Cameras used to use glass plates and flash used to be controlled explosions. Thankfully, we've embraced the relative safety introduced to the photographic process with DSLRs and mirrorless cameras, but can we say the same for AI?
TEDx UNC-Charlotte speaker Manav Majumdar, a student at the school, raises some interesting questions that all photographers should consider about AI.
Too often, new technologies are dismissed because they destroy an industry that was thriving in the old way. Whether it's electric versus gas cars or film photography versus digital, new is often scary.
But in reality, photographers have been using photography for a while, just not in obvious ways. For instance, software such those from the likes of Topaz have been resizing, sharpening, and denoising images for years using AI models. Where photographers started raising eyebrows and crying foul has been when AI began to be able to generate images from seemingly thin air, with the likes of Midjourney and DALL-E 2. These were often accused of cribbing from photographer's existing images to produce "new" work.
But Majumdar talks about other uses for AI that shouldn't be dismissed. For instance, he talks about the auto-culling and light editing of photos that can be done with AI, processes that save time, freeing a photographer from these mundane tasks to shoot more. Majumdar points out that AI can never replace getting a photo in focus, capturing the right moment, and all other things that lend authenticity to a photo.
I myself have found myself using AI in this task-oriented way. Instead of generating entire elements in photos, I've used it, I feel, judiciously. In my portrait work, I've used it to remove distracting background elements, simply highlighting and typing a command to do this. It's saved me hours of time that I would have had to do it manually. I've also used AI to extend the canvas of an image when I've made it just a hair too tight for social media or profile picture posts.
As Majumdar mentions, we've entered the "AI Age" whether we've liked it or not. The question is: how will you use it?
Back in the late 70’s, 80’s and until the great decline of film manufacturing, photography was already very complex and advanced with labs. In a way it was the analog then early digital version of AI. Those big labs most people have never heard about were capable of processing film, expose for prints in extremely large quantities, in the order of 20-50 thousand rolls in 5-8 hours/lab. That’s how they got delivered next day. Rolls were 24 or 36 exp so on average, 20k rolls x say 30exp for a smaller lab, that was 600000 exposure a night divided in a small amount of machines. Those rolls were for the most part over and under exposed, so you would hear the click variations from the shutters from frame to frame. The goal was to sell prints and the machines were quite good at bringing corrections, exposure and color filters extremely fast to be acceptable to the client and generate money. Minilabs did that too at slow paste. Scanners in prepress were able to analyze color and exposure the same way by the 90’s. The software Lino Color was the king of all, way ahead of it’s time.
Same story regarding advanced technology in photography with digital captures. A lot of people know that the first digital camera was created by Kodak and then nothing pretty much happened until the early 90’s when Kodak partnered with Nikon and then Leaf came out with it’s first professionally usable CCD digital back. In reality, the first portable digital camera was a Sony in 1981 and it was pretty much an SLR and looked like one. The only thing was that Sony build the Mavica for video (27 years before Canon’s 5DM2!) and had a multi frames type of image recording. However, the camera was actually capturing single frames exactly like a still camera. After that, Sony went to the VCR type of recording (invented in 1956).
My point is, AI isn’t new at all in photography. The one stealing images to generate outputs is the big talk, but even cellphones have used AI for over a decade. The magic wand in PS is nothing new, far from it. AI has been around for awhile for sure it’s just that to sell it as new thing, they gave it a name that sounds modern, advanced and all. It’s nothing but a gimmick and in reality is just evolution thanks to the larger memory and processing power available today.
"Artificial Intelligence doesn't exist in photography software"
Of course it does. AI-Denoising in Lightroom Classic for example use machine learning algorithms (as refered to Artificial Intelligence), the generative AI "firefly" in Adobe Photoshop Beta generates entirely new image contents.
The big problem here is to differenciate between
Generative AI and not generative AI.
Generative AI is drives Dall-E, Stable Diffusion, Adobe Firefly, chatGPT etc by "stealing" other peoples contents like images or texts, learning from it and generating entirely new images or texts by a user promt.