My take on it might surprise you. 🎨
UPDATE 2/2023
“AI Art” has been changing quickly. Over the past few months images created by AI are become more and more sophisticated. During this time the companies involved have not prioritized ethics or put the needed measures in place to prevent plagiarism. I am taking a break while they sort out the ethics. I see “AI” as Pandora’s Box and I don’t think it’s going anywhere. But we need to create systems that allow artists to opt out or limitations on how closely any image can resemble one of its reference images.
I would be very interested in an AI that only pulled from art that is in the public domain. This is what I used for my own prompts when exploring this new technology as you can see in my original post below.
I’ve been noodling around with “AI Art” this summer. Basically I’ve been doing this in the fragmented moments I would usually be scrolling Instagram.
Right away I noticed most of what I was hearing about “AI Art” was misleading. I think it’s actually pretty exciting (yes, even for artists) and deeply misunderstood.
I’ve been testing out this technology for a few months now, and I’m not remotely afraid that it will replace human artists.
But I do think it will change art.
Before we jump in let’s be clear…
AI is not an artist.
AI is not intelligent.
So, maybe its the nerd in me, but when I hear “Artificial Intelligence” I picture Data from Star Trek.
We are not there yet.
To simplify a long conversation with my computer engineer husband…
AI is just math.
The programmers created an algorithm (yes, just like on Instagram) to generate an image based on a prompt. And it’s following its programming.
It’s a glorified calculator.
I’ve generated images like the one above using Midjourney. The words in quotation marks are the final prompts. I’ve also included the date. This is important because of updates to the algorithm. The results I have now are drastically different than what I was getting in July.
When you see headlines saying,
“This Magazine Cover Was Designed by AI.”
It just… wasn’t.
Here’s what happens,
Human engineers create an algorithm.
Human inputs prompt, “Art droid.”
Mathematical algorithm determines what pixels to make what color.
Human is given four tiny blurry images to choose between.
Human makes a choice to: A. upscale image B. continue generating new iterations C. change the prompt
Human chooses final image.
Human must then upload this low resolution image into a secondary software to enlarge to print quality.
Human would then combine image and text to create print ready file for the magazine cover.
But “Human Uses New Technology to Design Magazine Cover” doesn’t have the same ring to it.
What’s important here is that the AI did NOT…
Choose to illustrate a magazine cover.
Choose the prompt.
Choose between the generated images.
Choose when to stop iterating.
A human made all those decisions. The human is deciding what to communicate and which image best does the job.
Its also important to understand that it can take dozens or even hundreds of iterations and prompt adjustments to get a final image. It’s like the longest eye exam of all time.
Without sentience AI cannot “make art.”
But it can be a tool for artists.
(Someone please write a sci fi novel where sentient AI want to make art and not war.)
Now, are there ethical issues? Yes.
But they are the same ones that plague every creative endeavor. (I’ll talk about those more at the end.)
Setting those aside, I see the development of AI as analogous to the invention of photography.
Before photography paintings looked a lot like this…
Shortly after photography was invented Impressionism was born.
Cameras did not replace artists. Instead the rise of photography led to innovation. It encouraged artists to pursue the type of art that a camera could never make.
I believe AI will be the same. It will push artists into making new and innovative kinds of art. Maybe it will even increase the value and appreciation of art made by hand.
Photography also became a tool used by traditional artists for reference images and inspiration. As well as an art medium in its own right.
Fine artists who are experimenting with apps like Midjourney are calling this medium “Generative Art”. They are using it mindfully (going through many iterations and utilizing advanced methods I haven’t described here.) These artists may go on to further alter images or combine with other mediums.
Random humans and techies might luck into some good images, but artists have the experience and vocabulary to intentionally push the iterations in an certain direction. Because of this artists are also more likely to be “good” at generative art. Everything we know about composition, color, texture, design, art history, etc. will inform the choices we make.
I think the potential for this technology lies in cases where original artwork would not be commissioned anyway. Newsletters, social media posts, etc. And as a tool for artists.
Here are 4 ways “AI Art” could be used for creative projects:
Concept art for writers and storytellers to communicate their process.
Textures and images for collage artists to use in their work.
A series of images that can be combined into an animation.
Uploading original imagery to iterate on your own work.
Now, coming back to the ethical issues.
As always, it’s important to be respectful of other artists.
While Midjourney can replicate (some) artistic styles you obviously wouldn’t want to copy the style of a living artist in your own work. That wouldn’t be ethical. Just as it wouldn’t be ethical to photograph someone’s work and sell prints.
But you might juxtapose the styles and mediums of historical artists to create interesting effects. The same way these artists can influence your traditional artistic style. I think as an artist its best to really know the source material you are invoking so you can make the call on when it is too referential and when you are on to something new.
It is my understanding that Midjourney, in particular, has been trained on a specific data set of images and does not continuously refer to Google. (My understanding from this article with its creator.) They have been very mindful of giving it a limited set of images to create a more painterly style (and prevent photorealism).
The founder also mentions the possibility for artists to “opt out” of being part of the data set in the future. (Similar to how artists can opt out of Pinterest.)
Another ethical issue is that any AI trained on human culture will inherit its cultural biases. Because of this algorithms are much more likely to depict white humans. And to gender roles in an outdated way.
The last update made some huge improvements depicting humans and I hope they will continue resolving this. For now Midjourney suggests specifying the race or gender you want depicted and it does a pretty good job following that direction.
Overall I think it’s worth a play to understand what “AI Art” really is.
What I’m finding most interesting at the moment is using Midjourney to develop concept art for the book I’m writing. Similarly to how I would have used Pinterest images to make mood boards in the past.
I’d love to share what I’ve learned with anyone who wants to play.
So I’m hosting an AI Character Challenge in October.
You can participate with Midjourney’s free beta test (which comes with around 25 free prompts / images.) Prompts are for character creation, but you could tweak them for other projects.
I think this will be so fun! Especially for anyone planning to do NaNoWriMo.
Let me know in comments if you’d like to join in! I’m still deciding on a platform. Leaning toward Voxer or Discord so we could easily share images and be inspired by each other.
Here are a few of the characters I’ve been playing around with…
Leave a Reply