Artists and designers have recently been flooding social media platforms like Instagram with surreal, often abstract images featuring an amazing aesthetic. These works are unlike anything any artist has ever produced before, and that’s because the art is actually created by artificial intelligence.
My very cursory understanding of how it works is that an AI engine built upon a neural network is prompted with text, and based on its “knowledge” base of thousands of images it constructs a composition informed by the words you’ve entered.
Prompts are ideally highly descriptive strings of text (e.g. “verdant forest with fairies glowing in the sunset”, “choppy sea with rugged ship tossed by giant waves”, or maybe “gothic character portrait of Victorian gentleman in stovepipe hat on foggy London streets”).
The AI uses the prompts to delineate its output and call upon patterns it has “learned” from its bank of images.
While not all AI engines are created the same, some yield fairly good approximations to the prompts, although you shouldn’t expect a highly coherent or photo-realistic image from even the best at the moment.
Still, the images generated by AI can be very exciting if framed as a sort of impressionism. Some of the forms are only hinted at, while others come shockingly close to what you intended for it to generate.
Regardless of the sometimes left-field and whacky outputs, AI Art has fully arrived.
The burgeoning world of AI art sparked my interest, as I have loved art since I was a boy. Although I never quite developed my drawing and painting skills to a level where I’m satisfied, I love to design things like album and book covers, and am constantly blown away by the skill of modern digital artists.
I decided to try AI art for myself, as I wanted to see how close AI could get to the works of masters both classic and contemporary.
In the gallery below, I had a play with a fun, totally free AI art app called “Dream“. The caption for each image is the prompt I used to inform the output.
As you can see, the results range from nearly abstract to bordering on total coherence, and even when the forms aren’t fully comprehensible, your brain fills in the gaps.
Dream is a joy to play with. For being free, it produces some interesting and surprisingly tasty images.
Other AI art apps to explore at the moment include Disco Diffusion (open source through Google Colab) and the incredible MidJourney (accepting beta-users through Discord, offers subscription plans).
Hop on board with AI art and have a go yourself! You can designate yourself an “artiste” even if you haven’t a lick of artistic talent.
But wait—isn’t this a sham? A farce? An indignation in the face or real artists?
Cheater, Cheater, Pumpkin-eater!
Checking out some Instagram posts recently, I witnessed heated arguments in the comments sections between clout-chasing posters and die-hard, starving-artist types which repeatedly raised the question—isn’t it “cheating” to pass off AI art as your own brilliant work?
Of course it is!
You’re providing prompts and exercising some level of meta-design, but you didn’t make that art, obviously.
You didn’t pick up a paint brush or a digital pen and give life to those forms, stroke by stroke. Anyone with a modicum of insight could appreciate the difference between handmade art (eventually we will need this distinction as AI ascends) and that generated by an artificial neural network.
AI art at the moment is a curiosity, and there is a thrill in waiting to see what wild composition the program will spit out after you enter your prompts and press “GO”.
The wise, future-proof artist will use the AI in such a way that it injects some healthy chaos into their paintings or concept art, combining hand-made forms with elements generated by AI versus relying solely on the neural network to feign a muse.
Yet, what is the endgame here with AI art? To eventually replace artists? Do these developers foresee empowering stiff, business-suited marketing types with the ability to generate endless art on the fly for their capitalist campaigns whilst true-blood artists whither out in the cold?
The Death Knell for Artists?
Any time technology encroaches on a certain field, the exponents of that field get nervous. Their knees knock and their hands are wrung as they anticipate the rising and unstoppable Machine to bring about their obsolescence and irrelevance. To be honest, I felt the same way as an aspiring musician the first time I stumbled upon AI-generated music.
Taking a closer look under the hood, however, you’ll find that none of these AI’s can create spontaneously (at least yet). The aforementioned music-composing AI required many, many patterns (i.e. pieces of music composed by humans) to construct its reference base and produce something even approaching coherence.
The same is true for art. Right now, these AI rely on your prompts and an ever-deepening neural network that analyzes thousands of images across eras, schools, movements, and styles to inform its creations.
Once AI can start spontaneously generating art, perhaps we might have reason to worry, as we may be faced with a neo-sentience devoid of morality or conscience. Sounds like a cool plot for a dystopian sci-fi story, actually ::strokes beard::.
If an AI art engine can eventually create an image with superior aesthetics, more perfect forms, and a higher emotional impact, in 1/100th the time it takes a flesh ‘n’ blood artist, does that mean that art—and by extension creativity—is nothing at all special or unique to humans?
As a creative, I feel that a lot of my muse or inspiration comes from some place outside of myself, some place—dare I say—spiritual.
If AI meets and exceeds our natural, anthropic capacity for these wonderful things like art and music, do we finally lose our souls?
I actually explore these kinds of questions in my writing, especially in the Dark Sci-Fi short stories I have authored and ideated (hopefully they see the light of day soon).
Poetry as Prompts
I thought it would be fun to feed two of the more heavy-duty AI’s—Disco Diffusion and MidJourney—a few of the descriptive lines in some poems I have written. I tweaked some of the prompts to ground the output a bit so that it would produce more concrete forms, but generally the samples of my poetry constituted the majority of the input.
What’s interesting about Disco and MidJourney, is that in addition to the usual prompts, you can enter the names of famous artists (e.g. Gustav Dore, Salvador Dali, Rembrandt, etc.), and the engine attempts to replicate their characteristic styles.
In the gallery below, you can see some of the results.
These poems and potentially the corresponding AI-art will be part of a book of poetry I’m working on as an aside to my debut novels. It’s a collection called Obscurus and the writing process for it is, I think, very unique as it stems from stream of conscious exercises I created for myself, like yoga for the writer’s brain. Maybe I’ll blog about these exercises in a future post.
In the meantime, AI art is a curious little tool to help stoke the imagination of working artists and hobbyists, or to manifest ominous horizons of doom for all of us sensitive, creative types.
Either way, the future of art is upon us.
What do you think? Do you think AI places us creative types under threat? Does it cheapen our craft?
What does this foreshadow in broader terms for humanity’s relationship with tech? Endless possibilities, evolution, or extinction?
Here’s a fun little nightmare seed for fellow authors—what if AI could write novels?
In fact, I think there are already such engines out there that attempt to compose things like e-mail and keynotes. Maybe the novel is next to be soul-sucked?
Let me know your thoughts!