Weekly site maintenance is scheduled Thursday 04:00–06:00 UTC.
Editing will not be possible during this time, and there may be other brief disruptions. (additional info)

Forum:Standards and documentation/AI Images?

From OpenGeofiction
Jump to navigation Jump to search
ForumsStandards and documentation → Standards and documentation/AI Images?


I want to pose a question regarding site policy: What is the correct stance that we, as a community, should hold regarding AI generated images on the wiki/map, or other officially-sanctioned platforms that may exist in the future?

It is my firm belief that OGF should not condone the use of AI as part of the wiki or the map. AI images are obviously extremely tempting for those of us who cannot photoshop or find a fitting image from the web. However, to me there are two main issues to using AI generated images as part of OGF’s environment.

1. Legality. Currently, there are multiple lawsuits (led by such corporations as the NYTimes and Getty Images) in progress against AI companies including StableDiffusion and OpenAI in the US and UK. The crux of these issues arises from the fact that AI companies scrape the web for data, collecting images including revenge porn, violent imagery, and medical photos (see this VICE article). They also scrape copyrighted content off of websites like Pinterest and Getty Images. Whether or not the images are of copyrighted content, the use of all such content non-consensually raises an ethical issue, as well as the obvious legal one. While it is unclear as to how the courts will rule on this issue, I don’t believe that it’s ethical for companies to do such a thing, as you can easily bypass so-called “guardrails” and create illegal images such as those of Taylor Swift that were recently spread across social media. Is participating in this system really the sort of social message that this site wants to distribute?

2. Creativeness. I’m sorry to all those who love AI, but typing in a prompt is not creative. This site is built off of hard work — hours and hours of it. The day is fast approaching when some enterprising fellow can scrape OSM and OGF and create a tool to generate maps that could then be traced or uploaded outright to this website. If my mapping was amalgamated into an AI map-creator I would feel cheated and betrayed. I would certainly not want that mapping to be uploaded, created by a prompt, to sit next to the mapping I have poured myself and my thoughts into. The people who have thrown their lives into photography and art do not deserve to be treated like that either. By putting AI-generated art onto our wiki, as a community we are endorsing a philosophy that diminishes the creativity and love that is put into this site every day. I fail to understand why it is impossible to simply use some elbow grease and either find an image, draw it, or composite one and appropriately cite the sources with the correct copyright licence. It’s not hard, and it’s what this site should be built around.

I don’t have a real issue with using AI off-site for inspiration, to be clear - it's cool stuff! However, on the wiki and map, using AI feels like an under-handed and lazy trick. I realize that it is, as I’ve said, appealing. But it’s not creative, it sends the wrong message, and it doesn’t even look that good. I know some users have uploaded generated images already: I obviously strongly disapprove and advocate for their removal and the prohibition of AI content on OGF. I would assume that AI generated work on the map would not be allowed - why should we allow it on the wiki? My ultimate conclusion is that as a website and community, I think we should not be promoting or supporting what is, in the end, in the words of Simon Stålenhag, “derivative, generated goo”.

I would like to know what the community and admin team's thoughts are on the matter.

--Lithium-Ion | [1] (talk) 20:11, 27 January 2024 (UTC)

I think using AI on the wiki should be permitted, as the wiki helps to illustrate and explain what's actually on the map. While I'm personally not a terribly big fan of AI imagery, I understand that it may be the best way to properly approximate what an OGF place looks like since photos are not possible, and real-world photos may not appropriately represent part of the OGF map. The wiki has always been subordinate to the OGF map, and from a creative perspective for things that only exist on the wiki (such as "photos" of OGF places) I don't think it needs to be held to the same standards of mapping, where using AI should be much more stringently regulated or outright prohibited.
That said, from a legal perspective I think we do need to ensure that AI imagery is not only appropriately labeled as such but also disclosed as to what generator was used to create the image and to ensure said generation software is compliant with Creative Commons licensing. --TheMayor (talk) 20:31, 27 January 2024 (UTC)