I don't think it's intentional. This behavior is a straightforward consequence of training something on pairs of "real" and "degraded" images. It learns what images look like and does the best it can. Philosophically, replacing the moon with a better moon is no different than any other computational photography technique that uses world knowledge, like reducing red eye from flash by detecting faces, or enhancing blurry text using knowledge of the alphabet.

The problem is philosophical - people are upset by certain types of enhancement error, but not others, and they can't even articulate the difference to themselves, let alone to an AI.

> people are upset by certain types of enhancement error, but not others, and they can't even articulate the difference to themselves

People are upset by enhancement errors that produce an image that is meaningfully different from the "true image". Replacing the sun with the moon or replacing the letters with different letters (e.g. https://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres... ).

The philosophical problem here is that you can't get at the true image, you can only imagine ideas like "what the human originally taking the photo saw with their naked eye".

What constitutes a "meaningful" difference? If you replace flash-induced red eyes with a more plausible (yet inaccurate) eye color, people are perfectly happy with that, even though the camera indeed saw red photons. They're not happy if you run the exact same algorithm on a photo of someone with genuinely red eyes. It's not always about the "true photons" - the cause of the red eye is relevant, and so is the intent of taking the photo.

"what the human originally taking the photo saw with their naked eye" is a tempting standard, but a misleading one - there might not be a human eye near the camera at all, and at any rate there certainly isn't one co-located with the focal point. And shall we have cameras that render only a few degrees of high resolution color, and have a large-ish blind spot? It's what the human eye sees.

> What constitutes a "meaningful" difference?

Well, replacing the sun with the moon constitutes a meaningful difference. Let’s start there and work our way down.

> If you replace flash-induced red eyes with a more plausible (yet inaccurate) eye color, people are perfectly happy with that, even though the camera indeed saw red photons.

I believe this is already covered by:

  what the human originally taking the photo saw with their naked eye

Camera saw red, you didn't.

When there's no human nearby, people may not know what the truth ever was, but we'd like to pretend that someone might have. Well, usually. I have yet to see anyone complain about astronomical photos using false colours, for example, but this may be because astronomy nerds understand the "why".

Blind spots are indeed our brains filling in the blanks, however we mostly avoid this (and foveation limitations) with saccades; on the other hand, part of stage magic is to manipulate us into looking in the wrong place, and if you're using the same tools without the audience knowing they're getting fooled it's seen as at least a bit antisocial.

>Camera saw red, you didn't.

But, again, you cannot consider flawed human perception to be the ground truth! Photonically, the red eye is the ground truth. If you were looking through the viewfinder, you should also have seen the red eye effect, if only for an instant. If you didn't, it's only because your eyes were too slow. Or perhaps you weren't looking through the viewfinder at all and therefore not in a position to see the retroreflection, in which case you might as well have been facing the other way for all that counts.

Or consider taking a photo from a moving vehicle. You can't resolve your blind spots with saccades because the scene moves. Your subjective impression is probably that only the thing you're tracking is crisp, and everything else is a vague, blurry mess. Is that what the camera should capture, irrespective of shutter speed?

I think it's clear that culturally we expect photographs to be an indexical trace of genuine photons, not a replication of subjective experience. That is the value they provide over an "artist's impression".

> But, again, you cannot consider flawed human perception to be the ground truth!

Irrelevant. The entire point I'm making here is that I think the criticism isn't about objective ground truth, but subjective: users want the camera to show what they saw as they saw (/would have seen) it.

> the red eye is the ground truth

Only if you happened to be looking at the moment of the flash itself, while the subject's retina is most dilated, and somehow the rapid change of intensity didn't mess with your own vision.

> I think it's clear that culturally we expect photographs to be an indexical trace of genuine photons, not a replication of subjective experience. That is the value they provide over an "artist's impression".

Some, perhaps many, people might. I think I was disabused of such notions even by the time of my GCSE in art*, which involved a film camera because digital cameras were very rare in 1999/2000, and thinking about it also my GCSE in history given that showed us all the editing done by the Soviets. I wouldn't have phrased it as an "indexical trace of genuine photons", but photos have been absolute liars from day one.

* where, ironically, one of the things the teachers tried to impress upon students who only slightly listened, was to actually look at the thing they were drawing instead of imagining it