Is there enough transparency on how artificial intelligence and machine learning editing tools actually work?

Some of the AI tools - not Topaz, but I can't remember the exact make now - have models which have been trained on real human retouching and do a very, very good job of replicating it.

If skin retouching skill is what is being judged in the competition I can't imagine those tools would be allowed.
A competition for skin retouching skills would be very rare however portrait competitions are common. It would be difficult to state no filters can be used and how could a judge tell anyway. I was attending a judging yesterday and on one occasion the judge thought a particular landscape image may have been constructed artificially but it was, in fact, a straight image. I think a retouched portrait would be easier to spot. It does suggests that it is best not to overdo any of these effects anyway.

On sky replacement, it is OK if you replace it with a sky which you have shot. I know PS can slot in one from its library. This would not be allowed in any competition I know because all elements of an image must be your own. However, I am not sure I see the need for this now. If the sky is bland because you turned up when the weather, season of time of time was wrong, you can always return. We used to get poor skies because the camera did not have sufficient Dynamic Range but generally most do now.

Dave
 
I started using a camera because my drawing abilities are woeful. Camera equipment was a way of overcoming my limitations, not the other way round.


Same goes for me actually!

But I think you misunderstand what I was saying. We both took up photography because of our limitations with a paintbrush or a pencil. I was talking about the limitations of photographic equipment.
 
But sky replacement doesn't require a lot of vision or skill in post processing. But its a filter down from that.
Actually it does if you don't want fake and crappy looking results that Jerry is still fuming about for a few consecutive years non stop
 
I'm not aware that Ansel Adams did do that other than by using traditional darkroom techniques. Did he actually combine negatives, for example?

I'm sure he would have been a master digital photographer and photoshop user, but I'd like to think he would have the integrity not to create digital composites which misrepresent what was in front of the camera.

No, he used traditional darkroom techniques. The results were derived from what was in front of the camera; they were a representation of what he saw in his mind's eye.

Whether they are a misrepresentation of what was in front of the camera is arguable, as per the sliding scale I mentioned earlier.
 
Surely if you shoot with a digital camera with a bayer sensor then doesn't it extrapolate detail from adjacent sites, and therefore guesses? I am happy to be proven wrong by someone who knows better :).
 
Actually it does if you don't want fake and crappy looking results that Jerry is still fuming about for a few consecutive years non stop
Yes, good composites are still technically and creatively very demanding, regardless of the help from AI.
 
I'm not aware that Ansel Adams did do that other than by using traditional darkroom techniques. Did he actually combine negatives, for example?

I'm sure he would have been a master digital photographer and photoshop user, but I'd like to think he would have the integrity not to create digital composites which misrepresent what was in front of the camera.
One day people will look back and say "he used traditional AI techniques"....
 
There have been a couple of cases that I'm aware of where photographers submitted composites to competitions ( eg Wildlife POTY and Landscape POTY) and were caught bang to rights for doing so. I wonder how much their reputations have suffered because of it?
There is of course an enormous difference between using composites to intentionally befraud, and using them as a creative and expressive photographic technique.

EDIT: to offer some examples

Brooke Shaden

Ellen Jantzen

John Paul Caponigro

Jerry Uelsmann (from the film days)
 
Last edited:
There is of course an enormous difference between using composites to intentionally befraud, and using them as a creative and expressive photographic technique.
I think it is all about this. When I do anything like that and it is by the way less than 5% of my landscape work, it is clearly intended only as decor of any sort and nothing more. Journos caught doing this should be doing prison time, and that includes staging scenes.

There have been a couple of cases that I'm aware of where photographers submitted composites to competitions ( eg Wildlife POTY and Landscape POTY) and were caught bang to rights for doing so. I wonder how much their reputations have suffered because of it?
I would also argue that there probably were and will be really good ones that have and will evade detection (I'm not party by the way, and I haven't even entered anything for a very long time). Getting caught, not only means what you imply, but also means your post production is substandard, junk level garbage with a definite proof.

Or the same flock of birds in multiple landscapes..... ;)

#Edit: Birds that are never seen in those landscapes.
So it was worth it then :banana:. We will never finish this. By the way, Glastonbury one was specially for you and it appears they are there https://middlewickholidaycottages.co.uk/great-somerset-crane-project/
Meanwhile if it ever sells on Photo4me, that's hardly a criminal case :p
 
But don't AI programs like Topaz work the same?
not really. They're trained using machine learning. The training builds an internal statistical model which is completely opaque - it's much harder to predict what the results of any given inputs will be.

A deterministic process is a predictable sequence of mathematical operations.

Small print: my inner mathematician reminds me that there are purely deterministic processes which do not have predictable outcomes but that's another story
 
Surely if you shoot with a digital camera with a bayer sensor then doesn't it extrapolate detail from adjacent sites, and therefore guesses? I am happy to be proven wrong by someone who knows better :).
Surely from adjacent sites it would be interpolation, and if that is what actually happens, it isn't guessing it's a mathematically repeatable and provable technique.
 
it's a mathematically repeatable and provable technique.
That's both true and false. It's true if you know exactly the starting conditions of each run but it's false because few if any real world users will have analysed the data to that level.
 
Photography has been about the manipulation of a view of reality since the days of Julia Margaret Cameron. I won't go so far as to say most but a large proportion of the photographs consumed today have been manipulated (filtered) in some way.
It did occur to me earlier that the proportion now is probably greater than at any time in history, when an Instagram filter is just a phone tap away.
A vast amount of academic brainpower is spent discussing how [an author's] personal perspective affects the story they are telling, particularly in history - which you might expect to be simply a record of fact.
Sure, a difference in perspective is one thing, but I think conscious manipulation is another. Sometimes it's obvious from the style or context of course, which is fair enough (not many people think Fear and Loathing in Las Vegas is straight reportage!). I've enjoyed quite a few exhibitions of obviously altered images and mixed media collages where the manipulation was the whole point. But when changes are entirely deliberate but neither obvious nor acknowledged, we are in much murkier waters.
I don't think anyone who buys a landscape print to put on their wall seriously expects it to be a documentary image. Personally, I don't how replacing a sky is any different to using a graduated filter to create two different exposures in a single frame.

However: that's all detail. My point really is reality in photography lies on a subtle sliding scale. Any attempt to be prescriptive about what is and what isn't real is bound to fail. As soon as we choose a subject, a viewpoint, a focal, length, an exposure, a crop in an enlarger or an image from a sequence then we're presenting a single interpretation of that reality. We just need to be honest about what we've done.

Steve McCurry and Tom Hunter got into trouble by presenting images as documentary when they really weren't - and by attempting to deny that they were manipulated.

Ansel Adams manipulated his images to within an inch of their lives but he was always perfectly open about it.
Personally, I see a profound difference between using a filter and replacing a sky. Most of us have probably seen descriptions of how photographs like Moonrise, Hernandez, New Mexico differ from the original negative, and it's no secret that Ansel Adams made extensive use of techniques like dodging and burning. After all, he wrote entire books about his process. But I'd be deeply disappointed if it turned out that Adams had been pulling the wool over our eyes all these years, and the whole thing was a composite image with the moon and clouds lifted over from Yosemite or somewhere. To me, that would greatly diminish the value and integrity of the image. Steve McCurry is calling himself a 'visual storyteller' rather than a documentary photographer these days, which is I suppose one way of approaching this.

I agree that 'we just need to be honest' but a lot of the time we are merely silent, and many images exist in a dubious no man's land between photographs and digital art. As in the title of this thread, there is often no real 'transparency' about what has been changed.
 
I agree that 'we just need to be honest' but a lot of the time we are merely silent,
I think you're implying that anyone who posts a retouched image should declare the fact. Is that your intention?

I've been considering making clear the level of retouching in images I post to instagram. I think it would be a valuable addition to the aspirational fashion work I do.
But - as discussed above - I feel I should then make clear the other technical choices which have been designed to flatter - lighting, focal length, viewpoint or whatever.

I haven't yet come up with a concise way of doing that without the description taking over, but I'm working on it.
 
Fascinating thread.

Trying to decide what is and isn't allowed in making an image seems pointless as long as there is no intent to deceive.

I just take pictures and then tweak them until I like the result for my own pleasure.
 
I think you're implying that anyone who posts a retouched image should declare the fact. Is that your intention?

I've been considering making clear the level of retouching in images I post to instagram. I think it would be a valuable addition to the aspirational fashion work I do.
But - as discussed above - I feel I should then make clear the other technical choices which have been designed to flatter - lighting, focal length, viewpoint or whatever.

I haven't yet come up with a concise way of doing that without the description taking over, but I'm working on it.
I think in some circumstances that sort of declaration would be very helpful. There are now laws in some countries that require retouching of commercial fashion images to be declared because of body image issues, and there's currently a Private Member's Bill that, if it progresses, would bring in something similar in the UK:
 
Does vaseline on the lens count as retouching? Or a diffusion filter? What about makeup on the subject, surely that is "retouching"?
 
I haven't yet come up with a concise way of doing that without the description taking over, but I'm working on it.
I'm of the opinion that any picture is a lie or at best a partial truth.

Whether created with pencil, paint, film or digital a picture is artificial and only shows what the person creating it chooses to show, That being the case, there's no real need to describe everything that led to the image as displayed, other than in legal proceedings.
 
A few years ago I saw an idyllic looking image on the front of a brochure for holiday cottages in the Brecon Beacons. I thought the location/viewpoint would be a great one for one of my postcards. Using all the clues I could find on the image I studied maps and online to try and find it, but eventually had to give up. I then asked someone who knew the area well and she told me where it was. I arrived to find it almost unrecognisable from the photograph. It looked like it had been taken from a helicopter, and various extraneous landscape features had been removed , like electricity supply poles. If I had booked that property from the photograph I'd have been very disappointed.

But going back to the original post, whatever Ai processing tools actually do to an image must be very minor compared to what some photographers do!
 
Last edited:
It did occur to me earlier that the proportion now is probably greater than at any time in history, when an Instagram filter is just a phone tap away.

Sure, a difference in perspective is one thing, but I think conscious manipulation is another. Sometimes it's obvious from the style or context of course, which is fair enough (not many people think Fear and Loathing in Las Vegas is straight reportage!). I've enjoyed quite a few exhibitions of obviously altered images and mixed media collages where the manipulation was the whole point. But when changes are entirely deliberate but neither obvious nor acknowledged, we are in much murkier waters.

Personally, I see a profound difference between using a filter and replacing a sky. Most of us have probably seen descriptions of how photographs like Moonrise, Hernandez, New Mexico differ from the original negative, and it's no secret that Ansel Adams made extensive use of techniques like dodging and burning. After all, he wrote entire books about his process. But I'd be deeply disappointed if it turned out that Adams had been pulling the wool over our eyes all these years, and the whole thing was a composite image with the moon and clouds lifted over from Yosemite or somewhere. To me, that would greatly diminish the value and integrity of the image. Steve McCurry is calling himself a 'visual storyteller' rather than a documentary photographer these days, which is I suppose one way of approaching this.

I agree that 'we just need to be honest' but a lot of the time we are merely silent, and many images exist in a dubious no man's land between photographs and digital art. As in the title of this thread, there is often no real 'transparency' about what has been changed.
I entirely agree with your post, Retune, with one exception: that digital manipulation of the sort we have been talking about could be described as "art" in any shape or form .
 
I entirely agree with your post, Retune, with one exception: that digital manipulation of the sort we have been talking about could be described as "art" in any shape or form .
Attempts to define art have, so far as I can tell, never ended well.
 
I agree, but perhaps it is easier to define what isn't "art".....?
Generally this would be anything mass produced, anything without intended aesthetic value. Btw Tate modern I'm looking at you
 
I agree, but perhaps it is easier to define what isn't "art".....?
Yes, I reckon the only credible definition of art is that it's something that can't be defined. If it's something that can be defined, it probably isn't art.

It's the lack of boundaries and lack of definition that makes art what it is. Whatever that is !
 
There are very widely used an accepted definitions of art, the problem isn't that art can't be defined, it is that some people don't like the definitions.
 
I think in some circumstances that sort of declaration would be very helpful. There are now laws in some countries that require retouching of commercial fashion images to be declared because of body image issues, and there's currently a Private Member's Bill that, if it progresses, would bring in something similar in the UK:
I know: unfortunately blanket statements like that are idiotic. Almost every single image will have a label; it'll become as ubiquitous as cookies warnings. And they'll be used in the same way - a tiny number of people will burrow into the detail, the rest will just become blind to the label. And that's one reason I haven't done it yet: I need to come up with a key or rating system to explain what's been done as per my post here.

And as I've said: retouching is only one small part of the story. Selection of the other aspects - including the model - will have considerably more influence. This is perhaps most obvious in fashion but it applies to all image making.

Does vaseline on the lens count as retouching? Or a diffusion filter? What about makeup on the subject, surely that is "retouching"?

My point exactly, thank you.

shrugs is my opinion

You win a biscuit for most valuable contribution to this thread.


The original question was:
Is there enough transparency on how artificial intelligence and machine learning editing tools actually work?

My contention was that - at present - AI tools aren't significantly different to any of the other processes involved in creating an image. I don't need to understand the detail of any of those processes. I haven't yet read anything to change my mind.

And I still don't see how you could - for example - prescribe against an AI sharpening tool but allow an octave-sharpening based scheme. They are similarly obtuse in their working, but one is stochastic and the other deterministic. That reasoning extends to other tools, too, including sky replacement. Is it ok to bracket and merge shots taken seconds apart - but not replace with the sky with a shot taken an hour later? Or a day? Or a year? Or from a slightly different viewpoint?

Any attempt to be prescriptive about what is and what isn't real is bound to fail.

I still think this is true. You might ban individual tools or techniques on a case-by-case basis but I can't see a way of codifying that in a general statement.
 
I know: unfortunately blanket statements like that are idiotic. Almost every single image will have a label; it'll become as ubiquitous as cookies warnings. And they'll be used in the same way - a tiny number of people will burrow into the detail, the rest will just become blind to the label.
Depending how it's done, we might well become blind to it, or never register its meaning. Dr Evans mentions the example of 'the ‘P’ symbol for product placement ... seen on commercial television in the UK' to which I reacted 'the what?'. Maybe I'm not alone. This is from 2011, when it was apparently introduced with a TV campaign to publicise it:
But there is a serious point to address. Being constantly exposed to heavily manipulated and sometimes anatomically impossible images of models and influencers can't be doing the mental health of a generation of young people with a worrying prevalence of body image issues and eating disorders any good. How should this be dealt with?
 
Back
Top