New Forum rule: No AI-generated topics and replies

Meta
1

The purpose of this forum is to be a place for human exchange and interaction about all things Typst. Unfortunately, this purpose is being jeopardized by an increase in AI-generated topics and replies. This comes with a few issues: Firstly, LLM-written prose is exhausting to read for many people. Secondly, such posts also indicate a lack of effort from the author to invest the time into a proper write-up.

To counter these two problems, we are introducing a new Forum rule:

Please do not post AI-generated topics or replies. If you want to share something with the community, we expect you to put in the effort of writing the message in your own words. If you depend on machine translation, we recommend you use dedicated translation tools instead of LLMs.

Notably, this rule does not mean that you cannot post about projects or other kinds of topics that involve AI or have been created with AI help. But we expect that for something that you consider worth sharing with the community, you also put in the effort to share it by yourself.

31 Likes
3

Which translation services do you recommend? Is DeepL Translator | World's Most Accurate Translator an option?

4

I’m also a bit confused about the boundaries of the new rules.

For instance, I once posted Why does `pagebreak(weak: true)` result in a empty page when wrapped in a `#page[]`? - #2 by Y.D.X. I basically pasted the question to AI, read and verified its reply, and then simplified the answer. Is this kind of post still allowed?

5

I basically pasted the question to AI, read and verified its reply, and then simplified the answer. Is this kind of post still allowed?

That’s okay as long as you write it out in your own words instead of pasting or tweaking the AI answer.

4 Likes
6

I also use DeepL, and yes that’s ok. The purpose of that rule (or that sentence of it) is to not paste a sentence into a chat bot to get a translation. While LLMs can translate, the result will still be text generated from your original. Dedicated translation tools should be better at preserving your phrasing and not just the overall meaning.

2 Likes
7

So while this nears the borders of self-censorchip, it’s still true that “my forum my rules” holds. People who don’t like this posture are free not to post.

That said, “If you can’t be bothered to write it then I won’t be bothered to read it” also holds.

So while I wholeheartedly find LLMs useful, copying and pasting LLM output verbatim into a public forum is a poor practice, even disrespectful. So I agree that rewriting the output results of an LLM in one’s own words means it’s your voice and you can write your own sweet words into the forum.

1 Like
8

While I appreciate the exclusion of AI slop and “chatbots” (generalized term here) from the deepest depths of my heart, I cannot go without raising an accessibility issue this rule in its current form and wording imposes.

While it is deeply disrespectful to slop some AI text and expect other people to read, understand, unravel the slop and reply to it in a meaningful manner (just to possibly again wrangle their work through another LLM), some people might actually use AI for a meaningful purpose: as an accessibility aid.

While I myself am not personally affected, I would appreciate an exclusion from the very general prohibition introduced by @laurmaedje for uses of AI as an aid for access and participation to help and discourse.

Access to computers - and anything which might teach you something about the way the world really works - should be unlimited and total. Always yield to the Hands-On Imperative! - CCC e.V. (and many, many friends and others)
CCC | Hacker Ethics

Please do respond with your views on this matter since LLMs and their uses, dangers and social implications is and continues to be a very disputed social topic.

Please be reminded: I am a human. This is close to my heart. I took the time to write this rather long post. In a second language no less. Please be kind and understanding and constructive.

:)

Thank you all for reading and considering my concern,
Proliecan.

3 Likes
9

That’s a good point; I drafted the addition to the rules and did not consider this. Given that e.g. NLnet too has an accessibility exception, and presumably invested more resources and determined this is necessary, I am for it. I will propose revised wording to add to the rules.

Edit: I unfortunately don’t see a way around requiring disclosing the use of AI for accessibility, and thus also disclosing the fact that one has a disability. The alternative would be indirect disclosure due to the fact that a post was not removed. I hope that is an acceptable tradeoff…

1 Like
10

I didn’t consider the issue of disclosure before.

Your argument about there being indirect disclosure due to unremoved posts seems partially valid only; it is limited by the difficulty of reliably recognizing AI-Posts in the first place.
However this limitation offers no solution to the issue at hand.

I myself can’t think of one and thus can’t offer it here.
I hope others with more experience, expertise and ideas introduce us to their thoughts on this.

11

An alternative would be private disclosure via a mod message.

I have brought it up in the mod team and we will see what kind of solution we can come up with. In the meantime:

We definitely don’t want to exclude people who rely on a11y tools, so if you do and the letter of this rule would exclude you, please reach out. It would be great if your experience can inform a rule that gives us necessary moderation options, without the letter of the rule discriminating. Until then, our moderation practices will reflect that our goal is to be inclusive.

2 Likes
12

Thank you @ensko and @moderators and all the others involved for being open and supportive in this issue and for supporting inclusive access and participation above all!