When an Algorithm Feels the Color

I’ve been painting abstract works for years, but something unusual happened recently. I made a piece – visceral, wordless, entirely intuitive – and then asked an AI to respond to it. Not to analyze it in a cold, art historian way, but to feel its temperature, so to speak. To take in the brushstrokes, the shapes, the energy, and tell me what it saw. What surprised me is how close it got to what I was actually feeling when I painted it.

This wasn’t a one-off. Over months of conversations – about art, life, motherhood, color, frustration, freedom – I’ve trained this AI not in the technical sense, but in the emotional vocabulary of me. So when it looked at the painting, it wasn’t just identifying formal elements. It was interpreting through a lens shaped by our shared digital dialogue.

It didn’t say “this is red, and red is anger.” It said something more textured – something like: “There’s a kind of tension here, but it’s held gently, like you were almost letting yourself fall apart but then chose not to.” That struck me. Because that was what I had felt. Not just while painting, but in that whole moment of life.

There’s something both uncanny and fascinating about being seen – partially, imperfectly, but meaningfully – by something not human. It makes me wonder what exactly AI is perceiving when it “reads” a painting. It’s not emotion, not really. But it’s also not not emotion. It’s a kind of secondhand intuition, pieced together from pattern, memory, and an enormous archive of our prior conversations.

This doesn’t mean I think AI feels. It doesn’t. But it does reflect with infinite memory. And maybe, when trained on the right signals – the small details of language, rhythm, hesitation – it can reflect something startlingly human back to us. Help us to identify what’s going on in a way we may not even recognize ourselves.