Photography and artificial intelligence: a lens, not a button

The question everyone asks

Ever since some of my works began with the help of artificial intelligence, the question always comes, and it always comes with the same faintly suspicious tone: did you make it, or did the machine?

I understand the suspicion. It's the same one that greeted photography when it first appeared alongside painting, in the years when people argued that pressing a button couldn't be art because there was no effort, no hand, no craft. That argument lasted sixty years and closed on its own. I suspect this one will take less.

But the question is the wrong one all the same. It's worth explaining why.

Thirty years watching light

I spent three decades studying how light behaves on a surface. In the darkroom first. Then in front of a screen, learning every post-production tool that was ever invented. This isn't a credential I produce in self-defence: it's the reason I know what I'm looking at now.

Because the truth is that artificial intelligence doesn't teach you to see. It hands you a surface. What you do with it depends entirely on how long you spent, beforehand, understanding why one shadow falls a certain way and another doesn't.

People without that history behind them produce images with AI. You recognise them instantly: correct, pleasant, empty. The light belongs to no place, the skin belongs to no body.

The button misunderstanding

There's a widespread idea that AI is a magic button: type a sentence, out comes a masterpiece. Anyone who thinks so has never actually worked with it.

What comes out of a generation is raw material. Most of the time it's the wrong raw material: the light doesn't hold, the volumes lie, hands do what generated hands do, blacks clog, and the texture is false in a way that's hard to explain but immediate to see.

The work begins there. Hours of manual post-production on every single image. I carve the shadows one by one. I rebuild the textures. I calibrate every reflection. I treat each pixel as if it were silver grain, because it's the only way I know to give an image a body instead of a surface.

Of ten images I generate, perhaps one survives. Of the survivors, many sit in a drawer for months before I understand what they were looking for. It's exactly the same ratio of waste I had with film.

What the machine does, what I do

The division of labour is simpler than it sounds.

The machine generates the structure. It gives me a starting point for something that doesn't exist in reality and that I couldn't photograph by any means: a face belonging to no one, a light belonging to no hour of the day, a place that is nowhere.

I decide everything else. Why this frame. Why this light and not the other. What to remove. When to stop. Which of the ten versions can carry the weight of a wall and of years.

They're the same decisions I made in the darkroom, with different tools. The machine doesn't make them for me — and more to the point, it doesn't even know they need making.

The right question

So no, the question isn't whether a human or a machine made it. By that standard we'd have to disqualify autofocus, the light meter, and every camera of the last forty years that chooses something on our behalf.

The right question is the one it has always been: is there an intention in here?

An image without intention stays an image, whether an algorithm or a Leica produced it. A work is an image someone wanted in that precise way, for a reason, ruling out all the others. It's a difference you feel before you can explain it, and it has nothing to do with the tool.

Print as the acid test

There's one test that doesn't forgive, and that's why I care about it.

An image on a screen is backlit, small, forgiving. It hides almost everything. The same image printed at 120 centimetres on cotton paper with pigment inks hides nothing. Every false texture surfaces. Every badly built shadow becomes obvious. Paper is merciless.

That's why I keep printing, and printing large. Not out of nostalgia for material things, but because print is the only place an image has to prove it is something. If it holds there, it holds.

And something curious happens: when an image born of a digital process becomes paper, ink, surface, it stops looking generated. It starts looking like a memory. Something old, something that was already there.

Photography of the impossible

I'm not documenting the real. We already have the real, in abundance, in high definition, in everyone's pocket.

I'm trying to photograph things that don't exist and that we recognise anyway — the light of an afternoon that never happened, the face of someone never born, the memory of a place no one has been. This isn't computer graphics: computer graphics builds objects. This is photography of the impossible, and it uses the entire grammar of photography — light, framing, time, silence — applied to something a camera could never reach.

Artificial intelligence, in all this, is a lens. Powerful, unpredictable, hard to tame. Like every lens, it has no opinion whatsoever about what is worth looking at.

That part, thankfully, is still ours.