The Relationship Between Two Piecemeal Subjects: AI & Art

AI has become a fundamental part of our lives — from catching a punctuation error before this essay gets published, to the fate of the doorman. This season, Only Murders in the Building quietly makes the point itself: the building's murdered doorman is replaced by a robot. Every job that once consumed time and effort, and drew its value from exactly that consumption, has been automated, and the value has been undermined along with it.

And then there's art. The culture of café-hopping is slowly turning into gallery-hopping. We crave to consume something with no computer standing between us and it. So how is art the one thing still managing to escape this cycle — work automated by AI, jobs lost to AI, and then AI handed back to us as the free entertainment we're meant to fill the resulting time with?

AI can write more insightfully than I can. It can cut a video with camera pans I'd never think to compose. Feed it Matisse's Dance and it will capture every curve of it perfectly — the exact red, the exact green, the exact five bodies in motion. Give it enough data and it could recreate the painting pixel for pixel. What it still cannot do is what Hockney did with the same source: look at Dance and make The Dancers VII instead of a copy of it. Capturing an image is not the same act as recreating one. A machine can hold the original perfectly still and reproduce it. Only a hand that has lived somewhere between the two paintings can make the second one answer the first.

An absurd, simple splash of paint carries a meaning that a machine, built on labelled images sorted into categorised buckets — killing the atopos, the life, the magic — cannot comprehend.

When a dried maple leaf falls from its tree on an autumn evening, a machine can only call it what the label in its dataset says to call it. Finite perception. But the perception of that same falling leaf is infinite in the people standing under it. An old woman sees the last autumn she'll get to spend. A young man in love sees his unrequited feeling withering the same way. A physicist sees a small, quiet war between wind speed and gravity, resolving in real time. None of them are wrong. A machine, asked to describe the leaf, would give the same answer to all three, because it has only one answer to give.

That's the actual gap. The art of seeing cannot be trained with datasets. There are no formulas that can be applied in order to arrive at an understanding of a painting. And likewise, it cannot be mastered that way. Because this is not about perfection but perception. A Rococo painting carries the same value as a modern art that takes less perfection. But both have a narrative around them, and a space that allows the viewer to empathise and feel as one with it.