Thursday, December 15, 2022

Resources for Artists Against Unethical AI Development



Who knew that 2022 would be the year artists would be having the discussion of whether or not they would be replaced by Artificial Intelligence?  But here we are in a very normal looking dystopian future!  Rather than re-hash what other more eloquent folks who have already stated, I wanted to put together this article to gather educational resources and advocacy efforts that I hope will be helpful to the art community.  I'll try to keep it updated as and when I can.  Feel free to suggest more resources in the comments!

Image Source: Pixabay
TL;DR for the folks who don't already know.  AI image generators like Stable Diffusion, MidJourney, Dall-E, and others have trained their technology on data that has been scraped from copyrighted sources without the permission of the owners of those images, including not only artists, but even the medical records of unwitting individuals as well.  To make matter worse, this tech was irresponsibly proliferated via Open Source code before regulation was even considered for its potential for copyright abuse.

We should not promote and condone any positive uses of this tech until the unethical side of it is regulated first, no matter what its helpful potential might be.  It is too harmful to the rights of creators and the creative field, in general, until regulation occurs.  Whether ethical or not, this tech was created to replace the need for artists and that will change how art and professional industries evolve over the next few years, for better or for worse.

And now on to the resource links!


For General Education

  • The Unethical Origins of AI - A thorough and well-researched thread on the unethical elements of AI generation with a breakdown of common defenses of the tech and accompanying explanations.  Complete with links to further research. (Facebook mirror link of this thread in case Twitter goes down).

  • Why AI Models are not inspired like humans. - An article by Karla Ortiz, a respected artist in the entertainment and illustration industry who has done her homework on this tech and also links to outside experts in the field.  This article explains an important distinction between how artists and machines learn, a common defense in letting AI tech go unchecked.

  • Invasive Diffusion: How one unwilling illustrator found herself turned into an AI model - An article discussing how an artist was targeted for her popular style and a model built on her art, which was then released for Open Source use, causing direct harm to the artist's livelihood and brand recognition.  It also discusses the potential for fraud and security issues with AI's power to simulate convincing photos of anyone after training on just a handful of photos of one's face.

  • The Future of AI Art and Automation in Creative Industries (Added 2/23/2023) - Written by prolific commercial photographer, Jingna Zhang, this article discusses the current and future uses of AI in the creative industries, what trends we can expect in job losses, and more considering this comes from someone extremely active in many branches of the creative community.  Zhang is also the founder of Cara.app, an ArtStation replacement that is committed to ethical use of AI.

Okay that was a lot so let's take a break with a meme!  Because the humor is keeping me alive right now.

I hate drawing hands too, MidJourney.

Resources for Artists to Protect Their Work

It may seem like a lot of existential doom right now.  Lord knows I've spent a good chunk of my week not sleeping much and pondering what my illustration career is going to look like in the coming years.  But there are some tools in development to help artists keep up with the quick advancement of this technology.

AI generators function off of using training sets (or datasets) which scrape images from everywhere online, including, but not limited to Google, Pinterest, DeviantART, etc.  The following links help you find your work in these training sets, with HaveIBeenTrained also allowing you to opt-out from training sets that respect this organization's wishes.  It's not guaranteed everyone will honor these tools, but it's at least progress that the major developers, like Stable Diffusion, are stepping up to honor the Opt-Out system.

  • HaveIBeenTrained - This site allows users to register for free and opt-out their images from being scraped for AI training.  It's clunky in its current state because you have to opt out images one by one, but it's better than nothing at the moment.  This site was created by Spawning, a group of artists advocating for ethical and consensual development of AI technology.  They are also currently working on Opt-In tools in addition to Opt-Out.

    You can search for your images manually or search by image upload.  The site claims to not store images used for search, which we will just have to trust is true.

  • What is DeviantArt’s new “noai” and “noimageai” meta tag and how to install it - This article discusses the advent of the NOAI meta tag which can be used to tell image scrapers to not use your site and images for training.  Including instructions for how to install it via popular website management tools like Squarespace.


What Can We Do?  Advocacy Efforts


Right now, what artists need most to combat the mass proliferation of this technology without any oversight is an organized effort which can navigate the many legal challenges involved with this technology.

EDIT: 2/2/2023 - Class-Action Lawsuit, Artists vs Stability AI et al. - A recent class-action lawsuit filed on 1/13/2023 by Sarah Ander­sen, Kelly McK­er­nan, and Karla Ortiz against Stability AI, DeviantART, and MidJourney which purports that these tools are trained by using copyrighted images, which makes the creation of art via the AI's diffusion methods collages works from the training set, making its output derivative works.  Read more details at the link, as it's an excellent simplified breakdown of how the tech works as quoted directly from its own developers.

EDIT: 2/2/2023 - Getty Images vs Stability AI - Another lawsuit which sprang up a few days after the aforementioned class-action lawsuit.  Getty Images claims that proper licensing was not pursued by AI training sets utilizing their stock images.  Another suit to watch for setting a possible legal precedent.

The Concept Art Association Fundraiser - Thankfully, there is currently a recent effort by the Concept Art Association to hire lobbyists, experts, and more to pursue litigation and advocate for ethical development and coexist with this technology.

Check out their Gofundme here.



The Class Action Lawsuit Against Github, Open AI, and Microsoft - The Verge has a good write-up on a court case currently being filed where several class action litigants are arguing for their rights to their copyrighted material being used for training because it is a violation of Open Source terms.  This case may set a precedent, while also examining the loopholes of training sets hosted and developed by non-profit organizations.

If you know of other advocacy efforts or useful links I should add to this resource list, please leave a comment!  For now, hang in there, friends.  It's rough right now, but there will always be artists and people who appreciate the effort that goes into handmade work.

PLEASE NOTE!  I am not interested in debating the facts about AI in this article, as I stand firmly on the side of the scraping of images without consent being unethical.  The main reason I am sharing this post is to gather resources to help artists.  All comments made here that start pointless debate about the merits of AI will be removed.  This isn't the space for it.

No comments:

Post a Comment