Tag: AI in photography

AI, Identity, and the Bridge That Doesn’t Exist

Posted on Posted in Black and White PhotographyTagged , , , , , , , , , Leave a comment

I asked an AI to show me what ijmphotography.net looks like, and it gave me back a Parisian bridge that doesn’t exist in Nantes. But it got everything else right: the cat, the tea, the film camera, the books that shaped how I see. It nailed the feeling while getting the geography wrong. Which made me think about what it means to ask a machine to reflect who you are, and what gets lost in the translation.

What I Asked For

AI-generated portrait of bearded man wearing glasses and straw hat, Nantes France photographer
The person asking an AI to explain who they are. I always thought I had a better skin that this illustration lets on…

I gave it nothing. Just the domain name: ijmphotography.net. Except that wasn’t quite nothing. A domain name is a vector pointing to years of published work: hundreds of thousands of words, dozens of photographs, camera reviews, personal essays. The AI didn’t guess in a vacuum. It had access to the entire body of work I’d chosen to make public. My voice. My obsessions. The point is: I didn’t hand it components to arrange, but I did hand it a very specific signal.

What came back included everything I actually care about: the cat, the tea, the film camera, the books (Cartier-Bresson, Freeman, the entire visual canon), the light, the sense of someone who thinks carefully about seeing. All of it inferred from that signal.

What It Actually Made

It understood the assignment. The cat was there, alert in a way that suggested agency. The tea was steaming. The camera was right; not some generic film camera but rendered with the kind of specificity that suggested it had looked at actual rangefinder cameras. The books were there, their spines readable, the right titles. The light was cool and controlled, nothing warm or golden.

And then there was the bridge. A perfect Parisian arch over dark water, stone weathered by actual centuries. It was impossibly Parisian.

Nantes doesn’t have bridges like that. We have the Pont de la Madeleine, which is a lot of things: industrial, efficient, carrying traffic from the island to the mainland; but it isn’t that. It doesn’t have that romantic geometry. The AI had never seen Nantes. It had seen Paris. And when asked to imagine where a photographer lives, it defaulted to the thing it knew how to render as picture-perfect.

What Actually Happened

Here’s the thing that stopped me: the prompt was just “I would like you to create a pen drawing that represents ijmphotography.net.” That’s it. No details. No aesthetic direction.

And it came back with everything: the coherent visual identity of someone who thinks in photographs. All of that was inference. The AI had to synthesize an entire person from a signal.

That’s what’s remarkable about where we are with these systems now. I didn’t hand it components to arrange. I gave it a category—a domain, a name—and it generated a complete aesthetic representation. It looked at ijmphotography.net and thought: This must be someone who thinks deeply about seeing. Someone shaped by a particular visual tradition. Someone who probably drinks tea and owns a film camera and has read the books that matter.

It got that right. Unsettlingly right.

Then it added the Parisian bridge. And that’s where the inference becomes visible. The AI had to place this person somewhere, and the logic it applied was: Serious photographer. European visual tradition. Thinks about light and geometry. That’s Paris. It didn’t know Nantes. It didn’t have the data that the person actually is Nantes, that the grey light and industrial past and the edges that don’t fit the romance are the point. So it filled the gap with the most statistically likely answer.

The bridge isn’t a mistake. It’s the moment where you can see how the inference works. The AI understood enough to synthesize an accurate aesthetic, but not enough to know that the specificity—the actual geography, the actual city—was part of what makes that aesthetic mean something. It knew what you looked at. It didn’t know where you looked from.

What Gets Lost in Translation

The weird part is how much the mistake tells me about myself. Because I would never have put Nantes in that image on my own. I would have defaulted to Paris too, probably. There’s something embarrassing about that gap: what I actually am (someone who photographs in Nantes, shaped by this specific grey city) versus what the visual language of my influences suggests I should be (someone drawn to the romance of European capitals).

The AI was doing what all of us do: trying to shorthand who we are by reaching for references that other people already understand. It works, most of the time. But it costs something. The specificity of Nantes: the industrial past, the way the light sits differently here, the edges of it that don’t fit romantic narratives; that becomes invisible. The machine sees the category you belong to and fills in with category defaults.

The Bridge and the Rest

I’ll probably use the image anyway. Not because it’s perfect, but because the flaw is more interesting than perfection would have been. The bridge that doesn’t exist is, in a weird way, more honest than an image of an actual Nantes bridge would have been. It shows the gap between how the world sees you and who you actually are: between what your influences suggest you should look like and what you actually look like.

And maybe that’s the point of asking an AI to show you yourself in the first place. Not to get an accurate portrait, but to see the particular ways you don’t add up. To notice where the machine defaults to assumption because it doesn’t have the specific data. To see yourself through the gap.

Everything essential remained. It’s just rendered in front of a bridge from another city, and somehow that’s more true than geography would have been.

The Uncanny Part

But here’s what I can’t shake: the AI inferred all of this from a signal. Well enough that I recognized myself in something made by a machine that has never seen me, never thought about photography, never sat with tea and a camera trying to understand how to see.

It got the shape of who I am right. It just can’t understand what that shape means. It saw the pattern and filled in the category defaults. And because the pattern was sophisticated enough, and the defaults were sophisticated enough, it worked. It looked like understanding.

That’s what’s uncanny. Not that the AI got it wrong. That it got it right by accident. That it can synthesize an identity without any capacity to comprehend identity.

The machine understands enough to be useful and unsettling in equal measure. It reads you well enough that you believe it knows you. Then you realize it doesn’t know anything at all. It just knows how to look like it knows.

It knew what I looked at.

It just didn’t know where I was looking from.

AI Isn’t Magic — It’s a Tool.

Posted on Posted in ReportageTagged , , , , , , , , , , , , , , , , , , , , , , , , 3 Comments

Here’s How I Use It Without Losing Myself.

A creative’s guide to using AI wisely — with the A.C.T.D. framework. 

I will seem a little controversial in this article, but AI (artificial intelligence) is becoming a constant in this creative world.  Do I use it?  Yes.  How?  When I don’t know how to do things, like speak Chinese.  When I need to have a copy editor to check my grammar and spelling.  When it comes to photography do I use it?  Not in the creation of a photograph, but sometimes yes to edit for me.  There are tools in Photoshop that are useful to the photographer like generative fill, for example..

But what is AI?  The large language models, the generative image tools?  Does it write code for developers?  Yes it can, but as most things it is a machine and never forget that.  It is learning, and think of it as being an eager student ready to learn.  It doesn’t get it right all the time, and it can, like any human, make mistakes.  It is a computer.  It does exactly what you tell it to do.  Does it understand British understatement,  sarcasm or banter?  No, because it still needs training.

But like most things, does it live up to the hype?  No, especially if you don’t know how to talk to it.   When I talk to people about how I use AI, I always share my A.C.T.D. framework. 

A is for Actor. I tell AI: “Act as my editor,” or “Act as my literary agent.” AI thrives on context — so give it one. For example: “Act as a specialist in vegan cooking.” (Yes, very controversial — I did warn you.) 

C is for Context. AI thrives on it. Give it background — what you’re trying to achieve, who your audience is, or why this matters. 

T is for Task. Be specific. Tell it exactly what you want — not “help me write,” but “write a 300-word intro about AI for photographers.” 

D is for “Think Deeply.” Ask AI to reflect before answering — in ChatGPT, click “Think”; in Qwen, you may need to select a reasoning-focused model.

And here’s the truth no one talks about enough: AI doesn’t create — it remixes. It feeds on human-made content — blogs, photos, code, songs, even tweets. And as anyone who’s scrolled the internet knows… a lot of that data is complete bollocks.

It’s trained on our best ideas — and our worst. Our genius — and our garbage. So if you feed it junk, it gives you junk back. If you give it context, clarity, and care — it can help you refine your own voice. Not replace it.

But tell me Ian what can’t it do?  It seems to be replacing everyone in translation and creative industries?  I get it, we all sorry about that, but you can learn how to spot AI content on the internet.  People once said a photograph never lies, but we soon found out that it could lie to us.  It can’t replace that personal interaction and accompagnement, such as sitting with a client before a shoot.  It can’t replace the feeling you get when you receive your prints from the photographer, and you realise that all that prep was worth it.  

I have met people who have told me that I have a beautiful camera, and that it must take beautiful pictures.  Well?  It can’t.  Like the person who says yeah but I can do that on my mobile…  Generally they can’t.  We still have our craft.  A computer can’t do that for us.  I can’t replace our artistic vision, despite trying to tell us that it can.  At best it can only provide an emulation.

I first wanted to learn about the internet when it first came out.  I wanted to know how to use it before my children would tell me.  That was over 25 years ago.  I approach AI in the same way.  It doesn’t have all the answers, but when well used, it’s a bloody good tool to have in your kit.

This isn’t about rejecting AI—it’s about reclaiming agency. When I give it clear context and a specific task, it becomes a mirror for my own creativity. But the song? The photo? The feeling in your hands when you hold a print? Those aren’t AI’s work. They’re yours. The tool sharpens the craft—but only the human holds the vision. So use AI like a darkroom chemical: a helper, not the artist. And never forget: the most revolutionary thing in your kit isn’t silicon. It’s you.

A illustration to a song written by AI about my Moly and my wife. That bitch who stole my man…

P.S. The song above? Written entirely by AI. It’s witty, structured, and even tugs at the heart—proof that AI can imitate craft. But it didn’t feel heartbreak. It didn’t choose to love a rescue dog more than its human partner. That irony, that ache? That’s still ours. AI remixes our stories. We give them meaning.