7 min read

Our Man Not In Havana

Our Man Not In Havana
Photojournalism's deep dreams are made of this (detail from a post by @Michaelchristopherbrown)

A reformed Magnum man proves that old habits die hard

It's hard to move in Photoland at the moment without bumping into some pearl-clutching goof asking what it means when AI steals their job. The latest iteration of this (though by no means whatsoever the worst or the biggest) is Michael Christopher Brown and his 90 Miles, a project so weird I have returned to it multiple times over the past couple of days and have yet to comprehend what possessed its maker to publish it. (That is, unless you count the NFTs it's slinging, because of course it's a testudo of solidarity slogans concealing a crypto grift).

The title refers to the narrow strip of water separating Cuba from Florida, a castle moat which has claimed the lives of so many of those who have tried to cross it on improvised vessels. Describing his decision to ask Midjourney AI create a series of images of Cuban life, as well as to imagine what the journey might look like for those fleeing the island, Brown states;

"Anyone may now create lensless, photo-realistic reportage illustrations on any subject anywhere, at any time, collaborating with a collective history of photography to illustrate the photographed world and create a vision of what was, is or can be."

The project is being pushed by AIRLAB, a company whose website explains that it was "created in response to the growth and influence of A.I. and to understand ways of utilising the technology in reportage illustration." AIRLAB is in the business of making and selling NFTs, in case you were wondering.

Imagine the prompt which had to be rationalised in order to arrive at this image

In principal, 90 Miles is actually the sort of use-case which you can imagine Silicon Valley seizing on to explain the virtues of neural net image generation. A photographer wishes to produce images about a topic, but cannot because of restrictions placed upon reporters in the state where the images would need to be made. The topic at hand is important, and thus showing people pictures of it ("raising awareness") must be inherently virtuous. So the photographer turns to the benevolent and objective gods of artificial intelligence and ask them to dream up images of fictional humans in fictional situations of fictional terror and distress which can then be circulated to raise awareness of real humans in real situations of real terror and distress.

The resulting images are exactly as disturbing as you'd imagine them to be. Traumatised faces and imperilled brown bodies cling to disintegrating rafts in stormy seas. Men with deep-lined faces and sailors' tattoos worry about unseen threats, while spasmodic violence in a tropical anytown draws a crowd of nearly-human figures. Fidel Castro makes an appearance but he is played by a lookalike, and while the prosthetics are good, his face is strange. Because neural networks can only regurgitate what they are fed, the refugees take to sea in vessels which resemble vintage cars and buses but are actually only smooth-cornered boxes with period greebling. The colours are perfect, the eternal golden sunshine and Kodachrome pastels of National Geographic's Havana, and over everything looms palm trees which look almost normal, but not totally.

Raul? Is that you?

At the Pakhuis De Zwijger in April, Babusi Nyoni gave a presentation about the potential of Midjourney and other generative imagery, saying;

"It's the worst now that it will ever be. It will only keep getting better and better."

In recent months much ink has been spilled about the visual strangeness of generated imagery, so I will assume you've already read other detailed visual critiques. Suffice it to say that, as photographs, these are no different- they're fine but not great, and as usual the hands are weird. They resemble photographs from Cuba because they are in large part made up of images of Cuba. They're composites of millions of images of Havana, of Castro, of misery and decay and vintage cars and tanned joggers and sunsets over seascapes, all melded and blended and then gobbed out as a new image. Yes there's a copyright issue there, but I came of age after the death of that particular business model so colour me uninterested. Instead what I want to talk about is the same thing as Brown (though for different reasons); what do these images mean for photography?

Specifically what I am fascinated by in these images is what they reveal about the production of a specific kind of photography. Let's call it the shallow dive, a photo essay from a far-flung place made by a person with a platform. Brown describes the need to bear witness to Cuban suffering, but in creating a shallow dive through Midjourney he unwittingly reveals something about the form. These images are creations of the mind of a photographer just as much as any fictional image exists first in the mind of its maker. That is to say, these images depict the scenarios which Brown would seek to photograph if he were allowed to photograph freely in Cuba. They show the things he wants to photograph, and the ways in which he would photograph them if he could. They show the topics he would choose, the perspectives, and even the framing and lighting.

National Geographic intensifies

I haven't seen his prompts, but I would guess he specified scenarios, colour palettes, subject types and lighting, before tweaking and combing the results to find images which matched the scenes he saw when he closed his eyes. He is making photographs, but instead of going out at sunset with a camera he is instead asking a machine to assemble a collage which resembles what it would look like if he did. These are 'real' photographs in every sense except the mechanism of their creation. We can therefore deduce that, were he granted free movement with his camera, his photographic process would be one of collection, not exploration, and it's unlikely that the dive would have reached much deeper even if Brown had actually boarded a plane to Havana.

It is this checklist approach to the story which reminds me of the business of international photojournalism. The two processes seem eerily similar, minus the peril. Stick a pin on a map- a country, a story. Fly in, then ask a fixer where the action is. Go to the protest, the site of a recent bombing or wherever the shooting is. If it's a migrant story, exchange riding to the front line in a beat-up Corolla for taking a Zodiac out past the breakers and scanning the horizon for orange dots. On days when there's a lull, enter the home of a local family and photograph a quiet feature about ordinary life in war's shadow. The client pays by the day so enough time is never given, the fixer charges by the day so enough time is never offered. The highest profit comes from making the most iconic images in the shortest space of time. They don't need to be deep, just iconic.

Now we've got you feeling emotional about some imaginary person's imaginary grandad and his melty melty arms.

That's because the job of the roving photographic reporter is to act as the proxy of the reader and translate often-bewildering realities into digestible visual forms. In this moment of rolling deadlines, shrinking budgets and scrolling feeds, it's no surprise that the form leans heavily on symbolism, and Midjourney images are just a faster way to achieve that same intention. This giant sightless eye has consumed all of visual history and now reflects back at us all the beauty and horrors of ourselves, teasing our dreams and fears from the ether into solid form, like cotton candy on the salesman's wand. It composes, stages, lights, positions and poses. It adds peril, or salvation, or synthesises expressions which can be read as displaying certain emotions. But it doesn't actually discover anything. It can only repeat what it is told, can only trace pictures on top of other pictures, and if a given result is particularly compelling that is because the commands which summoned it were made by a prompt engineer with a keen eye for cliché.

We should not be surprised that successful photographers turn out to be competent prompt engineers. The problem with these images is not the images themselves, it's what they reveal about the industry. If you're a photographer creating work which speaks to those it depicts, which surprises or reveals or even just truly, honestly witnesses, this bell tolls not for thee. Money-wise you're still going to have to eat shit the same as you ever did, but it's not going to get worse because oF tHe RoBoTs. Community centres will not start exhibiting work by AI, because AI can't envision and run a community storytelling project, in part because it doesn't care.

There's a reason why all the weeping noises are coming from the jet-set brigade- their business is built on the shallow dive, and it is precisely the quick-sketch impression story that these machines render worthless. And yes, that might mean that Land Rover chooses to save money and plump for an AI campaign to advertise its latest crude metaphor, but maybe that's for the best.

Bonus: Read Guillaume Bonn whingeing about the same topic here!

We would like to commission somebody to write an article about instagram accounts like this one and what they mean. Would you like to write it? Would you like to write something else? Just hit Upload!