I recently had a lot of people debating whether the picture I uploaded was AI-generated or Photoshop-created. The image link is here: www.inaturalist.org/observations/336776060
This image was so unusual and deemed unlikely that its validity was questioned. Yet there is no process to clear a picture on this forum. IMO, it is only a matter of time before “Bad Actors” will upload images that are designed to show something that is not real in an effort to gain notoriety or to publish false science. I feel it will become critical to create a group of people where some image of scientific significance is uploaded and deemed so important that those involved can ask a forum group to scrutinize the image and its history. This will allow those on the forum to trust what they see.
Tied in to this issue, I recently had several people questioning a picture that i uploaded and had used “Photostacking” on my Olympus TG-7 and did not realize that the “Ghost” images created by the movement of some of the eggs in left side of the image created a false idea in some of the viewers that I did not intend to create and were not reality. They looked at the image and thought Eggs were coming out of the sponge on the left, and we had quite a debate until I understood what they were looking at, and the false idea my image had created due to image stacking in an effort ot increase depth of field and a clearer image. It was all cleared up quickly once I mentioned that the picture wa smade usng image stacking.
A few weeks ago I was browsing flags on iNat and happened to see an AI flag and took a look at it. The image was an incredibly realistic image of a hoopoe - the main tells were that it was kind of too perfect and inconsistent with the user’s other uploads, maybe some discrepancies in the feathers but I don’t think it would stand out to me if it were mixed in with high quality photos.
Sadly, this already happens, but so far, it appears to be a limited problem and the images can often be detected and the users banned. You may be interested in these other threads:
Personally, I think people should never delete observations that are correctly represented just because someone asks if the date/location/subject is incorrect, especially in a case like this where you are positive that everything is correct. I often ask people to confirm that some detail is correct, because for every deliberate attempt to mislead on this site there are dozens of accidental mistakes, and having the original poster confirm that the details are correct is immensely helpful.
At current, no one can be 100% sure of.. anything, but especially whether something is real or AI generated. With AI “Upscaling” being included in many phone photos, sometimes even the person that posted the image doesn’t realize that it’s been altered by AI. iNaturalist is fundamentally a site to help people connect with nature, not a rigorously reviewed scientific database. The people identifying observations and dealing with curation of the database will always be volunteers trying to plug a whole in a dike with their finger (metaphorically speaking), so I don’t think that there will ever be a group of people specifically dedicated to reviewing where images are AI generated or not.
Firstly, it would never occur to me that your photo was AI created for one primary reason… it’s not good enough. Oh dear, that does sound terribly harsh and PLEASE please don’t be offended, let me explain how I mean it… to create seemingly perfect images with AI is relatively easy these days with just a few basic skills, but AI still finds it quite difficult to create convincing flaws and imperfections, such as in your case, with a muddled background and parts of both the spiders intertwined and blurred. To me it couldn’t be anything but an image captured in the field under challenging conditions.
I suppose it could have been created with double exposures and Photoshop, but fiddling around with layers, merging, blending, retouching and who knows what else would require an inordinate amount of time and skill, why on earth would someone bother!? Yes, it does illustrate interesting and unusual behaviour, but I can’t see that Sheldon has anything particular to gain from its publication, other than the pleasure of sharing a fascinating find with like minded people. I bet he’s half wishing he hadn’t!
The really sad thing about this is that we are here talking about it at all. There is such an air of suspicion around photography at the moment. I take great pride and care in getting the clearest images I can of the plants and animals I come across, but I’ve become reluctant to post them on social media, as I know I’ll end up having to defend them from the accusation of AI creation.
On the other side, when I encounter particularly good images of especially interesting organisms, or behaviour I sense is unlikely (imagine those cringe-inducing images of frogs holding water lily leaves to shelter them from the rain), I will nearly always run them through an AI detector before liking or commenting on them.
Whether we like it or not, we are in what has been dubbed the post-truth era. It’s difficult to see where it’ll lead, but in the meantime, it’s absolutely right to be suspicious, but let’s try and avoid paranoia.
On a footnote, in the photography community, it is no coincidence that various schools of art photography involving the creation of deliberately imperfect images are becoming increasingly popular.
We have a new DQA against AI images - Evidence accurately depicts organism.
It would be good to pre-emptively UPvote that, against future identifiers who may miss this thread. I would also link to this forum discussion from a comment on your obs.