My phone (motorola edge 50 pro) has the same option, generally I find it sharpens the image, sometime (low light) it can intensify the colour and a few other aberations when backlit but in my experiance it does not alter the identifying features or generate content (no internet connection required).
I’ve never noticed any major differences between having the setting on or off other than making the images generally a little better and clearer.
In situations where it matters I take photos with it on and off as well as with a seperate camera (rx100 sony) as points of comparison
I think you are getting alarmed by the term “AI” which doesn’t seem to mean what you think it does here. The phrasing “automatic tuning and AI scene detection” seems to be a Motorola-specific thing. This option is explained for example (not your exact phone model) here: https://en-us.support.motorola.com/app/answers/detail/a_id/161570/~/photography-settings---motorola-edge-(2021). Basically, the “AI” only applies to “scene detection” – it apparently uses an AI pattern-matching algorithm to classify the scene, then applies some standard (non-AI) photo enhancements specific to that type of scene. It does not appear to use generative AI to alter the image.
That sound reasonable. I also wasn’t sure that it was using the internet or not so turned it off to save internet costs as well. If it has all the scene parameters in the phone already it doesn’t need to use the internet. But phones aren’t very powerful and can’t really do much AI so they may just be using the term AI for marketing purposes anyway.
Ok for starting soft and trying to be understanding but iNat has become huge and the larger is will become the harder it will be to manage. So, I think that in certain cases there must be rules that are indisputable.
Of course it depends if we want iNat to remain a faithful depiction of reality or not. Consider that there are so many unmarked observations of non-wild organisms that raise so many doubts on the reliablity of the data.
Thank you!
Those are a good start. I would add some some discouragement for erase, retouch, smooth and extrapolating steps without submitting the original.
Those rules are great. They don’t mention noise reduction, and I think that’s good. Noise reduction on Photoshop, Topaz, etc. is that gray area between traditional digital photo editing and AI-enhancement. I expect this will be a sensitive topic in the near future (already is in some circles). For what it’s worth, the acts of noise reduction and leaving noise in both have qualities that distort “reality” of vision, which is already subjective.
FYI you can respond to multiple people in one post by various methods (quoting, mentioning, etc.) which helps keep the threads cleaner than making a sequence of individual posts.
The research that biologists and citizen scientists are able to do from the information in real photographs as well as the information provided with those photographs. The guidelines of iNAT should be driven by that potential research. AI or extremely enhanced photographs do not play a roll in that type of research.
I see that messiness on my screen also. I apologize for not knowing various methods to reply. It seemed to me that when I clicked “reply” with the arrow under the comment…and then the reply I type shows me it is directed to that person…that the reply would show as such. So, here goes again.
What you can do is select the text you want to reply to and click “Quote”. That will put their text your post like you see here. Then you can write a response below. They will be notified you quoted them.
And you can quote multiple posts from multiple people (just quoting @cthawley here to demonstrate, I’m not actually replying to their text)
For sure. FWIW I do use DXO’s Denoise when I feel it’s needed and I haven’t seen anything alarming when I do. I tested out Lightroom’s Generative Fill and that definitely, I think, goes too far when it comes to evidence for iNat observations.
Looks like some companies like Canon are already including AI enlarging in-camera, and OM System is thinking about it. I’m not a fan and I don’t do it, but it seems like companies will be trying to market it at least for a while.
I agree, and photographs have also never been perfect simulacra of reality. Lenses distort and compress, each company has different “color science”, etc etc. Technology is always going to be a moving target, which is why I’d like to have just some sort of DQA vote on accuracy rather than try to make a million rules for every eventuality.
After reading this thread, I think I finally get it, photography on iNat isn’t about documenting reality. It’s about atoning for it. I’ll begin my next observation with a ceremonial bath, smear mud on my lens, and whisper “unaltered entropy” three times into the wind. Then I’ll upload my perfectly honest, blurry mess, untainted by Lightroom, blessedly free of focus. And if the day ever comes when we’re required to skew our histograms penitently to the left and shoot blindfolded at ISO 6400 with a lens smudge to prove our humility, I’ll take another bath, this time in ethically sourced rainwater. Under a waxing gibbous. While reciting, “RAW is law.”
I think we do want to show reality here. But we can manipulate photos somewhat. I crop them (all I have the skills for) and that’s certainly OK. What else? “Photos” the AI make up are certainly NOT OK. But where to draw the line – and how to explain what’s OK and what isn’t? I think we all have an idea what’s OK – a little brightness, yes, actually changes colors or shapes, no – but we deal with people who at best might read the instructions very literally so instructions have to be very clear without making the rules unable to cope with technological changes. Harder than it seems at first.
Exactly. I think there are many uses for AI and forbid all of them would be counter productive. For an extreme example, if a strict no AI policy was enacted here, I would not be able to use my camera anymore, as my OM system body includes AI scene detection to assist focus.
Of course, I am all for forbidding all AI generated AND most AI “enhanced” images. I would also tend to agree with a three strikes policy.
Nature is beautiful in itself, messy and imperfect, but as someone said “there is grandeur in this view of life…” Cheers!
Phones have tiny lense(s) and tiny sensor(s) and would not be able to produce acceptable quality photos without stacking, noise reduction and (interpolating) infill. Some will reduce glare and apply other algorithmic filters.
The only reason we don’t have all that tech in cameras is that the processing required for those increases by the power of 2 to 3 with the pixel count.
High end phones and cameras also transform the photo to correct the geometry (so that parallel lines remain parallel in the picture).
I don’t think there is any reason or justification of rejecting functionality that the majority of observers can’t turn off.
It is only the “Fun”, “Creative” and sometimes “Portrait” features, that put eyelashes on fish, need discouragement and some post-processing actions.
Forcing it? They are doing it in response to popular demand: “I want to take my phone and point it at something (landscape, kitten, lasagna) and upload it to [social platform] and get bazillions of likes/engagement/followers/fame.”
Any output that is not colorful, shiny, smooth, eye-candy, screaming for attention = customer disappointment & loss of market.
I believe that the only solution to this and to all other issues arising as a result of AI generated content is for there to be a universal requirement for all AI content to be clearly labeled as such, and for penalties and fines to be levied against anyone who does not.
There is a growing need for this kind of action, and some platforms are already taking action, such as YouTube demonetizing the overwhelming majority of AI content on that platform. As this awareness grows, I believe government regulations will come about, such as those that were eventually imposed on drones, relegating the use of drones for licenced individuals only, with stiff restrictions on where, when, and how they can be used.
This thread (along with similar threads, particularly this one) has just been discussing how this not really feasible because there is a whole spectrum of AI enhancement ranging from minor sharpness/denoise to entirely recreating the image. How do you draw the line?
All these threads about AI have moved me. I shall feed an AI model all the pictures of myself that I have, and have it generate a photorealistic profile avatar of me.
