Page MenuHomePhabricator

Adding prompt for AI-generated media to info requested by UploadWizard
Open, Needs TriagePublicFeature

Description

Feature summary (what you would like to be able to do and where):

When uploading an AI-generated media, it should be available to users the possibility of adding the prompt that was used in generating such media, since it's relevant metadata that would be lost and unrecoverable later. Ideally, this prompt should be recorded through the recently introduced {{Prompt}} template.

Use case(s) (list the steps that you performed to discover that problem, and describe the actual underlying problem which you want to solve. Do not describe only a solution):

All uploads of AI-generated media by Commons users.

Benefits (why should this be implemented?):

It would save important data to be had while uploading the media, such as the prompt used.

Event Timeline

Thanks for creating this task! Regarding "should be available to users the possibility," I think that's too weak of a nudge. AI-generated images with prompts are much more useful than those without, so it's a piece of info we'd really like to have. And yeah, there may be some small subset of AI-generated images that were made in a way that didn't involve a prompt, or some small portion of users who won't complete the upload if forced to give the prompt, but on the whole, it's something that I think users will provide if we ask for it.

So I'd prefer the software to be designed in a way that really encourages users to provide the prompt, and makes it possible but not super easy to opt out. (I might feel differently about this if AI-generated images were hard to create and something we were really lacking, but the opposite is true — we're being flooded with them and trying to decide how to determine which ones to keep. So if a strong ask for uploaders results in fewer uploads of AI-generated images, that's not a big loss.)

We're not at all getting flooded by them. We're getting flooded with low-quality photographs of mundane things which we already have a thousand times over, not a handful of low-sized AI images that are usually among the first in their respective categories. An issue with requiring prompts is that people who have little experience with and apparently bias against AI images will delete (nominate or vote) them based on their flawed understanding of how these tools work in more sophisticated processes. And they're not necessarily easy to use if you a) have something specific in mind and b) that specific thing is not as simple as "a cat on a table". I'd be all for extensively encouraging prompts but that just leads to it getting nominated because it happens to have some objected or misunderstood term in it.

The general sense I got from reading this very long discussion (in which you were an active participant) is that many editors do feel there is an issue. I don't see the connection between an image having a prompt recorded and it being more likely to be nominated for deletion.

I didn't say anything about that. Re your second sentence I explained why in rough terms. Here is one concrete example but there are also many other cases where people would probably assume the text in the prompt exactly matches what the prompter meant to illustrate/visualize; for example because my explanations in that DR have not even been understood.

In your example, the fact you attached the prompt to your upload helped the community identify that the image was a derivative work (and thereby a copyright violation), and the community decided to delete it over your objection. Sorry, but that looks like a good thing.

No, it's not a derivative work. You just don't get how prompts are being used. Naming some examples in the prompt can be helpful if the tool doesn't get the concept otherwise and the image didn't look like the term in the prompt at all. Ignorance is not strength.

Only to report that User:HaeB noted in discussion that in Commons:AI-generated_media#Description, guidelines require the following:

Whenever you upload an AI-generated image (or other media file), you are expected to document the prompt used to generate the media in the file description, and identify the software that generated the media.