I recently set up a challenge for DALL-E 2 (or for myself?) to mimic one of my photos using a prompt. I didn’t think it would really work, until it did. Most of.
The left part of the photo above shows the AI’s interpretation of the “popular daytime shot of Fire Island Lighthouse with few clouds and reeds with a shallow depth of field in the foreground”, which is the best way that I can describe the actual photo that I took with One Canon EOS M6 Mark II And Sigma 56mm f/1.4 DC DN . Contemporary Lens.
To get that photo, I drove about an hour to the number 5 parking lot of Robert Moses State Park. I parked my car, carefully choosing which lenses to bring for my trip (and yes, I brought it too Camera brand camera) and then walk for about a mile along the boardwalk to the lighthouse, then walk along the trail along some dirt roads to find this interesting sight among the reeds. I made the wise decision to shoot at ISO 100 and set the aperture to f/1.4 to get a shallow depth of field. I had to think to order a 3 stops neutral density filter on my lens so that the camera’s maximum mechanical shutter speed of 1/4000 can properly expose the scene.
But I’d be damned if AI didn’t come close enough to make me wonder: In a few years when computers really get it right, is it worth the “shot” when the “shot” can be produced with just a few keyboard clicks?
To be fair, AI is a long way off, and in the case of the main photo in this article, I think it’s a combination of a detailed enough prompt and a bit of luck. Sometimes I let go, like with this shot of the Montauk Point Lighthouse that I tried to recreate with AI:
The reminder for this shot is: “Low angle shot of Montauk Point lighthouse covered with white Christmas lights with a long exposure of water flowing over rocks and lighthouse reflecting in the water at dusk with clouds.” It looks like Montauk Lighthouse has seen better days in the AI image, however, as in the Fire Island Lighthouse photo, it seems the AI is too far away to really figure this out.
Some might argue, especially in the case of recognizable landmarks like these two lighthouses, that the AI is basically just stealing images and tweaking them a bit. While it’s impossible to know what’s going on behind the scenes of the software, I wouldn’t be surprised.
That said, I’ll never be able to capture “A true-to-life image of a tyrannosaur wearing sunglasses driving a red convertible through a car wash”, and so I have to hand it over to DALL-E 2 on it.
However, with the ability to create worthy (possibly) photos in the near future through AI, is it worth the hefty expense to make one?
For me, the answer will always be yes, because I want to experience and see the very thing I’m shooting, which is something AI will never be able to replace. But is that your case? Is having a photo generated by an AI prompt good enough for you?
Leave your thoughts in the comments below.
