Tabs below contain the full essay, a short version for quick reference, and a sources page that is easy to extend.
Social media often spreads persuasive but incorrect claims about AI. Always do your own research and read other people's peer-review research.
>Watch for confirmation bias. Humans don't like being incorrect or feeling foolish and we will look for evidence that we are correct. =Algorithms can keep you engaging on the platfrom by feeding content that reinforces your starting view
As soon as you upload your work to the internet you give up control over how it is used. Anyone, including bots, can download your work. This has been true since the start of the internet and may explain why some older artists, who were young when the internet began, are less upset about their work being used. We have always known the risks and never falsely assumed we could control how our creative activity is consumed, downloaded or used.
This is especially true on social media platforms or art sharing sites. In order to use these sites, users, you, sign agreements that give platforms the right to use and sell their work. The standard user agreement gives social media platforms these rights "perpetual, irrevocable, royalty-free, fully-paid, non-exclusive, transferable, sublicensable right and license to use, copy, modify, delete in its entirety, adapt, publish, translate, create derivative works from and/or sell and/or distribute such content" You voluntary sign these when you make an account. The AI didn't steal your work, you gave the platforms the right to sell it to the AI companies.
Think of it like this. If you leave your artwork in the middle of Lincoln Road, anyone can come along and take it, they can reuse it by drawing over it or taking a picture of it, they can physically take it and put it in their own home. On the internet, anyone can right click save and take your work. How many times have artists used google image searches to find source material? We can not dictate other's use of our work when our work is in public. Many people will see it as up for grabs. Taking a photo of your work or downloading a copy is not stealing if it is left in public.
Data scraping by AI systems is not copying or stealing. Data scraping is just the automation of right click saving. An action done on a larger scale doesn't fundamentally change what that action is. If you have ever used an image from the internet, then you have data scraped. Artists have been data scraping since the beginning of the internet If information is held behind a paywall, then it may be considered stealing and their are lawsuits about companies breaching paywalls. Those are humans at companies doing that, not "AI." AI has to have instructions to operate and those instructions are given by humans.
Before you upload your work to the internet, consider where you upload it. Read the user agreements that you sign. Often you are signing away your rights to the use of your work.
This is a fact. AI does not copy art. Machine learning fundamentally differs from human copying. It is learning, not copying and it is actually impossible for it to copy. Machine learning does not store or reproduce images directly. The training process for AI is similar to how students study and learn. It learns and creates new work.
If you watched the video by Dr Alan Warburton titled The Wizard of AI (that I assigned and showed in front of class) you know the basics of how the technology works.
AI can generate work in a similar style if and only if a human user direct or trains it to. Instead of blaming the tool, the blame should be on the human user.
Using another artist's style has always been done. The quote is "Great Artists Steal"
How much of a particular artist's style an AI can learn is dependent on the amount of artwork freely available on the internet of that artist's work that the AI can actually recognize. The fraction of an AI’s training data that is “art-related” depends a lot on the model, its objectives, and the choices made by its creators. Given what is public: In a large model like Stable Diffusion, the estimate of art data is a low double-digit percentage (say 5 % to 20 %) of the image portion is “art or illustration / creative art content.” For text, the “art theory / art history / criticism” slice is likely much lower — maybe well under 5 % of total tokens in a general language corpus (unless it’s an art-specialized model). For specialized models (e.g. fine arts style transfer, museum digitization, or artist tools), the “art portion” could be much larger (even dominant). But I stress: this is speculative. The model creators’ filtering, weighting, and sampling schemes heavily influence the result.
It can not copy a style without a human telling it to. This is similar to me assigning you a master copy. Without my assignhment, you wouldn't copy that particular style. Without any style prompts, AI models rely on their data. With style prompts it can mimic a look in the same way a human imitates after study.As stated above, if you put your work on social media or artwork sharing platforms, you sign agreements that allow these platforms to have a lot of rights. For example, this is instagrams standard agreement:
… a non-exclusive, royalty-free, transferable, sub-licensable, worldwide license to host, use, distribute, modify, run, copy, publicly perform or display, translate, and create derivative works of your content (consistent with your privacy and application settings). You can end this license anytime by deleting your content or account. However, content will continue to appear if you shared it with others and they have not deleted it.
You still hold copyright to your own work. If someone else claims they made it or uses it without enough transformation to qualify as fair use you can sue them after registering your copyright. Always keep your original files and documentation.
Fan art is a copyright violation unless the original creator gives written consent. Calling AI copying while making fan art is not consistent since both practices borrow from other artists.
Copyright disputes often consider whether material was behind a paywall or on a password protected site and whether the use is fair use. Remember that humans control AI and humans decide to type prompts that name another artist style.
Some artists are losing jobs because of AI. Many of those jobs are in design not in fine art. This program is for fine artists not for commercial design pipelines.
To support your career find a day job that helps you finance your art animation practice. Create work that cannot be mistaken for AI. Pivot away from styles that non artists with AI tools can generate quickly. An artist who uses AI will create stronger work than a non artist using it.
Many employers in the creative field now require AI use. If they must choose between an open minded artist who can create with or without AI and a person who refuses to use AI showing close minded thinking they will choose the former over the latter.
Across history new tools have displaced jobs. Portrait painters struggled with the camera. Carriage drivers resisted cars. These stories repeat.
I lived through one such shift. In my high school and college years Wikipedia and the wider internet were banned in classes. Many professors believed that everything online was false and students were accused of cheating for using it. Some even failed classes. This should sound familiar.
We adapted in time just as painters and carriage drivers did. The practical lesson is simple. We may not like change or job loss but our options remain the same. We can pivot adopt or adapt.