I have been uploading AI generated historical portraits of my ancient ancestors
However, another user keeps deleting them completely from the memories section. How can this issue be addressed? (I have reached out to the user concerned inviting his input on the portraits and offering to make adjustments to them.)
Answers
-
Original Post response to @Ian8662 included in the post
MandyShaw1✭✭✭✭✭November 5edited November 5IPv6Ian, to help with that, can you answer the question I asked you here0 -
@sc woz When you split this topic from the original thread, you left behind a couple of comments that were specifically responding to the question you have moved to this new thread. Could you please move those comments to be here in this new thread?
Here are the links to those comments:
1 -
Original Post response to @Ian8662 included in the post
It's impossible for another user to delete pictures that you uploaded. What may have led you to think this happened is one of three possibilities:A FamilySearch administrator deleted your pictures.Another user detached the Memories from the person profile you had attached them to. In this case, there will be an entry in the change log for that profile. But the photo you uploaded will still be in your Memories Gallery.If you're talking about a portrait, then any other user might have removed or replaced the portrait. Unfortunately, these actions create no entry in the change log for the affected person profile. But the associated Memory you uploaded will still be safely on your Memories Gallery0 -
I am confused at where these images are coming from but you may not have the lisence to use those photos.
0 -
@jamiehadlock AI images typically cannot currently be copyrighted as they lack human authorship. However, users of generative AI could be subject to claims of IP infringement, as these platforms are widely trained on copyrighted works without the permission of the copyright holders of those original works.
1 -
The key thing here is whether FS changes its memory upload policy to reflect the relevant suggestions passed to them by @AmberML1 (see the original post linked at the top of this one, which was split from it).
Amber, @RaniM's point above re AI image copyright etc. would be an excellent addition to your discussions with those responsible for the policy.
3 -
Just had my first encounter with this subject.
A user just added a fake AI image to L63D-JV1
https://www.familysearch.org/en/legal/familysearch-upload-guidelines-and-policies reference 2nd to last paragraph: "Photos, Documents and Audio Recordings may not be edited in such a way as to make them inaccurate, false, or misleading."
I detached the image on the grounds it is inaccurate, false and misleading.
1 -
Following Marvin's point, what if a user clearly states the uploaded image is AI and submitted purely for reference purposes, i.e. illustrating what uniform a person profiled who held a certain historical office would have worn (with e.g. citations to Wikipedia)? Is the latter allowed if so specified? I've also encountered a phenomenon involving AI where users, doubtless with the best of intentions, have uploaded highly stereotypical racialized images of non-white people generated with seemingly older or less sophisticated versions of AI in lieu of actual photos or otherwise representations of them, especially in the case of certain Latin American people who lived a century or more ago. I don't want to open a can of worms, but perhaps people can be reminded to use common sense when uploading imaginary AI images of people, especially in the context of speculated racialization, and especially when the images come across as excessively caricatural.
1 -
How many people will take the time to read a disclaimer that the image is used only for reference purposes? The many who don't review the attached sources without making changes aren't likely to review background notes on a photo before capturing it for use elsewhere, declaring the find of a photo of Uncle X.
4 -
@LudovicMarsillach - I'm sympathetic to researchers loading AI generated images to illustrate something physical, such as uniforms, without people, because the risk of people thinking that "It's Uncle X" must surely be reduced.
However, it's also unclear to me how this is ever an advance on ordinary photos - if you want to illustrate WW2 battledress for the British Army, you don't need AI, you only need a photo with a suitable copyright permission. And if you do generate an AI version of it, the chances of it being accurate are slim. To see that, let me take a neutral example of generating an AI image of the British science fiction program, Doctor Who. I can tell you from personal experience that the TARDIS (a British police phone box) is roughly the right size and shape but the lettering is anywhere between totally illegible and mostly illegible. And as for the infamous Daleks, any enthusiast would laugh you out of court. And this is with loads of Doctor Who images on the internet to train AI
Basically, AI's current accuracy for illustration is poor. So right now, I'm dubious about my own suggestion.
2
