I am astounded daily by AI-generated art. I also understand the position of artists who rage against the theft of their visual style, loss of earnings, and disrespect for human creativity. There are some tricky ethics to consider. I won’t reach a conclusion in this little post, but you can be sure I’ll revisit the topic.
This installment of the excellent Everything is a Remix series by Kirby Ferguson explores exactly that. Take a look.
I am an amateur artist: sculpture, drawing, painting, some digital. I’m mediocre, a beginner with a touch of inherent talent but not enough dedication to practice. When creating, I am inspired by things I have seen all my life and the techniques I’ve learned. The final result feels new and is sometimes unexpected.
Using AI to create images delights me; it feels like what I would make using paint and brushes if I could. I’ve designed holiday cards, posters, and t-shirts in the past, for myself and work, and my results were fine, but with the help of AI I’m going to have kickass holiday cards this year. That’s exciting.
If I were a professional artist who had developed a consistent signature of my own, of course I’d be pissed off if anyone could prompt AI to create in my style, without getting my consent or paying me, and the results were good. That would be a direct threat to my livelihood, and earning a living in the arts is far from easy. As someone who owns original artworks and admires creative people, I dread the potential consequences.
Mohrbacher stated that he would like AI companies to seek affirmative consent from living artists whose work is used as reference images, and that seems right to me.
As I’m typing, I’m watching Peter Mohrbacher do a live demonstration of his digital painting techniques in the Midjourney Discord. Mohrbacher stated that he would like AI companies to seek affirmative consent from living artists whose work is used as reference images, and that seems right to me. He also pointed out that copyright in the entertainment arts is often held by corporations, not the creators, and there’s a danger of large companies controlling access to those images for training other models and stifling progress. Mohrbacher has a perspective that he admits is a “spicy take” now, when hostility toward AI is the default in the artistic community: he feels that being in the sample set, being appreciated and influencing new art, is validating. It’s an achievement.
(Tangent: With the endless possibilities that can be envisioned with AI, the number of female images in the Midjourney showcase — sometimes with prompts that specify things like “low-cut dress”, “tight bodycon dress”, “wide hips”, “cute lips” — is unsurprising yet disappointing. Almost every prompt I read for the image of a woman included “beautiful”. That said, I remarked to my husband that I’m glad I’m not in the online dating pool now. If you thought camera filters contributed to catfishing and fraud, just wait until it’s easy to generate a photorealistic avatar and reuse it in other settings as “proof” of reality: at the beach, at dinner with friends, hiking with a dog, etc. It will be soon.)
Perhaps it’s important to consider that AI doesn’t need to be the end of the process. It’s a tool, and the work done by a human before and after is key. In the Midjourney Discord, in Twitter threads, and in Reddit discussions, I’ve seen people help each other tweak prompts or suggest how to alter AI images (paintovers, blends, etc) to get the desired result.
It makes me think of virtual world photography. I choose not to retouch photos I take in Second Life because I want to show what can be experienced. The viewer software and all the things I see on my screen were created by others. The combination of angles, lighting, location, composition, and styling are the human creativity I add that differentiates a virtual photo from a screenshot. Enhancing the result in Photoshop or Procreate can apply a human touch to transform the photo further. Altered SL photos can be jaw-dropping. Is that art, and if so, who is the artist?
But, back to AI and ethics. AI systems are rolling out so quickly that even following helpful curators, I cannot possibly keep up. The debates are just beginning, not only in art, but also text, video, voice, and anything else that AI can generate. The bell cannot be unrung, but the knell can still be contained.
I’m not naïve. I see people trying to generate images specifically in the style of working artists — largely fantasy and comic/anime artists — and AI-generated works are flooding into self-print online marketplaces. But I also see people diving into the new tools with glee, able to create the images in their dreams for the first time. I see people making book cover art for friends, stickers for their kids, posters for their bands, backgrounds for their business plans, illustrations for their comics. It’s not all appropriation or cold profit-motive.
I am enamored with the possibilities even as I squirm at the potential repercussions. I subscribed to Midjourney so I can learn more, but only after reading their policy on DMCA and takedowns. You will see AI-generated art illustrating this blog on a regular basis. My personal choice is to never use the name of an artist in my prompts or attempt to replicate someone’s style. I’ll continue to support artists working in various media and will watch the evolution of policies around AI closely. That may not be enough, but it’s what I can do.
Update: Just after publishing this, I popped over to Reddit and saw a thread from yesterday where someone used Midjourney to make their Second Life avatar look more realistic. I’ve already seen a lot of images like that in profiles, but undoubtedly more are coming. Then, email brought my daily issue of The Download from MIT Technology Review, and this piece: Chinese creators use Midjourney’s AI to generate retro urban “photography”, which touches on some of the ethical questions (with links to more specific articles).