I recently posed a challenge to DALL-E 2 (or to myself?) to mimic one of my photos using prompts. I didn't think it was actually going to work, until it did. Almost.
The left part of the photo above represents AI's interpretation of a "photo of Fire Island Lighthouse during the day with few clouds and common reeds with shallow depth of field in foreground" which is about the best way I could describe the actual photo I took with a Canon EOS M6 Mark II and Sigma 56mm f/1.4 DC DN Contemporary Lens.
To get that photo, I drove out about an hour to Robert Moses State Park's parking field five. I parked my car, carefully picked out which lenses to bring with me for my trip (and yes, I also brought along Camera brand camera) and then walked about a mile along a boardwalk to get out to the lighthouse, and then walked off the path along some dirt trails to find this interesting scene from amongst the reeds. I made the conscious decision to shoot at ISO 100 and set my aperture to f/1.4 to get the shallow depth of field. I had to think to put a 3-stop neutral density filter on my lens so that the camera's maximum mechanical shutter speed of 1/4000 would be able to expose the scene properly.
But I'll be damned if the AI didn't come close enough to make me wonder: In a couple of years when the computers actually get it right, will it be even worth it to get "the shot" when "the shot" could be produced with just a few clicks of a keyboard?
To be fair, AI is a long way off, and in the case of the main photo in this article, I think it was a combination of a detailed enough prompt and some luck. Sometimes, I struck out, like with this photo of the Montauk Point Lighthouse that I tried to recreate with AI:
Some may argue, especially in the case of recognizable landmarks like these two lighthouses, that the AI is essentially just stealing images and tweaking them a bit. While it's impossible to tell what's going on behind the scenes of the software, I wouldn't be surprised.
That said, I'd never be able to photograph a "Photorealistic image of a t-rex wearing sunglasses driving a red convertible car through a machine car wash," and so I've got to hand it to DALL-E 2 on that.
For me, the answer will always be yes, as I want to experience and see the very thing I'm photographing, which is something AI will never replace. But is that the case for you? Is having the photo produced by an AI prompt good enough for you?
Leave your thoughts in the comments below.
One of the results that Midjourney spat back out...
Came here to say that Midjourney v5 is in an entirely different league from Dall-E or any of the other AI image generators.
Pretty much Jurassic park 1993. - I don't mean this in a pejorative way. I am trying to understand the connection between you and the image. Technically, it is not your creation so where does the personal relation and feeling start and end if any. Had you shot this, I wouldn't make the 1993 remark or any comment.
Had I shot this, I'd have been someone's lunch. :D
The connection between me and the image is that I took the prompt in the article, and plugged it into Midjourney 5. I then shared the result here so that people can see the difference between what Dall.E 2 can do, vs Midjourney. There's absolutely nothing else that needs to be read into this. :D
I don't know that I can agree. You don't have the sunglasses and your car is not a convertible. I see water in a car shop, not a carwash. But you have photorealistic on your side with a very strong Jurassic Park look. Could be that Dall E filters out movies and is more careful regarding copyrights. I google searched something as simple as Jurassic Dinosaur and something like 2/3 of the images that came up are from the movie for example. Learning from others images happen to be the basic process how those AI apps learn. Apart from the imitation aspect, yes the 2 images are drastically different but they don't seem to follow the same rules.
How much satisfaction will 'just a few clicks of a keyboard give you"?
AI is just totally fake imagery and always will be.
AI will be the panacea of, or death of, the crap artist.
Part of it is your decision to be known for your "AI creations" or as a photographer. I think testing is one thing but making too many can hurt too, something that probably cannot be undone 100% ever.
I was following a person who was doing somewhat interesting photography. One day when Dall-E was hot news, that person started showing extremely creative images and the next ones were even more crazy. Over time 50 to 100 were posted. I asked if AI was used but just got a vague reply. From there, I noticed the likes were going down including mine. Eventually the person started posting real images but I have lost interest even in that.
For me it is simple. Even if I am only traveling for work (200 miles or less one way), I like to photograph what I see. It is part of the experience of going somewhere and seeing something new. If I want to create something that doesn't exist, I'll take up drawing or painting. I may not be good at it, but it will my work.
Photography as an activity is about more than the final image, at the risk of sounding cliched it's also about the journey.
AI may be able to create interesting images, particularly of places that have already been photographed, but the images are not "real" and that authenticity still has a draw. If that remains the case in the future or not... I expect that people will turn against AI generated images in a few years. Already you can see that some people are valuing authenticity to the point of dogmatism (severely limited edits only).
I spent a large sum of time and money on my last trip to take some photos in a jungle. Could AI create a passable image replicating one of mine? Sure. Wouldn't have enriched my life, given me many more photos that I couldn't have preconceived or allowed me to meet some interesting people or become enamored with the destination though. The experience is part of the photo.
"Some may argue, especially in the case of recognizable landmarks like these two lighthouses, that the AI is essentially just stealing images and tweaking them a bit."
Do you think?
Let's try to remember that this so-called AI never sees the real world - just images created by people. Its only source material is the work of artists and photographers.
I think it is not valid to ask that question without referring to the goal of aquiring that picture.
If I want this picture just to have it, then of course using AI is as good as driving and shooting (assuming the AI is good enough one day). But then I can just look on Flickr or Instagram and download such a picture from any photographer. Then at least it would show the real thing.
If I want this picture in the sense a photographer would, the question is not a question at all as AI would be excluded as a means of aquiring it.
Oh I was just going to say try again on Midjourney but Susheel Chandradhas beat me to it
I sure as hell ain't going to let AI 'mimics a photo that I, me...personally, want to take. What's the point? There's much more to Photography than people actually think and, to me, getting ready, preparing all the necessary things to take photos, which includes travel arrangements, is one of the more exciting and anticipatory things involved. The thought of getting out on the road/air/water/train only builds more desire to go out time and time again. From this article, I would feel like a lethargic Android just clicking here or there to try and emulate the beauty of what I wanted to see in person...with my own eyes, to feel, with my own heart and experience with my own soul. That can never be achieved through AI. And never will.
When I review my several decades worth of photography, each image represents a fraction of a second of a particular holiday or outing. The photograph brings back the associated memories of how I felt, what we did before, what we did afterwards and a vast panoply of emotions that an AI generated image is never going to replace. So even when AI can perfectly reproduce any photograph I can imagine, I'm still going to want to travel and take my own photographs, flaws and all.
Reminds me of the 80s, when sythesisers could "create any sound", and were only "a couple of years" away from "making obsolete" all the traditional instruments. Turned out that synthesisers were useful to create their own unique, recognisable sounds, and became just another tool in some musicians' arsenals.
I think AI will end up being the same for photography; it will be simply another tool, with its own distinctive style, that some will choose to use.