The influx of AI art represents a profound symbiosis between human creativity and machine intelligence, reshaping the boundaries of authorship and challenging our fundamental conceptions of what it means to create.
The above sentence is AI-generated. The argument is concise, the language is succinct, and the structure, impeccable. And sometimes (I worry) AI does a better job than the clickity-clacks of my keyboard. For the purpose of this article however, you may have to bear with the imperfections of a human hand as it weaves in and out of the contradictions and redefinitions of what it means to be an artist today.
Remember, remember, the 30th of November. Something new, arguably revolutionary, took the world by storm as OpenAI released an early demo of ChatGPT. Within five days of its release, ChatGPT had a million curious users. A blessing for students, a nightmare for professors, and a fascination for linguists, Generative Artificial Intelligence had the world reeling with possibilities. One didn’t have to seek the self-proclaimed poet’s help to impress a crush anymore, nor did one have to pay the SEO-trained content writer to further one’s business. A new form of creativity, one that calls for a reconsideration of the very word, had been invented.
Generative AI doesn’t just get creative with language. 2022 also witnessed the advent of AI art, and with text-to-image generative technology, your most bizarre dreams could now be explained to the therapist in a picture. And thus, with a blatant nod to a certain sharply moustached Spanish surrealist, Dall-E 3 was born. Cats at a tea party? Painted in seconds. A dog happily writing a thesis? No problem. A neoclassical sculpture with cinematic lighting in the style of Michelangelo? Done, and done. DALL-E 3 was now capable of understanding “significantly more nuance and detail” a little too seriously. Uh… the future had arrived.
AI art wasn’t new. Apparently, an AI-generated portrait had been sold for over $400,000 at a 2018 auction. But it was now available for us to goof around with! This was too much to resist even for a person as immune to social pressures as I am (I haven’t watched GOT yet). One Google search and five minutes later, I had on my screen the atrocities of my imagination, some of which might fascinate you, dear unsuspecting reader.
Cats at a Tea Party
Oh, the poise! Oh, the grace! How could my amateur paintbrush ever compete with that? And there I went, spiralling down a rabbit hole of imposter syndrome (one of the many effects of the wondrous generative AI).
I had prompted Deep Dream Generator to make a picture of cats at a tea party. There was also a plethora of modifiers to choose from - “photoreal”, or “very cute” to “Van Gogh” or “Monet”. My silly little experiments, painted by the Impressionistic greats. The peculiarity deepens.
How does AI generate such pictures? How original are they? Once the shock wore off, the frilly little felines seemed strikingly similar to the style of puzzle artist Linda Picken. Paintings like Cats and Christmas, Autumn Cats, and even Little Bloomers had previously featured whimsical animals much akin to this one. And there I went, spiralling into a frenzy of Google search.
Generative AI is trained on vast datasets of existing artwork. Much like human art students, AI ‘studies’ the various artistic styles, techniques, and motifs of artists, be it Da Vinci or the up-and-coming artist you follow on Instagram. If an artwork is publicly available, AI can access it for its own perception. Once trained, AI uses complex algorithms like Generative Adversarial Networks (GANs) to analyse patterns, colours, and textures from the training data. The AI system then begins to generate original art by combining and modifying elements it has learned from the data. It can produce variations, mix styles, and create entirely new compositions based on the information it has gathered. Think of it as a glorified collage (yes, we can, and will, argue this point).
This aspect has its own can of worms. Does publicising one’s work automatically mean consenting to AI training? AI’s art styles aren’t limited to those of great artists (who were successful but are dead now), it can just as accurately replicate the styles of lesser-known living artists who could do with the money. Thus, copyrighting in the era of Generative AI is an unresolved legal nightmare.
An Indian City (ft. Van Gogh and Michelangelo)
This painting, while not as surreal as I had anticipated, readily shows the influences of the artists from which Deep Dream Generator draws its data. The prompt was “An Indian City with People”, in the style of Van Gogh and Michelangelo. Sure enough, the Taj Mahal is the quintessential postcard vision of ‘India’. Interestingly, Van Gogh’s characteristic brushwork can be observed in the distinct blue and yellow hues. Much like in this artwork, swirling clouds, tactile texture, and unnatural colours are typical of his post-Impressionist paintings. Intricate realism can be observed in the cloaked human figures, on the other hand. Although Michalengelo’s famous anatomical accuracy is not apparent, the drapes of clothing have a touch of chiaroscuro that might remind one of the three-dimensionality of his paintings. AI played a surprising peacemaker conciliating the two diverging artists.
I had an alarming number of art styles to choose from - micro shots, and Renoir, and hyperreal, even Pixar. Arguably, ‘Many Rainbow Bubbles’ is the creepiest.
However, to my chagrin, AI threw up bewildering sketches when prompted with abstracts like “harmony” or “freedom” (or it may just be my own poor understanding of abstract art). The algorithm needed something concrete to illustrate it, it couldn’t interpret and illustrate abstracts by itself.
And here comes the question of creativity. Collage artists take patchworks from newspapers or fabrics or magazines to bring to life an original idea. They may interpret (much unlike AI) a concept like ‘peace’ with colours, gestures, texts, or illustrations that end up featuring a dove flashing a peace sign with its talons. While AI does create a collage combining various artists’ styles, pictures, and existing works (which is arguably a kind of creativity), the prompters still supply the originality. Of course, that may change in the next few years, rendering my argument obsolete. AI’s ‘creativity’ is a result of pattern recognition and data manipulation, and the question of whether this is true creativity in the human sense remains a subject of philosophical discussion.
Until then, we might entertain a disturbing thought - why are we automating creativity? As an amateur artist with a few masterpieces on her classroom desk, I would argue that the pleasure of creating art is as much, if not infinitely greater than, the pleasure of viewing it. Putting a turbocharger on the creative process thus comes with its own set of social implications where consumption reigns supreme.
Our economic systems and societal orders have a huge influence on the direction of technological progress. Automating AI does not solve an existing problem, but instead caters to immediate gratification, a deliberately manufactured demand of consumerism. It practically has a negative social value.
That said, I continue to be awed at art’s resistance to definition. As we grapple with these shifts in the creative landscape, the interplay of AI and art poses new challenges to the already complicated question of Art and Aesthetics. It’s a promising start to the ever-evolving nature of what it means to create and consume art in the digital age, and the showdown will be one for the ages.
Share This Post On
Leave a comment
You need to login to leave a comment. Log-in