AI / Machine Learning
February 10, 2023

Stop AI - Why Are Some Artists So Upset?

Image generation tools have become quite popular nowadays, and the most talked about are DALL-2, Midjourney, and Stable Diffusion. These tools have the power to create images from text prompts. Some of these generated images are outstanding, creating things that go well beyond the imagination.

While DALL-E 2 produces high-quality, realistic, and detailed images, Midjourney excels in generating art with a human-like artistic touch. Alternatively, Stable Diffusion is an open-source model which is also very good for detailed artistic illustration.

As the results are impressive, there is a huge debate around these tools. Artists are angry and complaining that their job is not being considered. The images that AI is generating are not generated from anything. The AI models are based on millions of images, including artists' work. 

When AI uses artists’ work, they do not receive any return attribution. For this reason, some artists have taken to social media and posted on ArtStation to protest against the use of AI in art. 

On the other hand, not all artists are against AI. Different groups of artists are integrating AI as a tool for creation and inspiration and generating amazing artwork. In this protest, those artists are being asked not to use AI.

In this article, I will provide some context around the problem and share my perspective on the issue. This is only my opinion, and I am not representing any particular site in this debate. All thoughts and ideas are my own. 

Data Laundering

When asking AI to create an image with an artist’s style, the machine generates it with high accuracy; and you can sell that art or use the image as you wish. However, artists did not agree with this. They never knew that their data was being used. When using data without asking, it's called data laundering.

So, the question is: are AI image generative tools data laundering?

The AI’s output is a new image, not one of the original images, but the authentic images were being used to train the AI. Therefore, there are two sides here. One says that AI is stealing artists’ work without permission, and the other side says that AI is just“being inspired” by their work.

The Ethics Of AI-Generated Art

Finding inspiration from others’ work is not illegal; in fact, artists do it all the time. Artists might be inspired by others’ opinions, drawings, and writing to create their artwork, creating your own art and adding your unique touch. Is AI being inspired? Or is it copying and pasting?

AI of concept of a 3D human face on a dark background.

I recommend watching the Everything is a remix series on YouTube, where Kirby Ferguson talks about the foundations of creativity: copy, transform, and combine. Ideas don’t come out of nowhere; they come from other ideas. By applying some variations to the original ideas, we can be creative. Taking different ideas and combining them into something new is creative.

This is the way innovation, and advances in science and technology take place. Researchers build up the unknown, creating theories and performing experiments to validate their hypotheses. So, scientists who know which experiments did not work can try other things and speculate about why they did not. This is how we evolve and make improvements discovering the unknown or creating something new.

AI Algorithms Are Not Magic

We learn by interpreting reality, and AI mimics how we process by learning patterns from large amounts of data. The algorithms take some data as inputs, perform some calculations (transformations) and return a final result. In the case of image generation, the output is a sequence of pixels representing a new image.

But, AI does not copy; it interprets all the input information and learns concepts from patterns. For example, from many images containing a bear, it can learn what a bear is and the characteristics and attributes that compose it.

The only way the AI might copy something from the training set instead of learning a concept is when it needs to be more balanced. Overfitting happens when the AI learns too much from the training set and becomes too detailed in its learning. In the case of image generation, the output image might be very similar to the original one.

As AI is being used in many fields, this issue that is now present in art might happen elsewhere. The same issue with art also happens with code. AI algorithms like OpenAI Codex, trained on huge amounts of code, can generate new code and functions that perform what you ask for. They might use a piece of code from someone else and not cite the m.

On the other hand, programmers also copy and paste all the time. Why would they re-do some code if someone else has already done the work?

What is Art? Is AI Creative?

Holographic silhouette of a human. Conceptual image of AI.

Another discussion regarding images generated by AI is what constitutes art and whether the images generated by AI can be considered art or not. Art is very subjective, so it is challenging to come up with a unique definition. There is art when someone is trying to communicate or express something, whether an idea or an emotion.

So, AI by itself does not express anything, at least not yet. It needs someone to use it and try to communicate something through it, as well as someone on the other side that interprets the art. Therefore, art without communication is not art.

To be able to generate images with AI, you need to be creative as well as practice with the tools. It is not just choosing a text and passing it to the AI. Consequently, the artist is not the AI but the one that uses it.

Who Gets Credit When AI Generates Art?

Nowadays, we share a lot on the internet and social networks. When doing this, we need to be aware that anything we post - whether an AI might use an image, text, video, or audio in some way. Assuming that, it sounds like the AI has the right to use those images. That is not an issue when protecting privacy and following copyright policies.

The real issue is that it might not be clear to people how they can or can’t use the tools. Remember, AI is just a tool. Therefore, policies should be around using the tool instead of banning the tool for something. Just as if an artist plagiarizes someone else’s work is wrong, the same occurs when someone uses AI to plagiarize.

Another point to consider is when creating any dataset with data from the internet to train an AI. Developers should take into account only information tagged with a creative commons license. The problem with image generation models was that, at first, they were going to be built for research purposes, but now they are starting to be used for monetization. For instance, Stable Diffusion is an open-source tool, and you can use it and build your business on top of that.

AI-generated image by Stable Diffusion

So, some image generation tools are charging for usage rather than for the images. That is fine because, according to the tool’s license, you have the right to monetize the generated images. But I understand that several parties are involved who might want to get credit. 

  1. Original artists: when using a prompt that plagiarizes a certain artist, credit should be given to the artists.
  2. The team that developed the tool: the team should be mentioned whenever the tool is used.
  3. The final user of the app: the one who thinks creatively about the prompt and generates the image, should be considered the image owner.

Regardless, these are only my thoughts. For now, there are no policies around this issue being considered. You pay for the use of the tool. What happens before and after needs to be clarified. It should be clearer for all the parties involved what the rules are in this game.

My question is, who should decide what these rules are? And how? Lawyers and other leaders need to resolve this problem, and we need to work together to address it.

My Final Thoughts

AI cannot be stopped. Even if the government decides that AI generative tools are illegal, AI cannot be controlled. The code is already out there, and it is here to stay. I believe it’s crucial to start having these conversations.

The challenge lies in turning this issue into arguments instead. We need to create interdisciplinary teams to address this and other ethical concerns where AI is taking place. As technology advances rapidly, we must act before it’s too late. 

Are you interested in learning more about AI and its benefits? Check out my article on how AI is for everyone.