Photography and A.I.

It was bound to happen. At some point, I had to write down my thoughts on the evolution of AI.

When I talk about A.I., I don’t mean the Artificial Intelligence found in Lightroom, Photoshop, or any other software. You know the type—select subject, select background, etc. It might surprise you, but these are tools I honestly use very rarely, if ever. I still go by the Dodge & Burn principle. You’ll read later in this article why that is.

When I talk about A.I. here, I mean the kind of A.I. that is making the job of a photographer obsolete—software like MidJourney and DALL·E.

Am I anti-A.I.? Definitely not. Most people are anti out of ignorance, and that’s exactly what convinced me to look into it and experiment with the models mentioned above.

WHY WOULD I WANT TO CREATE FAKE PHOTO'S

A long time ago, I graduated as a computer scientist and have always been interested in the latest technology—especially when it involves images.

During one of the many ads on social media, I saw a photo of E.T. pass by that I wouldn’t mind seeing on my wall. Yes, secretly I’m a fan of that sweet little extraterrestrial. Unfortunately, I never got a reply to my many emails, and then I thought: maybe I can create one with A.I. myself. Using DALL·E version 3, which you can try for free via the Bing Image Creator, I typed: “Fine Art Black and White studio portrait of E.T.” and soon I got an astonishing result.

You can understand that I was—and still am—hugely impressed! What a beautiful result, and I could never have photographed it like that myself. Especially since E.T. has been back on his home planet since 1982.

The images above were generated by DALL·E from OpenAI, the same company behind the well-known ChatGPT, which I’ve been using extensively for several months now. Apparently, MidJourney is considered much more professional when it comes to photography… uh, I mean… generating images.

There was another (artistic) image I wanted to hang on my wall but couldn’t find anywhere on the internet. So I decided to have that one generated as well.

Made with MidJourney Prompt: an artistic portrait of a happy bordercollie on a yellow background, splatter paint art style

Made with Dall-E v3 Here I wanted a combination of my two dogs: a black-and-white border collie (brown eyes) and a red merle border collie (blue eyes).

These last two images are, of course, not photorealistic—and that wasn’t the intention. But I think you’re starting to see the power of A.I.

Let’s take it one step further...

WILL A.I. REPLACE THE PHOTOGRAPHER?

Yes, unless regulations are introduced. I’ve already lost part of my income because of this new technology. Do I hate this new evolution because of that? Absolutely not. Automation has made many jobs obsolete in the past, so it was only a matter of time before the artistic sector was affected as well. What’s important is to keep up, understand it, and stay flexible.

Part of the income I’ve lost comes from stock photography. About 15 years ago, I uploaded many photos—photos I’m not proud of now—on Shutterstock to sell. Back then, this generated a decent income, and even though I haven’t been active in it since 2006, I still receive a check every year.

If you visit Shutterstock, you’ll see that customers have two options. Either you search among the existing photos for an image that somewhat fits your campaign, or you type your search terms into the A.I. section, and a photo is generated for you—free from portrait rights and all other administrative hassles. A.I. generates entirely fictional characters.

Let’s put this to the test and compare results from existing photos on Shutterstock with those generated by A.I.

We search on Shutterstock for images that I might use in a commercial campaign: “a woman behind the computer following an online course photography.” Unfortunately, there were zero results. We simplify the search to: “a woman behind the computer” (click here to see the results on Shutterstock yourself).

In MidJourney: a woman behind the computer

At the moment, I would still rather purchase a result from Shutterstock. Nowhere in the photo do I see a reference to photography—not on Shutterstock, nor on MidJourney. Let’s adjust the prompt (instruction) for MidJourney a bit.

Prompt: a woman behind the computer studying an online course photography, with a professional camera next to the computer and two landschape photography prints hanging on the wall behind her

Not exactly what I asked for, but I would now be more likely to use this one for the campaign than a Shutterstock result.

A.I. AS A NATURE PHOTOGRAPHER

Let’s take the computer out into nature and generate a photo of a robin. By default, the system produces square results unless you set it otherwise. I kept my prompt simple and typed: “robin on a branch.”

Dall-E : pretty but not really photo realistic

Midjourney: scary realistic

Let’s continue with MidJourney and create a more detailed prompt/instruction:

 “a robin on a branch, shot with a 600mm on F/4, nice bokeh and green background. Small warm backlight and a little rimlight on the bird.”

Scary....

WHAT NOW?

I think I speak for everyone when I say that we enjoy photography and would rather photograph a robin out in the field than generate a virtual one. For professional photographers, of course, it’s different—they have to make a living from it. That’s why I want to closely follow everything and see where I can differentiate myself so that my images don’t end up looking too much like A.I. creations.

PERFECTION IS THE KILLER

A.I. aims to do things as perfectly as possible, and that’s where we, as image-makers, can still make a difference. Perfection has been a bit of a trend in the digital age, although I’ve noticed a shift over the last two years. Many image-makers (photographers and filmmakers) are trying to move away from that perfection and want to put their soul back into the photo.

Generated by A.I.

One of my images

Let that perfection also be the reason why I never use A.I. in Lightroom. That “Subject/background select” is just too perfect for me—I prefer it when you can see that I’ve “painted” the image. I can even appreciate a sensor spot in some photos. Everyone has their own taste and preference, and thankfully so!

CONCLUSION

They’re not quite there yet with their A.I., but it’s definitely frightening. Keep in mind that the system improves every two weeks, and we’ve only been working with these kinds of models for just over a year. It will only get better and better, and in a year, it will be a completely different story. Just look at how Lightroom has evolved compared to LR3.

As a photographer, I’m convinced that we can still make a difference. We are storytellers, and each of us has lived a unique life and had unique experiences. This ensures that each of us has our own vision of a subject. And it’s precisely that vision, that story, that makes us unique.

To finish, here’s an image of Tuscany... I’ve never been there myself ;-)

What do you think of this evolution? Will you use it to create images?