The use of generative machine learning models to generate user facing content is banned.
Furthermore, the usage of tools whose primary use is generative machine learning is also banned for user facing content.

Generative: The model produces more meaningful information than was input into the model
Machine learning: An algorithmic system designed to replicate a given set of training data
User-facing content: All content directly viewable to the user (EG, text, images, audio video ect, excluding CSS, Javascript and other parts that communicate to the browser and not the user)

Under this standard:
things banned
GPT, midjourny
Auto captioning (we would need a policy for an exception here)
AI voice generation

Things not banned
Grammery, DeepL: Machine learning, not generative
The 001 image generation software: Generitive, not machine learning
ACR: The generitive part is not ML, the ML part is not generative
Smart layer blending/Smart infill/ AI image upscaling: Again not banned as not generative
AI generated CSS or Java script: not user facing content
the music stuff that was discussed in 04
Idea generation

There is some forms of AI image generation that is not covered by this policy, but anything that requires a week of work by someone with a masters degree to do is something I don't think we need to worry about banning.

LORDXVNV

Lt Flops

I do think it's probably also a good idea that the rule that forbids AI art still stays there even if it's superfluous as art is probably the most effected by AI-usage on the wiki.

"A maximum of 5 layers in a neural network" is not a useful metric.

I should have a more useful definition soon.

That said, seconding.

That said, I do support explicitly codifying what we're intending to ban in greater detail, rather than just slapping a "No AI" sticker on the Wiki and calling it a day.

I've argued in the past^{(1) / (2)} that when creatives, as a practice, rely on AI to generate creative works, there may exist stultifying effects downstream from engaging in that practice that stunt their growth as an artist/author. However, as it relates to on-site works, this is not the only possible use of AI. There are contexts in which you can use AI to do wondrous things to expand the accessibility and interactivity of works on the Wiki without entering the territory of machine-generated imagery (be that textual, artistic, or something else). The proof is in the pudding:

1. Many of the tens of thousands of images on the Wiki lack alt text.¹ There exist AI tools that can take image inputs and produce alt text that screen readers can parse. As a caveat, this technology is still limited: it's prone to make mistakes, act with bias, and exhibit context-blindness, so a degree of human intervention is necessary. (That said, the honest Wiki contributor should also intervene where necessary when using Grammarly and DeepL, because these tools are similarly imperfect.)
2. Hubs that contain lists of articles often include flavourful excerpts from said articles to pique reader interest. AI-based tools, like ChatGPT, are able to select (what is deemed) the most interesting excerpt(s) from an article, thereby maximizing the likelihood that readers go on to read the article in question.
3. Algorithm Curated Recommendations. Self-explanatory.

It also doesn't make sense to limit this discussion to the end-user. This policy extension, if enacted, should also concern staff. For instance:

1. Staff members may make use of tools that detect machine-generated text content, to help confirm whether select articles were AI-generated. These tools themselves make use of AI in their operation.
2. Staff members may also make use of tools that detect plagiarism, some of which make use of AI in their operation.

Hence, it would further benefit the Wiki as a whole to consider whitelisting in our policy the use of some or all of the AI-based techniques I just laid out.

TLDR: I support extending our existing policy into the domain of machine-generated sound content. But our opposition to AI shouldn't be dogmatic. We don't oppose AI along the lines of a neo-Luddite moralism. We oppose it only where it conflicts with the values we hold as a writing community.

This post was not written by ChatGPT.

Proposal: Stop beating around the bush. We're blanket-banning AI stuff.

This is a general policy discussion. I am uninterested in hashing out the definitions of AI, because that is not the assignment here.

The point is that we are extending our current policy of "blacklisted until otherwise stated" to encompass all AI-generated works, not just text or images. This will include, but is not limited to, AI voices and whatever else nightmarish stupidity techbros come up with.

We will grandfather into the whitelist the stuff already allowed in the original policies:

the use of tools which use AI or AI-adjacent scripting in their operation (such as Grammarly, DeepL or GPT) to assist with grammar, critique, translation and idea generation is allowed

Again, the point here is not to try and define what is and isn't AI. That can be debated later on. The point to make here is that we are expanding the scope of the current policy to encompass AI everything, instead of having two separate but effectively identical policies for text and writing, and nothing else.

tl;dr: blanket banning all ai stuff