AI Alchemy

TL;DR: This article traces the evolution of AI image generation from crude pandemic-era experiments to today's sophisticated tools. It explores both the technical advancement and cultural impact of these technologies, highlighting the ongoing debate about whether AI-generated imagery constitutes art, while suggesting that “AI Art” has emerged as its own artistic movement regardless of where one stands on the issue.

Writing this in mid-2025, it is amazing to see how far AI image generation has progressed in just a few years. Not so long ago we sat isolated in our homes during the global pandemic and noticed the appearance of odd images, both realistic and flawed, creeping into our social feeds.

A half-decent effort at a Romantic woodcut

Ever enthusiastic and curious about new tech, I sought out the source of this novel technology, wanting to play with it and understand how these dream-like images burst into existence from the latent noise space.

Back then I had to either wait my turn on a Discord server or OpenAI's site for a chance to type in the magic keywords that described, clumsily, my desired image. Then, seconds (or minutes) later it would appear like some sort of cyber-Polaroid, surprising and delighting with its novelty. And often—hideousness.

A less romantic interpretation of the Lady and.. a sea Otter with creepy fingers?

While this was both magical and terrifying, the real exploration began with the release of the open source model, Stable Diffusion. Sure, DeepDaze, BigSleep, and VQGAN-CLIP were around a year or so earlier, but their outputs were just not coherent nor versatile enough for general use. Stable Diffusion launched an entire ecosystem and soon enough you could go beyond the base model and train your own, mess around with the minutiae of settings, and delight yet again in the creations that now took on more realistic or stylistic dimensions. And it could run on your own machine, independent of anyone else.

From sketch to an image

Around this time, serious concerns about what is and is not Art began to circulate. Sure, the obvious copyright infringements and deepfakes were legitimate and established concerns. However, the line between what can be considered your own original creation and something produced by a process far removed from human creativity started to get fuzzy. As always, the critics decried AI as clearly not art. It was just machines rehashing digested (and often inappropriately acquired) artworks. But what of technical enthusiasts who, like myself, toiled for hours each day to squeeze and pull the algorithms every which way and shape into being their unique creations? Whether started from a pencil sketch or a photograph or a sequence of prompts, the more people got involved in the multi-step process, the less clear it became whether these stochastic probability machines or their directors (i.e., humans) were the artists.

A few days work, distilled in an AI alchemical flask

Wherever you place yourself on the spectrum of this debate, it is certain that “AI Art” is emerging as its own movement. Just like Impressionism and Photography had their bitter opponents and passionate proponents, so too will this create critics and fans worldwide. It will also continue to inspire debate, reflection, and new futures. And in the end, that is one of the key points of Art.