Is It Ethical for Artists to Use Generative AI in Their Work? This Creator Believes So -- See Her Images & Her Approach for Making Them

There's quite a lot of excitement about the potential to quickly create art through generative AI programs like Midjourney, but there's one small problem for starters: Most of these "artificial intelligence" programs are churning out images that are really just copies of art work by actual human artists without their consent or compensation.

Is it even possible to create AI-based, ethically sound images? One artist, who goes by the name "Milena Carbone", believes so. In real life Milena is a photographer who's shown her own work off in galleries, experimented with Second Life screenshots as a medium, and is now experimenting with generative AI.

The images you see in this post are from Milena, and by her lights, they're ethically sourced:

"It's impossible to avoid being influenced by an artist," she allows, "but this isn't specific to AI. When we create, we're always influenced by what we're attracted to, by the authors we admire, consciously or unconsciously."

To prevent outright copying, she declines to include artists in her Midjourney prompts:

"I forbid myself from copying or direct influence, out of ethics, but I should say out of respect for the artist and for myself," she explains. "[I]n the case of Midjourney, I prohibit myself from using the expression 'by' or 'in the style of' followed by the name of an artist. Even if I can claim to be part of [that] artistic movement.

"Another way of creating a personal style with AI is to define a set of parameters that define my style. This is what I do for all my images, with a 'recipe' that I don't reveal. Of course, it's always possible for anyone to come close, but it remains my personal touch, which I evolve as AI evolves."

Another ethical component to her practice is making AI just one among many tools she uses to create the final image:

W

Watch Milena's AI artist statement above

"In my latest work, for example, virtually all the images are assemblages of several Midjourney images, but also sometimes real life textures and Photoshop image processing." With recent updates to programs like Midjourney, DALL-E 3, and Firefly in Photoshop, along with new tools like Ideogram and D-ID, she believes digital artists like her will integrate multiple tools. This will lead to more well-rounded or complex works.

"All this, I believe, will contribute to these tools making it possible to create original works, and not just copies of existing artists, or, as is still the case today, works that you can 'feel' are produced with the help of an AI (which I recognize, as far as my current work is concerned)."

Lately she's heard friends and relatives who are artists complain about the rise of generative AI, and has felt the backlash herself. "I understand their reaction, and agree with them for the most part. Recently, I was even insulted and dragged through the mud on Youtube."

Ironically, the very fact that these images are called "AI" seems to give people permission to heavily insult the human creators who generated them, sometimes with great artistry and painstaking time.

But the high level of vehemence also suggests it's time to have a more productive conservation.

"Paradoxically," as Milena puts it, "these emotional overreactions show that these new tools are challenging us and that it would be a good idea to open a debate that would enable AI to serve art even better. I know I'm a pioneer, and it's not easy to be marginalized in this way, simply because you've decided to 'paint with a new brush'." 

Frankly I suspect the real conversation will begin when (if) art critics begin acclaiming AI-based works. Nettrice Gaskins' Deep Dream images have already risen to that level, but it will take others to reach that plain for that dialog to truly start.  

All images depicted here by Milena Carbone.  Normally I'd say they were "copyright" the artist, but given they're built on AI programs and claimed by an anonymous artist known only by her Second Life avatar name, I'll leave that to a judge to decide.

Comments

[Moon Beerbaum]

Personally, I think that there is nothing artistic about tools like MJ. An AI can only calculate on the base of data that needs to be fed first. That reference data base with billions of entries was filled with billions of stolen images (like the huge data breach via API on Flickr a few years ago). Otherwise an AI cannot calculate, it does not know what and on the basis of what, and towards what.

Do "artists" using such tools realize that their work, too, will be reassembled, reused, recalculated and comes with no copyright protection at all?

I loved Bryn Oh´s article about the matter, also, because she points out solutions:
https://brynoh.blogspot.com/2023/10/nightshade-vs-image-generative-ai.html

[Milena Carbone]

Thank you for this article, which reflects my thoughts and my work very well. On the question of style, I'd like to add that Midjourney's latest "/tune" feature is incredibly powerful for creating and stabilizing your own style. It also shows the Midjourney team's concern to address artists and illustration professionals.

[George Djorgovski]

I think that Milena is exactly right. AI is not an artist - it is a tool than an artist can use, requiring new skills, but also everything they learned in the art school about the composition, colors, etc. The output image reflects the creativity of the person who wields that tool.

This also enables people who are creative and have something to express, but are physically unable to hold a painting brush or whatever, to express that latent creativity. Why would you want to exclude that pool of talent?

Let me also dispel the common misperception that generative AI copies individual artist's images. Its output is based on averaging of sorts of millions of images found on the Web, a vast majority of which are photographs produced by non-artists, and if any artist-generated images are included in the training data, it is a negligible fraction of it, and pretty much impossible to determine.

Certainly, artists should be paid for their work, but the burden should be on the people or companies who put copyrighted images on the Web, and there are tools to do that. Even before the generative AI, anyone who downloaded an image of your artwork that someone else posted on the Web is not liable. The party who posts copyrighted material without protection is.

[mescota]

In the realm of generative artificial intelligence art, the problem that you are talking to is, without a doubt, a critical worry. A great number of artificial intelligence models are trained on datasets that include copyrighted artwork without the express consent of the artists who created the artwork. In light of this ethical concerns about intellectual property and the rights of artists are brought to light.