AI and "photography"

[billtils]

There's been quite a lot of traffic here on Adobe's AI noise reduction routine, comparing it to other noise reduction programs, including those that use AI.

The current issue of the newsletter of the PAGB takes a broader look at AI and photography - scroll down to the last page and look for the link in bold blue that begins with "Gareth Oakey" (if you don't know what the PAGB is, the newsletter will fix that for you). The article includes a link to a FaceBook group but I don't "do" FB so don't know how far it goes into things.

In brief Gareth is challenging us to consider where AI sits in the context of creative photography - if we are not in control of the taking of the image and its editing then is the photograph our work? For example, other than a few instances it could be argued that we shouldn't need any nose reduction software if we set the camera and lighting up correctly (and, yes, I take quite a lot of shots of a choir performing in a church in winter evenings and, to use a locus-appropriate expression, give thanks to Topaz AI Denoise every time I edit them).

Maybe we should all just sell our gear and buy a high-endcamera phone?

[Round Tuit]

...

Maybe we should all just sell our gear and buy a high-endcamera phone?

That depends on whether you enjoy the process of making pictures or are only interested in the end product.

[DanK]

That depends on whether you enjoy the process of making pictures or are only interested in the end product.

Exactly.

A techie friend of mine who has given up his MILC for a high-end iPhone suggested some years ago that instead of a camera, tourists could carry a GPS-aware device that would present them with an array of good photographs of wherever they happened to be, so they wouldn't have to take one themselves. He decided that he isn't concerned with creativity or with fine control, so he is happy with the iPhone's engineers taking control for him. I am not.

[billtils]

Depends on whether you want to record or to create ... and if "create", whether you want to be involved or a bystander.

[pnodrog]

Noise is mostly a technical issue with the existing sensor technology. At very low light levels "shot noise" will always be a problem. However I don't see any reason for a photographer not to take advantage of AI denoise options in editing. They are essentially just asking the computer to do what would be the ultra time consuming job of suitably spot correcting each speck of noise.

AI at a creative level rather than repetitive level is another matter.

[Manfred M]

That depends on whether you enjoy the process of making pictures or are only interested in the end product.

There are a lot of photographers that are "shooters", i.e., they like getting out with their cameras and capturing what they see. They may or may not tweak their captures in post. There is nothing wrong with that.

On the opposite side of the spectrum, there are photographers out there that still work the same way with their digital workflow as they did in the film days, so their post-processing workflow does not take advantage of the technology that modern technology provides. Again, there is nothing wrong with that.

I even do not have an issue with the people that use the new software to generate images using technology, rather than cameras. The only time I object is when they try to pass these digital creations off as photographs that they took with a camera, whether it is a phone or a state of the art camera.

[billtils]

There are a lot of photographers that are "shooters", i.e., they like getting out with their cameras and capturing what they see. They may or may not tweak their captures in post. There is nothing wrong with that.

On the opposite side of the spectrum, there are photographers out there that still work the same way with their digital workflow as they did in the film days, so their post-processing workflow does not take advantage of the technology that modern technology provides. Again, there is nothing wrong with that.

I even do not have an issue with the people that use the new software to generate images using technology, rather than cameras. The only time I object is when they try to pass these digital creations off as photographs that they took with a camera, whether it is a phone or a state of the art camera.

+1 to what you say here Manfred. I would add - and expect that in your role as a CAPA certified judge this is something you are going to be grappling with more and more - how are we going to be able to tell if an image is a product of the author's own work, perhaps using enhanced versions of tools for things like dodging and burning (which would be OK) or has been created via AI (which would not in competitions or journalism)?

[Manfred M]

+1 to what you say here Manfred. I would add - and expect that in your role as a CAPA certified judge this is something you are going to be grappling with more and more - how are we going to be able to tell if an image is a product of the author's own work, perhaps using enhanced versions of tools for things like dodging and burning (which would be OK) or has been created via AI (which would not in competitions or journalism)?

CAPA is developing a policy on this subject, but is waiting for the main organization that it belongs to, FIAP (Fédération Internationale de l'Art Photographique) to release a policy on it.

As the head of competitions for my photography club, I cannot wait for these bodies to make the decision. I plan to incorporate an interim ban on any AI generated content in photo competitions. This will be an outgrowth of a current club policy that requires members to submit images that they have wholly created. Ultimately, this will mean that 100% of the content in images will have to be photographed by the member. Machine learning assistance (anything from eye detect in cameras through to noise reduction or content generation (content aware fill, for example) in post will continue to be permitted. Pure computer generated components or inclusions of others work. For example, sky replacement will be legal, if the sky is something the photographer one of their own images is permissible, but stock skies are not.

[BobGilbody]

Hello Manfred,
What concerns me with AI used to alter images is not so much what happens in camera clubs or forums like this one but rather use of images to prevent crime, control access to countries etc.
In a camera club someone can make rules and the members will more or less adhere to them or head off to a different camera club.
When images are used as evidence to prosecute in courts the game has changed. In that scenario there are two [ hopefully ] conflicting groups, those held responsible for enforcing laws and the criminal element intent on gain by bypassing laws.
In the first case club members might feel annoyed, in the second the structure of society may well be jeopardised.
I guess in the end the dilemma has and always will be a contest between good and evil.

[Manfred M]

Hello Manfred,
What concerns me with AI used to alter images is not so much what happens in camera clubs or forums like this one but rather use of images to prevent crime, control access to countries etc.
In a camera club someone can make rules and the members will more or less adhere to them or head off to a different camera club.
When images are used as evidence to prosecute in courts the game has changed. In that scenario there are two [ hopefully ] conflicting groups, those held responsible for enforcing laws and the criminal element intent on gain by bypassing laws.
In the first case club members might feel annoyed, in the second the structure of society may well be jeopardised.
I guess in the end the dilemma has and always will be a contest between good and evil.

The ability to create "deep fakes" has existed for some time now, but the ability to "doctor" images and voice took longer and required more expertise. Videos, given the large number of frames that had to be faked took more money and resources, but would still be something that could be done, especially by state sponsored actors. Like anything else, the new technology has democratized this process and we are getting to the point where photo, video and audio evidence may no longer be acceptable as legal evidence. Just about anyone can create strong fake evidence now...

In the "court of public opinion", people will continue to believe what they want to believe, regardless of the quality of the evidence.

[billtils]

In the "court of public opinion", people will continue to believe what they want to believe, regardless of the quality of the evidence.

Sadly, only too true.

[Manfred M]

I just received this information from the Director of Competitions at CAPA.


By virtue of submitting an image into a CAPA Competition, the photographer asserts and agrees that:

For the purposes of our competitions, a photographic image is defined “as an captured image on a light-sensitive device/surface (e.g. photogram technique, film camera, digital camera, smartphone, tablet, etc…) and recorded on paper, film or in a digital format.”

An artificial intelligence (AI) generated image is not deemed to be a photographic image because it was created from scratch by the AI system and contains no image captured by the entrant. Therefore, this type of image will not be accepted into our competitions.

Use of AI features contained within a post processing application are permitted in our competitions providing they comply with a competition’s editing criteria and do not contain any elements which were not captured by the entrant (e.g. element, texture, sky, etc.). Permitted AI features are limited to: clear, de-noise, masking, sharpening, subject selection, tone adjustments, and/or upsizing.

AI sky replacement is permitted provided that the new sky was captured by the entrant. For all potential winning images, the Director of Competitions may require an entrant to provide all supporting images used in creating a composite image.

All components of the submitted image are their original work and does not contain the elements created or captured by someone else; and as such they hold and will retain the copyright for the image. For example: it is okay to take inspiration from other photographers or artists. However, if the inspiration contains the same colour, composition, element(s), perspective, angle, subject, and tone of the original then the inspiration is not deemed to be the original work of the photographer. Another example – A photographer will NOT capture a screen image or photograph an image found on the internet and incorporate it in a composite image that is later submitted into a CAPA competition.

[billtils]

I like it. Thanks for posting.

[billtils]

Sorry - meant to ask if this is based on FIAP feedback?

[Manfred M]

Sorry - meant to ask if this is based on FIAP feedback?

At this point it seems to be a CAPA only position, given the wording in the email that accompanied this text. The National organizations that belong to FIAP continue to look to that organization for overall guidance.

That being said, the position is well thought out and I suspect will require some further refinement and clarification. As an example use of "Content Aware" functions are excluded, but that would also seem to include "Spot Healing", which has long been okay for removing sensor dust.

[billtils]

That being said, the position is well thought out and I suspect will require some further refinement and clarification. As an example use of "Content Aware" functions are excluded, but that would also seem to include "Spot Healing", which has long been okay for removing sensor dust.

So if I remove an element I don't want, such as a power line, by using Affinity Photo's Inpainting brush, that would not be allowed but if I cloned it out or used a healing brush where I define the source, that would be OK?

(I have no complaints about this, indeed it makes sense as it's like Rembrandt or Constable painting over an area they don't like ... not that I'm suggesting anyone here on CIC is in that league ...).

[DanK]

I think the dividing line is hard to pin down, even though it's not hard to come up with examples that seem clear.

One person (only one) in my family is a competent painter. I've often envied the fact that he can alter things as he wishes as he paints. If a tree would look better 10 degrees to the right, he can put it there.

Most of us who make modest adjustments to our photos are now using AI-based tools. For example, I often use Adobe's automatic selection tools because they are faster and often more accurate than my efforts by hand. Does that mean that my photos are less "mine" than if I had drawn the selection boundaries myself? I would say no because the tool was doing what I intended. Likewise, I'll feel free to use Adobe's AI-based noise reduction, now that I've seen that it can accomplish what I want better than I can manually in some cases. However, these are all cases where I am using AI as a tool to create what I intend.

Even without AI, we have a lot of tools that cross the line and impose a different creative vision. Nik filters are an example. They mix an unrevealed set of adjustments to create a "look", and I think people sometimes page through them looking for an appearance they like. Those images are less "theirs" in my opinion, even though there is no AI involved. (I rarely use tools like that for this reason.)

We face a similar evolution in lots of aspects of our lives. When I grew up in the snowbelt of upstate NY, in a day when almost all cars were rear wheel drive and traction control didn't exist, it was a point of pride to drive a manual and take full control over the car. I haven't done that in many years. When I encounter an icy hill in my old Subaru Outback, which has a continuously variable transmission, I push a button to tell the computer that I need that particular assistance, and the car does what it needs to do to get me up the hill. I can't even fully describe what it's doing, but it works.

[Manfred M]

So if I remove an element I don't want, such as a power line, by using Affinity Photo's Inpainting brush, that would not be allowed but if I cloned it out or used a healing brush where I define the source, that would be OK?

(I have no complaints about this, indeed it makes sense as it's like Rembrandt or Constable painting over an area they don't like ... not that I'm suggesting anyone here on CIC is in that league ...).

In my view, these rules are fine for photographic competitions, but should not impact anything beyond that. I suspect that this will be an incremental process and the policy will evolve over time. This reminds me of the "no post-processing allowed" advocates that were there in the early days when Photoshop started to go mainstream. Even the most stringent rules I have seen in in widely recognized competitions on Nature and Documentary at least allow the edits that were made in the traditional "wet" darkroom. The removal of sensor dust is allowed as well.

I rather hope that we see evolution going in two specific directions - a separate category for images that are either partially or completely produced by AI.

The other is the acceptance that these AI tools accomplish the same thing that we were able to do in the past, only much more easily. With time and patience I can get results that are close to what I can achieve with the Clone tool in conjunction with the "Add Noise" functionality, just far more quickly. If it is legal to use the Clone tool then Content Aware Fill should be allowed as well, in my opinion.

[Mingmoos]

The essence of a photo lies not just in its technical precision but in the emotion and story it conveys. I believe it's less about the gear and more about the vision. Sure, AI can assist in perfecting certain aspects, but what is the raw emotion behind the shot? That's human.

As for the gear, I've always said the person behind the camera matters most, not the camera itself. The situation with the AI is the same; it's important what prompts you give him or how you use it, so it's also a human behind it. I've tried these days Mid Journey, and with the right prompts, I generated some great pics. Try it, play with it, and learn because this is our future!

[Manfred M]

The essence of a photo lies not just in its technical precision but in the emotion and story it conveys. I believe it's less about the gear and more about the vision. Sure, AI can assist in perfecting certain aspects, but what is the raw emotion behind the shot? That's human.

As for the gear, I've always said the person behind the camera matters most, not the camera itself. The situation with the AI is the same; it's important what prompts you give him or how you use it, so it's also a human behind it. I've tried these days Mid Journey, and with the right prompts, I generated some great pics. Try it, play with it, and learn because this is our future!

The way that this post has been structured borders on spam, so I have removed the link.

Had you posted some pieces and made a comparison to some of the other packages out there (specifically DALL-E or Stable Diffusion) with a useful commentary on why you prefer Midjourney, that would have been useful and interesting.

You over-stress the importance of input and while it does drive the final artwork, the amount of user control at this point is rather limited and it is difficult for me to suggest that the person doing the input is really the creator. You have also not mentioned that the default (and only) size of output right now is 1024 x 1024 pixels, which is fun to play with, but useless for much else. To become useful in my workflow, I have to run the output through an upscaling tool like Topaz GigaPixel AI.

In my own testing, Midjourney tends to be my tools of choice to work with for new content generation, but have found that I need to buy a subscription for decent access. While $USD 10 per month is not expensive, right now I would suggest that money would be better spent with a Creative Cloud subscription and access to Adobe Firefly software.

At this point, as a photographer, the most interesting product right now is Adobe's Generative Fill tool that is available to all Creative Cloud subscribers, if they download the Photoshop Beta that includes that functionality.