- Machine learning and AI are like super-advanced filters that tailor themselves to your images.
- Automatically “fixing” portraits will result in unreal body-image expectations.
- We haven’t yet found an AI duck-face removal tool.
Artificial Intelligence is the hot thing in photography right now. It means that your photo app does your editing for you, cropping, re-coloring, beautifying, and even changing people’s expressions. It’s making photography amazing, and it’s also ruining it.
AI and machine learning have already transformed photography, and with the new iPhone 12 Pro Max, apps like Pixelmator Photo 2, and Skylum’s Luminar AI launching soon, it’s set to get even worse/better. Have you noticed that the portraits you take with your phone are free from blinkers, and everyone is smiling? Or that the background is beautifully blurred? Or that all your photos are perfectly exposed? But the next wave of AI editing is here, and it promises to make your photos look amazing. But will it also make them look like everybody else’s pictures?
“No one cares about depth or soul or meaning anymore,” writes photographer Chris Gouge on Petapixal “It’s all about aesthetics and the hope that same other-worldly looking landscape or sunset will gain a lot of digital hearts or thumbs-up on the internet.”
What Can AI Do?
In photo editing, machine learning means that an app has been fed zillions of example images, and told to work out for itself how they’re put together. Then, it uses this training on your photos. In phone cameras, which have a computer attached, some of this happens before you even press the shutter—like smile-detection, for example.
“When editing photos, people usually spend 74% of their time on repetitive, routine tasks, which we call grunt work,” Skylum’s head of global communications, Maria Gordienko, told Lifewire via email. “Because of the boring nature of this grunt work, people come to think of photo editing as a more difficult and less satisfying process than it really is.”
Then there’s auto-cropping, and other basic edits. Pixelmator Photo 2.0, for example, has a machine-learning-powered “enhance” tool. “Many of the most important adjustments can be applied automatically, using a machine learning algorithm trained on 20 million photos,” says the product page.
But then things get murkier.
The Dark Side of AI
Deep fakes are one clearly dangerous use of AI photo editing, but there are less obvious possibilities. For instance, Skylum’s upcoming Luminar AI can analyze a portrait, then retouch it automatically. It can sculpt lips, narrow a face, change the irises, and remove skin blemishes with a click.
Taken alone, these seem like minor problems. Or maybe you love the sound of an automatic beauty filter. But what happens when almost all images have been tweaked like this? We worry about “photoshopping” in commercial photography. Ads with slimmed down bodies and perfect skin that we can aspire to but never attain. Instagram is probably the place where most photos are shared. What happens when our AI-enhanced selfies take it over?
“It’s what people want, I think. So many people seem desperate to look like each other—apparently inspired by fish faces.”
“I don’t really care what people do to their images when it doesn’t impact me/society,” photographer and writer Hamish Gill, the founder of photography blog 35mmc, told Lifewire via Twitter. “People have been editing practically since the beginning of photography. This stuff makes it easier, then great.”
And how about this: Tuning portraits to make the subjects look attractive sounds fine, but the AI edits that make people attractive are obviously based on sexual cues. That’s what we mean by “attractive,” after all. On adults, this is problematic enough, but what about the pictures of kids that pass through these filters?
It’s not all just moral outrage and body-image issues. AI photo enhancement has another downside: it makes all photos look the same. That’s just how it works. Now, Instagram influencers are already copycat cliché lovers, but even so, AI will take this to a new level. It won’t just be me-too subject matter, but me-too edits. Perhaps, though, this is exactly the point.
“It’s what people want, I think,” says Gill. “So many people seem desperate to look like each other—apparently inspired by fish faces.”
AI for Good
AI can also be a very useful tool. If you’re tasked with removing a surprise pimple from thousands of pictures in a photoshoot, then would you prefer to do it manually, one pimple at a time, or have software take care of it for you? And it could be argued that AI is just a slightly fancier filter, which is itself just a collection of edits saved into a preset.
I have a preset that mimics the looks of Kodak’s B&W Tri-X film, and then applies grain. Usually, I have to tweak the lightness of the image manually. What if an AI tool could learn how I apply these edits and do it for me. Is that a valid time-saver, or does it make all my future photos look like the previous ones?
“When editing photos, people usually spend 74% of their time on repetitive, routine tasks, which we call grunt work.”
In Luminar AI, says Gordienko, “Beginning editors can find inspiration through recommendations on how to edit their photographs. Experienced editors exercise more control, selectively applying AI tools as they edit their images.”
Like any tool, there are good or bad sides, but in the case of AI photo editing, it seems like the possible perils will be magnified to far outweigh the advantages to photographers. And for what, really? You’re not making your photos better. You’re making them similar.
The captions in the demo video for Luminar AI say it best: “Photo editing is tedious, stressful, and complex.” Why put in any creative effort when you can let the computer fake creativity for you?