Artificial selection

Artificial intelligence in photography? What happened to the ordinary guy?

AI is everywhere. Machine learning is, by implication, great because it means we don’t have to do any learning ourselves. Deep learning is the new alternative to what we actually do, and neural networks are an incredibly technical alternative to what our brain does while it sleeps. All this, just to save us thinking.

In photography, AI is used for everything from selecting objects to choosing a preset, determining focus, to optimizing camera settings based on a scene. It’s about making a machine do something so that we don’t have to.

The latest pro cameras like the Canon EOS R3 use AI to identify and track specific objects like birds or, in this case, cars. It has the potential to revolutionize sports and wildlife photography. (Photo credit: Canon)

(opens in a new tab)

When AI can really make a difference

AI can, indeed, do many complicated things much faster than us. It might mean focusing on a human eye, or a bird, or a train, or a blueberry (soon, probably) faster than we ever could. It’s easy to see how this could be useful. This leads to a revolution in sports and wildlife photography and the ability to track fast moving subjects and keep them in focus.

It’s not really new. Cameras have had subject-based tracking modes for a while – you identify an object or area and the camera can track it around the scene. What’s changed is that the cameras now know what those objects of interest are (lots of them) and can locate them for you in frame. This was accompanied by much more powerful AF algorithms and faster, more powerful AF actuators in the lenses.

The second is that the latest photo editing AI can save us the effort of manually making complex and difficult selections. Skylum Luminar leads the pack here, with AI sky selection and replacement, AI face and feature recognition for portrait enhancement, and subject and background masking. plane by AI.

Other software does the same thing. Photoshop has its own AI sky replacement feature, and Lightroom has an extremely effective sky and subject masking tool that’s eerily and unerringly accurate.

Lightroom now has AI-powered auto subject and sky masking. It is incredibly efficient and simplifies a lot of editing jobs. (Image credit: Rod Lawton)

(opens in a new tab)

When AI starts to take over

So far, so good. AI can certainly help us achieve our creative goals more easily. But what happens when it starts interfering with the creative process itself? Is the AI ​​starting to tell us what to shoot, how to shoot it, and what it should look like when edited?

Luminar will analyze the content of your photos using AI, then suggest templates (presets) to make them look beautiful. Lightroom offers AI-powered preset suggestions.

This all sounds innocuous enough, and perhaps great for beginners who are still exploring different visual styles, but it also leads people to a broad and generic “taste” suggested by a machine learning algorithm.

Are we perhaps heading towards a future where every image is “good” but the same. A future of universal competence, with visual value determined by an algorithm, coded by people who have never taken a photo but know how to analyze data thoroughly.

It’s easy to fall into this idea that there is a “good” universal selection, a “good photograph” or a “good” choice of style. It’s a great way to breed a breed of imagery automatons and ensure that we never have Vivian Maier, Henri Cartier-Bresson, Annie Leibovitz, or Bill Brandt again.

I like this image, but if I presented this as proof of my talent as a bird photographer, it would be a big lie. The place is real, the sky and the birds have been added – it barely took a minute. (Image credit: Rod Lawton)

(opens in a new tab)

Photography or forgery?

AI didn’t invent photo tampering, but it certainly made it easier. Personally, I’m obsessed with Skylum’s AI Sky replacement. In moments, you can transform ordinary scenes into something surreal, dramatic or beautiful.

But it’s good. The photograph does not have to be a forensic record of reality. Some photographers like to record what is there and what is real; others like to make images that evoke ideas or moods.

AI sky replacement, portrait enhancement, and masking are great tools for creating images of the second kind, but of course there is a crossover. Quite often a landscape will look better with a different sky, and a portrait painter might be much happier with a modified version of themselves.

This is where it gets tricky, especially in the new “influencer economy”. It’s now very easy for anyone to fake a photo while presenting it as “reality” (often because they don’t know any better). You can look good, your travels look amazing, and your life is wonderful — and all the time, your AI-enhanced photos are presented as proof of those things. It’s a bit nasty.

AI autofocus can automatically find your subject’s eyes, and your AI software can select facial features for individual enhancement. (Image credit: Rod Lawton)

(opens in a new tab)

Do we want likes or originality?

Are we really so lazy, anxious, glory-hungry, and ignorant that we’d willingly let a machine do the thinking for us? No one wants to take a bad photo, but it’s surely a risk worth taking over the alternative – surrendering to AI-driven universality where your best photos are the same as those others and you willingly trade your individuality for popularity.

It’s not all the AI’s fault. Social media algorithms are also to blame. They led us to imagine that popular is good, and instant impact is all that matters. It’s nice that AI helps us be more individual and creative, but not if all it does is reduce photography to a bunch of popular memes.

The history of photography is a delicious mix of the kitsch, the popular, the wild and the individualistic. We have to keep it that way.