusion paper, you'll see it's only a complex overlay of image + synthesis. This example is from the paper:
https://ommer-lab.com/wp-content/uploads/2022/08/article-Figure16-1.png
Seeing in that a new form of life that need same privilege than human about copyright is −Imo− pure ideology ( called posthumanism/AI TakeOver I think ).
It's not because the trick is complex that this machine 'think'. If it would be, the problem of 6 fingers rendering would be not one. 🤣
@gigantos@Tanath
Matthieu Weber 🇫🇷/🇫🇮
•But your remark about the copyright privilege is interesting: who should be the legal entity held responsible w.r.t. copyright? The human training the AI? The AI (probably not)? The user of the AI? I guess the question is similar for online meme generators (pictures taken from wherever with an overlay of text created by the user of the tool). Is there any legal decision on those?
a form of argument and an informal fallacy based on giving the impression of refuting an opponent's argument, while actually refuting an argument that was not presented by that opponent
Contributors to Wikimedia projects (Wikimedia Foundation, Inc.)David Revoy
•Yes, the responsibility of AI copyright is an interesting case. I'll see what people harvest from the recent provocation of Disney copyright ( like https://vmst.io/@selzero/109512557990367884 )
@gigantos@Tanath
Deniz Opal (@selzero@vmst.io)
vmst·io𝚓𝚎𝚎(Backup);
•Which one ? 👺
David Revoy
•This one: https://boulderbugle.com/aipyth-ifu3ebs5
AIPyth
boulderbugle.com𝚓𝚎𝚎(Backup);
•David Revoy
•Matthieu Weber 🇫🇷/🇫🇮
•If you consider that the training of an AI is like looking at published pictures (i.e. meant to be looked at), then it's the person who uses the AI as a tool to generate a new, derivative, inspired-by-what-it-has-seen picture who is responsible for respecting copyright law. Just like a person who is inspired by another artist is responsible if it produces something that looks too much like an original work.
Baskit
•source : https://creapills.com/bescherelle-orthographe-intelligence-artificielle-20221215
Bescherelle se moque de l'IA pour rappeler l’importance de l'orthographe
CreapillsDavid Revoy
•Timkongart
•David Revoy
•🚫 No NFTs
🚫 No A.I.
... in my Bio, maybe it will help the commenters to know it is now highly unrealistic I used any A.I. engine in my art. I'll see if it prevents anything.
David Revoy
•@gigantos@Tanath
Matthieu Weber 🇫🇷/🇫🇮
•The lawsuit seems to want to discriminate between humans who watch the source images and get inspired by it and software doing the same (but on an industrial scale). So it fights against the industrialization of the process, right?
David Revoy
•Website like https://haveibeentrained.com/ are new services returning if your art is part of the Laion dataset (or other).
It's just a big database of URL + description.
@gigantos@Tanath
LAION-5B: A NEW ERA OF OPEN LARGE-SCALE MULTI-MODAL DATASETS | LAION
laion.aiMatthieu Weber 🇫🇷/🇫🇮
•But as the dataset effectively contains images that LAION probably has no permission to redistribute, it would indeed breaks the law. But let's be accurate: the point of contention is the redistribution of the training set, not the AI.
David Revoy
•The real issue is with getting −without my consent− 10 times my Alice in Wonderland CG Award illustration on LAION-5B, training all AI 😔
https://rom1504.github.io/clip-retrieval/?back=https%3A%2F%2Fknn5.laion.ai&index=laion5B&useMclip=false&imageUrl=https%3A%2F%2Fi.pinimg.com%2F236x%2Fb3%2Fc5%2F23%2Fb3c5235c3016ebe0e16d6492d1cba88d--cheshire-cat-art-alice-liddell.jpg
@gigantos@Tanath
Clip front
rom1504.github.ioaeris 🏳️🌈
•Nodoka | trap is fine too
•Aral Balkan
•David Revoy
•Claudius (legacy account)
•phryk 🏴
•It's just too far removed from baseline reality for my brain to acknowledge people can actually think like that. Like, who do these people imagine is actually creating all this stuff?
In hindsight, it seems only natural that these people would now start to assume everything is done by an AI just because they can't do that particular thing. :/
lertsenem
•David Revoy
•*The* Paul Brown
•Lennart
•Dashtop
•Bartłomiej Garbiec 🇺🇦
•David Revoy
•David Revoy (@davidrevoy@framapiaf.org)
FramapiafUkuleleBard
•Ten Thousand Worlds
•I … I … I do not have enough middle fingers to properly respond to that.
PJ Coffey
•David Revoy
•I just wanted to show that for us, artists, we are now in this new era where this type of question is common.
If you want to discuss with AIArt enthusiasts, you'll find many after exploring the right hashtags here.
PJ Coffey
•I'd love a better tool to help artists but, in my limited understanding, it draws on artists copyrighted material to create things.
I'd love a tool that was trained on public domain art because I'd like stock art that wasn't pale, male and male gazey because in my niche, bootstrapping is a real problem.
Ajira
•