AI - we need to talk!
Hello everyone
I’ve noticed that some of us are experimenting with AI - sometimes photos look heavily AI processed, sometimes even entirely AI generated. These images can look stunning because the AI “knows” exactly what kinds of pictures people find appealing. In some “photo of the day” submissions, the AI influence was obvious to me. I’m fully aware that modern tools such as Photoshop and Lightroom embed AI features, and I use them myself. My favorite is the AI assisted eraser, and I’ve also tried the reflection removal tool (though I’m not satisfied with its results and generally prefer to integrate reflections or avoid them while taking the photo). I try to shoot deliberately and thoughtfully so that I need minimal post processing. I do change colors and contrast of my photos, to adapt them to my perception, or to make something new out of them. Processing, enhancing, and altering photos is as old as photography itself; some argue that the moment you choose a frame you already influence the viewer’s perception of reality.
As photographers, we should be especially conscious of AI’s growing impact on our craft and on how people perceive “beauty,” “normality,” and “reality.” We should reflect on how much post processing we’re comfortable with. We need to learn how to use AI responsibly (what does that look like?) and develop the ability to spot AI generated images. It helps to prompt some pics on one of the well-known AI platforms and experiment with it aware of what AI can and can’t do and how it does what it does. A quotation comes to my mind: «Shape or be shaped!». I don’t remember who said it.
Finally, we should discuss what it means to be a photo community today. Is it defined by the use of cameras, or simply by the creation and appreciation of pictures? The least I would like here on Tookapic is that photos created by or edited with tools like ChatGPT, Midjourney, DALL-E or Firefly would be clearly marked as ‘made by AI’ or ‘processed by AI’
Please share your thoughts about this...
leszek55 Jazzie, I almost often use Lightroom (mainly to improve details in shadows and highlights) and often Topaz Sharpen AI if the photo is slightly blurry and I want to sharpen it. I'm not familiar with the programs you mentioned (ChatGPT, Midjourney, DALL-E, or Firefly).
Lauren Huston @leszek55 The AI programs is where someone types in something like "a cat sits on a fence by itself, next to a giant mushroom. it's night time, with the moonlight sparkling off the nearby lake" and those applications generate the image for you. Or you might supply a photo of your backyard fence and type in the above and it'll alter the image and generate the missing content.
Using Lightroom to simply adjust shadows/highlights etc is generally not a standard use of those sorts of systems, and has been pretty much done since photography has been a thing.
Lauren Huston I look at photo adjustment in 3 very generalised categories:
Editing - i.e. cropping, colour/brightness/contrast etc adjustment.
Manipulation - introducing elements not in the original photo (adding reflections for example), or layering to achieve a certain effect, or removing elements that were originally there, in order to achieve the original intent of the photo.
Generation - completely modifying elements that shouldn't have been possible in person. e.g. was that unicorn always there in the sky?
In my opinion, I feel that the first two are important in photography, and have been around for a long time. Even when people take a photo with a phone (for example) and claim no editing was involved, there absolutely is editing and/or manipulation involved, it just wasn't done by the photographer.
Recently I had a project where I had to take photos of people on stage, and using AI to remove distracting elements was a life saver as I am NOT good at photo manipulation.
I don't particularly like AI art, both on an ethical level and often it just doesn't look very good, but I understand why people do it.
Personally I would prefer if images that were created by typing in a prompt were not part of this site.
But one consideration is there really a difference between say, someone being very good at "airbrushing" (that's a term I haven't heard for a long time!) and AI smoothing everything out? Other than the skill of the photographer/editor.
jazzie
Thank you for pointing out these three categories of photo adjustments, they make sense to me. Perhaps we should make a internal differentiation in the category 'generation':
- 'human made': a precise idea of content AND design I developed myself, maybe using pencil and paper. I then realise it the way I imagined it, for example by making a composite or 'airbrushing'
- 'machine made': an idea I had, then prompt it and look what THE MACHINE makes of it, which will never be exactly what you imagined. Or maybe I'm just bad at prompting 😉
Lauren Huston Yes absolutely. To me, it's the involvement of the machine doing the actual artwork that I have issues with.
Basia Fotografia od zawsze była w jakimś stopniu zmieniana w postprodukcji, czy podczas wywoływania obrazu w ciemni, czy przy użyciu programów do obróbki zdjęć cyfrowych. To zawsze jest fotografia. Mamy teraz nowe narzędzia, które dalej nam sztuczna inteligencja. Niektórzy z nas pewnie chętnie ich się nauczą i będą korzystać. Póki to nadal będzie tylko postprodukcja to nadal jest zdjęcie. Obraz całkowicie stworzony przez sztuczną inteligencję już nie, nie jest niczym co zobaczyło najpierw nasze oko. Jeśli chodzi o Tookapic nie chciałabym aby tutaj pojawiały się takie obrazy, bo mija się to z celem tej platformy. Jeśli chodzi o samą postprodukcję, to nie wiem czy w większości zdjęć nie trzeba by oznaczać, że są przetworzone przez sztuczną inteligencję, bo aparaty cyfrowe, czy nasze telefony to teraz małe komputery. Telefony w szczególności, bo programy do robienia zdjęć mają wbudowaną sztuczną inteligencję, by nasze zdjęcia od razu były "piękne"
jazzie
I like your argument about 'when your eyes first saw it, it is a photography'. But then, what if your inner eye had seen it, or you had seen something in your dreams? Wouldn't that be a photo if you made this visible to others by using AI? Or does the physical eye need to be involved?
What I'm talking about is photos that were changed so much by the use of AI that they look like completely made by generative AI and no more like photos. I think, what a camera does 'in-camera' is what fits into Lauren's category 'editing'.
Basia To jest odwieczne pytanie kiedy kończy się fotografia a zaczyna grafika. Nie da się jasno postawić granicy, że tyle i tyle ingerencji w zdjęcie powoduje, iż nie jest to już zdjęcie. Jest wiele kadrów, które zobaczyły moje oczy, a nie zostały odwzorowane na matrycy aparatu i nie mówię tu o oczach wyobraźni i wewnętrznym oku
radzikmr Personally, I use AI in both PS and Lightroom. So, consciously. However, as Basia rightly pointed out, currently advanced phones use AI without the user's knowledge. It will be difficult to identify photos that did or did not use AI in post-production, even for less experienced users. However, photos generated by these types of tools are certainly pointless. But why even be here to post something like that? It defeats the purpose, especially on our humble Tookapic. I could understand such an action if someone has thousands of Instagram followers, but here? Complete nonsense. Therefore, generating artificial photos is self-deception. There's another aspect – processing and post-production using tools like ChatGP. You can't generate photos immediately; you can request sharpness and color enhancements, which is standard post-production. And finally, I believe that you can still change a photo much more in PS and LR using traditional tools than AI can. It's a difficult topic.
jazzie
Thank you for joining the discussion. At the moment, I think you’re the user who is most clearly experimenting with AI, and I believe it would be interesting for others to learn more about what exactly you do. To me, your picture from 11 October tookapic.com/photos/986775 clearly looks like a generative AI image (and became the photo of the day!) - I didn’t even have to look closely to notice that. It has that smooth, hyperreal, aesthetically pleasing look that lacks the small “imperfect” details of reality, such as uneven ground with puddles and leaves, or a slight motion blur.
It also shows the typical mistakes AI still tends to make when you look at the details: For example, the green spot on the tree in the background, the incorrect shade of black on the branches of the fifth tree on the left, and the odd details on the benches.
What I find particularly interesting is that you show the “source” image, the one from 10 October "Untitled". Comparing the two reveals a lot about how generative AI “works.” I find it fascinating to compare these because it helps us understand how AI interprets and reconstructs visual reality.
For me, taking a photo and asking ChatGPT (or another AI) to rebuild an improved version already qualifies as generative AI.
radzikmr Here you go, let me explain. The September 10th photo only has a subtle LR filter. The September 11th photo has a number of AI-assisted treatments in PS: - Adding fog or haze: Using a "Gaussian Blur" layer on the top of the photo. Reducing the contrast in the background, making the trees in the background less distinct, using a gradient mask and selective blur. Next, applying adjustment layers like "Color Balance," "Selective Color," and "Gradient Map" to highlight the warm autumn colors (orange, red, gold). Desaturation of certain areas: The background is decolorized to focus on the path and leaves. Using a "Soft Light" layer with a subtle gradient to add a diffused light effect and a classic vignette. The changes on the tree are likely the effect of the "High Pass Filter" or "Unsharp Mask," I didn't notice. Finally, using filters "Oil Paint" to give the photo a more illustrative character, perhaps that's where the generative AI feel comes from? As for the green, I didn't even notice :), and that's a typical mistake when working with PS and masks; I messed it up and didn't clean it up :). AI does’t do such kind of mistakes :))).Thanks for pointing it out. In short, this isn't an AI-generated image; what's more, it's literally a different photo. This is my workflow; I can take several of these photos every day. Try to Upload a photo from 10 to Chat GPT or Copilot and give it a prompt to get it from 11; I'm curious about the result. But to the point. What's the difference between generative processing and advanced AI? Andreas Gursky completely changed the content of the photo in the "Rhein II" photo, removing factories, chimneys, and other elements. What difference does the tool the artist uses make? Because that's what conversation comes down to? And next time, I wish you more courage. If you wanted to call me out, you could have done it directly. After all, we talked privately and I have no secrets.
radzikmr And one more thing. If you want to see a photo edited with the prompt in the gpt chat, take a look at September 26th. Of course, I have a raw or jpg version of this photo, because I don't remember what settings I had on my phone. The prompt was something like "warm up the colors, boost the contrast, sharpen and highlight the steam rising from the cup, slightly blur the background more, without interfering with the image's content." I could have done all of this in PS, even better. But that day, I only had my phone and little time. So, the question is: ethical or not?
jazzie
Ok, then you got me - if that’s all photoshopped. After applying the Gaussian blur, you re-sharpened the image, and that’s when the shapes of the subjects changed?
The difference between the tools really comes down to what kind of mastery you’re aiming for: mastering a photo-editing tool, or mastering the art of prompting. Probably a personal choice. I still feel a photo is more mine if I’ve made all the changes myself, consciously and deliberately.
And about courage: I could have called you out, but I wanted it to be an open exchange of views, which might not have happened otherwise.
radzikmr No, as you pointed out, this is a clone of the fifth branch of the tree, intended to cover the urban fabric. Listen, if it's easier for you to assume this is an AI-generated photo even though you essentially have the original, then fine, because I'm starting to get the impression that the discussion isn't about the ethics of AI use, but about my particular photograph, which unfortunately became the photo of the day. To make yourself feel better, I've removed the photograph; it won't arouse controversy. As Oscar Wilde said, "Without freedom, there is no art; art lives from constraints but dies from coercion."
jazzie You got me completely wrong… That wasn’t the intention. I’m sorry!
radzikmr No problem.
sebi Hej, kochani nie jestem zwolennikiem AI w fotografii i nie bardzo rozumiem w jaki sposób można go wykorzystać w „naszym” ukochanym miejscu ?? Nie jest to czasem okłamywanie samego siebie ? Idea „złapania” aparatu każdego dnia nie bardzo a wręcz bardzo mi nie wygląda po drodze z AI ??? Może się mylę …. Ale dla atencji AI jest szybką i skuteczną drogą ;) Pozdrowienia !!!
Top Contributors
People who started the most discussions on Talks.
-
138 -
26
-
25
-
21
-
21
-
21
-
20
-
17
-
16
-
16
Unanswered Talks
Discussions with no comments. Be first to post a comment.
-
Paweł Kadysz posted
-
ponzu posted
-
Paweł Kadysz posted
-
Magda Korzewska posted
-
Nicolas Camenisch posted
grainydays Thanks Jazzie for posting so thoughtfully on the tricky subject of « Generative AI »
One of the reasons I shoot film is to go the opposite direction of where AI is going, in other words, to be more authentic and less artificial. Always nice to have a negative to prove a photos is real and not just the result of some AI prompt 😊
What I especially like and admire about the Tookapic community is the human side of most of the photos published. So, for me at least, AI generated photos don’t really make sense and have little worth.