Ethical flagging of AI images
Just following on from Julian's question on AI.
Am I right in thinking that as EXIF info regarding shooting details follows an image? If true, then couldn't/shouldn't this be used be used as an identifier to flag whether and image originated from camera or PC?
It seems to me that it would be very easy for software/social/media platforms to indicate whether an image (or video?) was generated by a PC (AI) or whether a camera was used.
This would help in case where AI was abused in the editing process of course, but at least it would flush out those who try and pass off AI-prompted generations as photography.
We have the technology folks!
Thoughts?
9 Comments
"Am I right in thinking that as EXIF info regarding shooting details follows an image?"
Yes and No. Some websites strip EXIF form the photo completely, for example: the Leaf photo was download from Fstoppers and Leaf 2 from a popular Social Media platform.
While AI "could" be embedded in EXIF it can easily be stripped as well. There would need to be as area in Properties that is non-editable.
Good point Dean. If mandated though we could mandate the addition of EXIF as a validation check.
A basic question: What or who may prevent me to add reasonable looking Exif data to an AI image?
True, but you would have to have knowledge to do so and this would be a clear attempt to defraud the viewer.
In today’s technology locking down access to EXIF details has to be a relatively simple task.
The fly in that ointment would be someone like me who works a lot in film. The majority of what i do is film! There is no EXIF data. At this point in time though, and I am sure this will change, i can often ferret out an AI image. They just have a different look.
I agree on the look of AI images, but they will only getter better/harder to spot and even now most non-photographers are fooled.
The film concern is valid, but if EXIF was mandated at least the vast majority of photos could be validated.
I can see a point in future where EXIF-like data has to accompany AI generated images as a way to help tell fact from fiction.
HI Alan! Very topical issue. I think Dg9ncc/portable's point is vital here. If someone is determined to pass off AI as real, fake EXIF would be relatively easy - or so I think, though I hope I am wrong.
I'm afraid this AI juggernaut is unstoppable, but like much new technology (like photography, "threatening" painting!), is initially perceived as a threat to established ways of doing things, but this settles over time into being incorporated into everyday life.
A bit like science fiction which rarely predicts future reality, it's really hard to predict where it will lead. I can see some crafts, e.g fashion photography, being hard hit, in that case perhaps only the catalogue-level sector. But AI cannot be creative, only a digital pastiche of what's already there, so I think creativity is safe.
Personally, I find the very idea of using AI in creative pursuits boring, about as interesting as having a robot clean your house. Yet I wouldn't be surprised if someone proves me wrong.
A robot that would clean my house? Where can I apply? You are correct in your analysis though. But that would automatically increase the stature and standing of "real photography" leaving AI as the "new guy". All of this was said in jest, BTW.
I think that EXiF information is embedded within the photo details. I don’t doubt that someone MAY be able to hack this, but the average guy who creates images from prompts is highly unlikely to have a skill set to do so.
Like anything else, the industry could lock this down should they choose to do so.