Photo Threads: Using AI OK?

[Darmok N Jalad]

Ideally, I want the final product to match my perception of the scene when I took the photograph, and I will manipulate accordingly. What that doesn't mean is changing the scene to be what I think it should be. For example, if I didn't get the right sky, well, I'm not going to swap it. Once you start going all-out, it eventually becomes a "why bother taking the photo in the first place" situation. Part of what makes any photo "special" is to manage to get something great out of what you have to work with. Once you start idealizing it, it just becomes less special to take photos at all. Might as well just ask OpenAI, Gemini, or whatever to generate the idealized photo for me.

Say I go to the Grand Canyon, and I don't get the sunset I was hoping for. Well, I could take a photo and switch the sky, but then that's not my memory of the place. Instead, I'll just take the best shot available. If I did swap it, in 20 years, will I look back and even fool myself? No, I'd rather remember it the way it was. Now, removing strangers from the photo is different, since I don't know them, they are not my reason for being there, and I might not be able to wait for them to not be in scene.

Just my 2 cents, of course. I just don't want to see machines rob humanity of its creative gift. It's one of the most rewarding things about being human, perhaps its defining characteristic, IMO.

 

[BanjoDudeAhoy]

Ideally, I want the final product to match my perception of the scene when I took the photograph, and I will manipulate accordingly. What that doesn't mean is changing the scene to be what I think it should be. For example, if I didn't get the right sky, well, I'm not going to swap it. Once you start going all-out, it eventually becomes a "why bother taking the photo in the first place" situation. Part of what makes any photo "special" is to manage to get something great out of what you have to work with. Once you start idealizing it, it just becomes less special to take photos at all. Might as well just ask OpenAI, Gemini, or whatever to generate the idealized photo for me.

Say I go to the Grand Canyon, and I don't get the sunset I was hoping for. Well, I could take a photo and switch the sky, but then that's not my memory of the place. Instead, I'll just take the best shot available. If I did swap it, in 20 years, will I look back and even fool myself? No, I'd rather remember it the way it was. Now, removing strangers from the photo is different, since I don't know them, they are not my reason for being there, and I might not be able to wait for them to not be in scene.

Just my 2 cents, of course. I just don't want to see machines rob humanity of its creative gift. It's one of the most rewarding things about being human, perhaps its defining characteristic, IMO.

Really well said and just about my stance as well.

There are tools for denoising that I consider okay for photography. Removing certain blemishes (caused by the lens or whatever) I can get behind as well.
As soon as we get into adding stuff that wasn't there to begin with, removing/replacing large parts of the image with "AI" generated stuff... that's crossing the line for me. That's no longer the photo you've taken.

The way this stuff works, you're using someone else's picture(s) and pretend it's yours. Not cool.

Regarding tools - I'm also not entirely sure where I stand on sharpening with "AI". If your image is juuuuust out of focus and an unsharp mask or some highpass trickery would have fixed it, ok. That probably didn't need "AI" in the first place.
If it's so blurry it's hard to make out details but the "AI" tool makes it something super crisp, that's somewhere in a very weird gray zone to me.

 

[decafjava]

Thanks all for the interesting discussion. I don't have anything to add except I agree with the general consensus that if we treat it as a corrective tool but not to replace the creative process. There is a quote out there about wanting AI to do laundry and dishes so we can do art (we do have those machines but I think you get the gist).

 

[Darmok N Jalad]

Regarding tools - I'm also not entirely sure where I stand on sharpening with "AI". If your image is juuuuust out of focus and an unsharp mask or some highpass trickery would have fixed it, ok. That probably didn't need "AI" in the first place.
If it's so blurry it's hard to make out details but the "AI" tool makes it something super crisp, that's somewhere in a very weird gray zone to me.

From what I've seen of AI NR and sharpening, it really can't quite save a missed focus, at least not that I have seen yet. If you don't have the right focus, it's pretty much game over as far as I'm concerned. I guess in some respects, it's still nice to have if you want to improve a photo you took of a memory, so you don't want to delete it and have nothing better to work with.

I do use DXO PureRAW for my files, and in my mind, it's the equivalent of about half a lens upgrade. I've also attempted to replicate the final results of DXO PR with a free editor like Darktable, and I can get it pretty close. It just takes a lot more time. A lot of times, a pass through DXO PR is all I even do with many photos, and I'll just take the time on the select ones I really like.

 

[Ctrlos]

It has long been thought that every photograph taken is a lie. A photo is a representation of a scene, now the actual truth behind it. This was true with the black and white photography of the 19th Century (because the world is colourful) and it is true of today where a press of the button tells a piece of software to capture a series of photons and then process it through complex algorithms into the best version of itself.

Photographs have been edited for decades. How is an AI object removal any different from the clone tool? Pure AI generation of an image are in many ways no different to the complex collages I see every day in say, a movie poster. The difference however is that collage started out as a human idea and was likely made of source images captured by the artist themselves. With human work, effort has gone into it. Learning has taken place. Models and actors have been paid. Costume designers were set to work. Set dressers practiced their craft. Humans place value on artwork equal to the work that has gone into it; a hand-made dresser is worth more than a flat-pack Ikea version.

Pure AI generation removes the effort from the artwork and in turn the value. It turns human ingenuity and creativity into a disposable commodity. A photo from the top of a mountain has value because as a human you recognise how hard it was to climb; an AI generated landscape has none because it hasn't seen it and the image tells no story.

AI edited photographs would be fine, you're just skipping a step. Anyone who thinks putting prompts into Midjourney facilitates artwork with value is kidding themselves.

 

[OldMacs4Me]

Having seen and had a few photos disappear from POTD without rational explanation, I can see why one would be a little jittery about describing any AI process. At least when posting to POTD. Being limited to one photo also makes it impossible to do a before and after comparison in that thread.

As to the weekly contest that is and should be entirely at the discretion of that weeks judge.

All that said my software is just too ancient to do any serious AI work, other than stitching panoramas. Even so there are a lot of manual tools available to me and for the most part I don't hesitate to use them. The exception being product photography. Misrepresenting a product does not fly with me.

One of the reasons I started the The Relaxed Photo Thread was to provide a location where the emphasis was on the images and the stories behind them. Certainly seeing before and after AI images would be most welcome there.

 

[C0ncreteBl0nde]

I appreciate all the interest in this subject and find that I *think* most of us are on the same page.
I'm going to attempt to upload a few photos to illustrate the point I was trying to make when I first brought this up in the POTD thread:

This is a photo of my granddaughter on her birthday with her paternal grandfather's arm and leg behind her. In the second picture, I removed grandpa by loosely making a selection around him and in the "generate photo" part of Photoshop's AI Generative Fill tool, I typed "Remove."

This next is a "picture" that came out of my head. In PS's "Generative Fill" box, I typed the words to what I hoped my scene would look like when generated by AI. There was no "photography" included in creating this virtual artwork (or whatever you wish to call it).

IMHO, and, if I'm correct in what I'm hearing from the rest of you, the first use of the AI generator should be okay to post here in the photographic threads as it would have been doable (albeit, much more difficult) to achieve BEFORE AI. The third photograph of the abandoned gas station, which I made up in my head and had no camera involved in it's creation, should NOT be seen in the POTD thread because it is NOT photography but, perhaps in it's own area devoted to this type of art. YMMV.

 

[Apple fanboy]

I appreciate all the interest in this subject and find that I *think* most of us are on the same page.
I'm going to attempt to upload a few photos to illustrate the point I was trying to make when I first brought this up in the POTD thread:

This is a photo of my granddaughter on her birthday with her paternal grandfather's arm and leg behind her. In the second picture, I removed grandpa by loosely making a selection around him and in the "generate photo" part of Photoshop's AI Generative Fill tool, I typed "Remove."


View attachment 2386798

View attachment 2386799

This next is a "picture" that came out of my head. In PS's "Generative Fill" box, I typed the words to what I hoped my scene would look like when generated by AI. There was no "photography" included in creating this virtual artwork (or whatever you wish to call it).


View attachment 2386802

IMHO, and, if I'm correct in what I'm hearing from the rest of you, the first use of the AI generator should be okay to post here in the photographic threads as it would have been doable (albeit, much more difficult) to achieve BEFORE AI. The third photograph of the abandoned gas station, which I made up in my head and had no camera involved in it's creation, should NOT be seen in the POTD thread because it is NOT photography but, perhaps in it's own area devoted to this type of art. YMMV.

That’s how I would do it yes. The first is a doctored photo (no issue with that)
The second is something other than photography.

Of course the bigger question is when does the ‘other’ image manufacture AI make photography irrelevant!

 

[C0ncreteBl0nde]

That’s how I would do it yes. The first is a doctored photo (no issue with that)
The second is something other than photography.

Of course the bigger question is when does the ‘other’ image manufacture AI make photography irrelevant!

THAT is the question that keeps me up at night!

 

[arkitect]

I appreciate all the interest in this subject and find that I *think* most of us are on the same page.
I'm going to attempt to upload a few photos to illustrate the point I was trying to make when I first brought this up in the POTD thread:

This is a photo of my granddaughter on her birthday with her paternal grandfather's arm and leg behind her. In the second picture, I removed grandpa by loosely making a selection around him and in the "generate photo" part of Photoshop's AI Generative Fill tool, I typed "Remove."


View attachment 2386798

View attachment 2386799

Personally I have zero problem with what you did here… if it saves me some painstaking time with the laso tool and hit-and-miss contextual fill I'd be happy. 🙂

This next is a "picture" that came out of my head. In PS's "Generative Fill" box, I typed the words to what I hoped my scene would look like when generated by AI. There was no "photography" included in creating this virtual artwork (or whatever you wish to call it).



View attachment 2386802

IMHO, and, if I'm correct in what I'm hearing from the rest of you, the first use of the AI generator should be okay to post here in the photographic threads as it would have been doable (albeit, much more difficult) to achieve BEFORE AI. The third photograph of the abandoned gas station, which I made up in my head and had no camera involved in it's creation, should NOT be seen in the POTD thread because it is NOT photography but, perhaps in it's own area devoted to this type of art. YMMV.

Luckily AI generated images like these are still reasonably easy to spot — or at least they still look like badly photoshopped images.

This one has some weird hoses that come in from nowhere… and the lighting, while it does have that "light painting" look, is very strange.
That bright light from above should be illuminating the front of the forward of the two fuel pumps. Same with the car.

I doubt it would pass muster under the eagle eyes of our resident photo-pros if it appeared in POTD. 😁
Right now it still looks like a bad collage.

 

[mollyc]

I appreciate all the interest in this subject and find that I *think* most of us are on the same page.
I'm going to attempt to upload a few photos to illustrate the point I was trying to make when I first brought this up in the POTD thread:

This is a photo of my granddaughter on her birthday with her paternal grandfather's arm and leg behind her. In the second picture, I removed grandpa by loosely making a selection around him and in the "generate photo" part of Photoshop's AI Generative Fill tool, I typed "Remove."


View attachment 2386798

View attachment 2386799

This next is a "picture" that came out of my head. In PS's "Generative Fill" box, I typed the words to what I hoped my scene would look like when generated by AI. There was no "photography" included in creating this virtual artwork (or whatever you wish to call it).


View attachment 2386802

IMHO, and, if I'm correct in what I'm hearing from the rest of you, the first use of the AI generator should be okay to post here in the photographic threads as it would have been doable (albeit, much more difficult) to achieve BEFORE AI. The third photograph of the abandoned gas station, which I made up in my head and had no camera involved in it's creation, should NOT be seen in the POTD thread because it is NOT photography but, perhaps in it's own area devoted to this type of art. YMMV.

I mean this as a sincere discussion point and not as a "wtf are you thinking" which is how it might come across. 🙃 ❤️ And I truly appreciate the spinoff conversation to have as a dedicated chatting area.

But I honestly don't understand why anyone would consider the second image a photograph. Yes, at some point, *someone* took photos of most or all of those elements and gave rights to whatever AI generator program you used. It's not a photo. It's a compilation of elements that may or may not have originated in some source photo. But some elements in the AI programs are probably also computer drawn, which are definitely not photographs then, they are drawings. (Admittedly, I realize that in just a few months or years, the believability of these kinds of images will be vastly better than they are now.)

I mean, yes, it *looks* like a photo. But it's not even well done. The gas pumps don't line up; the lighting doesn't make any sense (where is there a spotlight over the car??). This is totally different from a head swap where a single photographer took 5 images in a row in the same lighting with the same subject and one person blinked at the wrong fraction of a second so the photographer just swaps in a non-blinking head from a different fraction of the second, but possibly even the same second as the blink.

I don't think AI will ever truly replace photography. For one, the AI needs source material. For two, actual photography is too ritualistic for too many people and they won't want to give it up. We're still shooting film and tintypes! AI will become it's own thing, and it may be bigger than traditional photography, but it won't ever take it all over.

 

[katbel]

This next is a "picture" that came out of my head. In PS's "Generative Fill" box, I typed the words to what I hoped my scene would look like when generated by AI. There was no "photography" included in creating this virtual artwork (or whatever you wish to call it).


View attachment 2386802

IMHO, and, if I'm correct in what I'm hearing from the rest of you, the first use of the AI generator should be okay to post here in the photographic threads as it would have been doable (albeit, much more difficult) to achieve BEFORE AI. The third photograph of the abandoned gas station, which I made up in my head and had no camera involved in it's creation, should NOT be seen in the POTD thread because it is NOT photography but, perhaps in it's own area devoted to this type of art. YMMV.

I agree with you: your photos are natural even when you removed the arm and leg behind your granddaughter.
The Gas Station photo has something creepy, same atmosphere seen in a specific movie category.
I already don't like the movie scenes done by computer, too much artefact, not natural, you can tell it's not real.
At that point I prefer animation movies: you know what you are going to get.

Looking at the second rule of POTD: All photos should be your own work
AI Generator is not me or you it's a BOT
Bots should have a different thread: BPOTD 😉

 

[Jumpthesnark]

To me, the second photo and the AI-generated image are both problematic, for different reasons. Neither represent reality.

I know that is a minority opinion here. But reality is imperfect, and I'd rather see images reflect that than an idealized version of something that never existed.

 

[Darmok N Jalad]

...and gave rights to whatever AI generator program you used.

That is the big question. I've heard AI models have been busted for scraping content without permission. I joined MR back in 2017. There was no concept of AI generated art, but for all we know, what we've posted here for years has been fed into the AI hoppers through web scrapers, maybe with or without MR's blessing. Yes, we certainly knew that posting anything online has the potential for theft, but instead of worrying about some random guy claiming our work as his own, we probably have LLMs taking an amalgamation of our work "as inspiration."

Didn't Adobe recently change their TOS to allow them to scan your CC content, and that you had to intentionally opt-out? We're the unwilling prey in order for these models to learn. They can say it's for our benefit, but it won't be long before they become the generative "photographer" that people hire. Send in a few smartphone photos and you now have a family portrait at the Eiffel Tower, or on the ISS, whatever. Not gonna lie, but I really don't trust any of these companies anymore, especially when it comes to AI. The entire tech sector is trying to come out on top in an emerging, poorly-regulated environment.

Not to say AI doesn't have its potential. It looks promising in healthcare and other such sectors, but it's just way too wide open right now to know if it will be a net good for us.

 

[C0ncreteBl0nde]

You might find this interesting; this is a bit about part of what got me thinking about this subject:

I have an account on Flickr, the photo-sharing site, and I recently began to notice while perusing the Explore photos for the day (Explore is a daily group of 500 PHOTOGRAPHS that are supposed to be outstanding from the huge number of photos from around the world that day) that people are getting "Explored" for (what to me) are obvious AI creations and claiming them as photos. People commenting on these photos are praising the creator of awesome PHOTOS which leads me to believe they really think they came from a camera or a phone. This irks me. On these supposed photos, there is no "taken when" date, only "When uploaded" dates, no EXIF (or falsified EXIF) or camera data, no location, no tags or groups mentioning anything about AI (and yes, they have them) and where you enter what kind of content your upload is, they always have "Photo" chosen. So I did an experiment and uploaded two AI "pictures"...one is the gas station photo I posted here and one is of a dog in a forest. What would a dog be doing alone in a forest? That's kind of my point. I did this to see what, if any, attention they would get but I did things differently. I put AI in the title. I put several AI, Artificial-Intelligence, Digital Creation and such in my tags. I joined AI groups and no groups with the word "Photo or Photography" in them. AND, as content type, I set the uploads to "Virtual Photography" which is the only thing they offer that's close to what it actually is.

They've only been up a few days and they are coming down tomorrow, but I did get some "Favorites" and a few positive comments about the photography. I've only been on Flickr since January of this year, so I don't have much of a following, but the uploads have gotten enough attention for what I'm used to that it seems people CAN'T tell they are fake photos and/or they don't look at the details like the tags and content type. You have already pointed out the many ways the gas station creation is obviously fake and many of the "photos" that are showing up in Explore are just as much so...but obviously not to everyone. I'm talking about hundreds of comments praising the uploader for such great photography work. TBH...sour grapes? Yes, somewhat, but it's not like I have tons of photos "Explored". It's unfair to everyone; certainly to world-wide photographers who are deserving of being there and whose photos I would like to see.

I searched the site to see if this subject has been addressed to those who are in charge, and it has, but Flickr doesn't seem to know what to do about it. I realize that the photos chosen by Explore are done with an algorithm and not hand-picked by humans, but it seems to me that at this point in time, the site needs to be able to look for the missing items on REAL photos that I mentioned above. I fear this is just a very small example of a bigger problem. To whomever it was that said that there will always be a place for REAL photography with lighting and focus and all the things we've come to love about the craft, I REALLY hope you are right!

 

[Darmok N Jalad]

You might find this interesting; this is a bit about part of what got me thinking about this subject:

I have an account on Flickr, the photo-sharing site, and I recently began to notice while perusing the Explore photos for the day (Explore is a daily group of 500 PHOTOGRAPHS that are supposed to be outstanding from the huge number of photos from around the world that day) that people are getting "Explored" for (what to me) are obvious AI creations and claiming them as photos. People commenting on these photos are praising the creator of awesome PHOTOS which leads me to believe they really think they came from a camera or a phone. This irks me. On these supposed photos, there is no "taken when" date, only "When uploaded" dates, no EXIF (or falsified EXIF) or camera data, no location, no tags or groups mentioning anything about AI (and yes, they have them) and where you enter what kind of content your upload is, they always have "Photo" chosen. So I did an experiment and uploaded two AI "pictures"...one is the gas station photo I posted here and one is of a dog in a forest. What would a dog be doing alone in a forest? That's kind of my point. I did this to see what, if any, attention they would get but I did things differently. I put AI in the title. I put several AI, Artificial-Intelligence, Digital Creation and such in my tags. I joined AI groups and no groups with the word "Photo or Photography" in them. AND, as content type, I set the uploads to "Virtual Photography" which is the only thing they offer that's close to what it actually is.

They've only been up a few days and they are coming down tomorrow, but I did get some "Favorites" and a few positive comments about the photography. I've only been on Flickr since January of this year, so I don't have much of a following, but the uploads have gotten enough attention for what I'm used to that it seems people CAN'T tell they are fake photos and/or they don't look at the details like the tags and content type. You have already pointed out the many ways the gas station creation is obviously fake and many of the "photos" that are showing up in Explore are just as much so...but obviously not to everyone. I'm talking about hundreds of comments praising the uploader for such great photography work. TBH...sour grapes? Yes, somewhat, but it's not like I have tons of photos "Explored". It's unfair to everyone; certainly to world-wide photographers who are deserving of being there and whose photos I would like to see.

I searched the site to see if this subject has been addressed to those who are in charge, and it has, but Flickr doesn't seem to know what to do about it. I realize that the photos chosen by Explore are done with an algorithm and not hand-picked by humans, but it seems to me that at this point in time, the site needs to be able to look for the missing items on REAL photos that I mentioned above. I fear this is just a very small example of a bigger problem. To whomever it was that said that there will always be a place for REAL photography with lighting and focus and all the things we've come to love about the craft, I REALLY hope you are right!

Those EXIFs are becoming more and more important, unless of course the generative AI starts coming up with that, too. For a while now, we could easily question "followers" online. Not only can it be spoofed by humans for profit, but I'm sure now bots can handle the task of bumping your numbers and likes. It's all so predatory.

 

[splifingate]

I removed grandpa

That is a really amazing removal!

I understand your 'focus', but does Gramps know that he didn't have a hand in it?

 

[splifingate]

Personally I have zero problem with what you did here… if it saves me some painstaking time with the laso tool and hit-and-miss contextual fill I'd be happy. 🙂

This.

 

[mollyc]

Those EXIFs are becoming more and more important, unless of course the generative AI starts coming up with that, too. For a while now, we could easily question "followers" online. Not only can it be spoofed by humans for profit, but I'm sure now bots can handle the task of bumping your numbers and likes. It's all so predatory.

Leica (and probably others but I forget who right now) has implemented content credentials in their most recent camera, which proves that it's an actual photo. It gets keyed in like a password somehow. It's largely to prove something like photojournalism, but with the uptick of AI it's good for that as well. It's a collaboration with Adobe, and while I know people love to hate Adobe, they are working to maintain a photo centric approach to cameras.

 

[Chuckeee]

Those EXIFs are becoming more and more important, unless of course the generative AI starts coming up with that, too. For a while now, we could easily question "followers" online. Not only can it be spoofed by humans for profit, but I'm sure now bots can handle the task of bumping your numbers and likes. It's all so predatory.

EXIF are very easy to edit and fake. That is a very easy function for AI. Depending on EXIF would just lead to a false sense of security

 

[decafjava]

Sigh, maybe Frank Hebert was on to something when he wrote of a far future with no AI after the "Butlerian Jihad" destroying all except the simplest of computers - "Thou shalt not make a machine in the likeness of a human mind" or shall we say, image.

 

[Darmok N Jalad]

EXIF are very easy to edit and fake. That is a very easy function for AI. Depending on EXIF would just lead to a false sense of security

Yeah, that's the problem, isn't it? If generative AI is powerful enough to write code, then it could probably replicate any feature we create to authenticate a digital item. I think ethics are extremely important in this sector, and I just have serious doubts. There seems to be a lot going on behind the scenes at OpenAI, including firing a CEO over ethics. Microsoft is in the mix, and now Apple is, too. Humanity has a history of abusing tools, and I'm afraid we're going to see that with this tool all the same.

 

[katbel]

EXIF are very easy to edit and fake. That is a very easy function for AI. Depending on EXIF would just lead to a false sense of security

This is one of the reasons someone suggested to remove the EXIF every time you post a photo anywhere
This way if someone requires the detailed infos you have them

 

[splifingate]

Sigh, maybe Frank Hebert was on to something when he wrote of a far future with no AI after the "Butlerian Jihad" destroying all except the simplest of computers - "Thou shalt not make a machine in the likeness of a human mind" or shall we say, image.

Aye; a bit excessive, from our perspective, but I guess 'There comes a point...'.

Crazy Eddie, to be sure