Jump to content

Wikimedia Commons AI/Essay

From Meta, a Wikimedia project coordination wiki
(English) This is an essay. It expresses the opinions and ideas of some Wikimedians but may not have wide support. This is not policy on Meta, but it may be a policy or guideline on other Wikimedia projects. Feel free to update this page as needed, or use the discussion page to propose major changes.
Translate
We should protect human-made art from AI-generated art. In the near future, art made by AI will be considered more perfect than art made by humans, so art made purely by the natural creavity of humans will slowly be extinguished.

Technology is developing at a rapid pace. Every day, AI-generated media is becoming better and more realistic. Despite my belief that AI-generated media is also a form of art and human creativity, I find it important to make a clear distinction between media generated by AI and media created by humans, by which I mean any content or artistic expressions produced directly by individuals without the assistance of artificial intelligence or automated processes. My expectation is that in the very near future, AI-generated media will be indistinguishable from media created by humans. I find this regrettable because, to my knowledge, creating photographs, drawings, films and music requires more effort than simply telling an AI what to generate using a (relative) short prompt.

This is why I think Wikimedia Commons AI should be created; to protect pure human creativity and art, and ensure that it is passed on to future generations. The purpose of Wikimedia Commons AI is therefore to form a collection of media separate from content or artistic expressions produced directly by individuals using artificial intelligence or automated processes, for the purpose of protecting and preserving images, sounds and videos created by humans.

How Wikimedia Commons AI would work

[edit]

When people would like to upload files to Wikimedia Commons, they would first be asked whether the file was created by a human (such as a photo, drawing, screenshot, recorded video, voice, real-life sound, etc.) or generated by AI. If users state that the file was generated by AI, they would automatically be redirected to Wikimedia Commons AI.

During the process of uploading media files on Wikimedia Commons AI, certain requirements should be fulfilled. The following details need to be specified:

  • A name, description, human author, date and category, just as with human-created media on Wikimedia Commons.
  • The name and version of the AI software used (such as DALL·E 3, Craiyon V3 or Midjourney V5). The media file would then automatically be placed in categories like [[Category:DALL·E 3]], [[Category:Craiyon V3]], or [[Category:Midjourney V5]].
  • The input prompt (the original prompt written by the human).
  • The prompt as modified by the AI (the prompt with any additions and adjustments made by the AI).

All existing media files on Wikimedia Commons that have been generated by AI should be moved to Wikimedia Commons AI. Wikimedia Commons should no longer contain media files generated by AI. This should be monitored and moderated, possibly through the use of integraded detection tools, although I am not sure how long these tools will remain effective, considering the continuously improving realism of AI-generated media.

Policy

[edit]
A clear example of a misgeneration of an AI-generated image of a boy.

AI-generated media on Wikimedia Commons AI will be held to the same standards as human-made media on Wikimedia Commons. There are a few exceptions. This includes:

  • Lacking in originality of limited creativity in design, for example in the case where similar images already exist on Wikimedia Commons, making an AI-generated media redundant.
  • Incorrect grammar and spelling, like "PHILSOSHIGY" instead of "PHILOSOPHY". The only exception is to display, for educational purposes, how AI used to generate texts in the past.
  • Low-quality material, such as images that are clearly distorted or deformed. Also here, the exception is the purpose of showing how AI-generated images were generated in the past.
  • Deepfake and misleading information about a person, for example Morgan Freeman seeming to say things he has not said.

Interesting things to read

[edit]