Can we / should we add something about no AI slop to the community guidelines? Many open source communities just say no AI at all.
In my reading of the guidelines it could fit into a few of the categories, but isn’t explicit.
I’d be in favour of making this explicit, although exactly what constitutes ai slop is blurry…
Any thoughts?
If you mean using it to take something seriously that I am sure some people here (who want to help others) feel even “trolled”. Yes, we need a conversation about it.
This is easier to identify than to know if some priest used Copilot completion or whatever. (and a bit ridiculous to write that officially)
I think talking is always the most useful, not the “enforcement” itself.
Can we / should we add something about no AI slop to the community guidelines? Many open source communities just say no AI at all.
I would be very much in favor of it! At the very least, we should prohibit AI generated topics and responses.
For example, I really don’t want to see stuff like this: Suggestion: Developing a Cross-Platform Script for DSP Extraction, Conversion to Faust, and Integration with SuperCollider
Apparently, machine-generated content already violates our ToS: Proposal: Unity Asset Store or GitHub Package Integration - #14 by jamshark70
By making Content available, you represent and warrant that: … the Content is not spam, is not machine- or randomly-generated…
Does this apply to AI generated posts? (IMO it does.) If so we should make it explicit in the ToS.
(Side note: for some reason I can’t access https://scsynth.org/t/terms-of-service. Anyone knows why?)
My one issue I can think of is translation.
As a native English speaker I’d rather read ‘broken’ English (is there even such a thing as non-broken English, I certainly don’t speak it)… But I can understand how non-native speakers might struggle with reading less than perfect English. If someone has any first had experience with this it would be beneficial to the conversation to share!
I suppose this isn’t technically related to ai ‘slop’ but might be worth clarifying the difference, or where we as a community stand on it?
Me neither? but it’s here Terms of Service - scsynth
And the relevant passage
the Content is not spam, is not machine- or randomly-generated, and does not contain unethical or unwanted commercial content designed to drive traffic to third party sites or boost the search engine rankings of third party sites, or to further unlawful acts (such as phishing) or mislead recipients as to the source of the material (such as spoofing);
There are certainly people on this forum who are using AI for translation/editing – and I think that’s ok!
I suppose this isn’t technically related to ai ‘slop’ but might be worth clarifying the difference, or where we as a community stand on it?
I think the crucial difference is that the content has first been written by a human.
There is a corner case where you ask the LLM to “embellish” your original text, which can lead to annoying walls of flowery prose.
Me neither? but it’s here Terms of Service - scsynth
Thanks! Apparently the link has changed.
The tos might indicate this is not allowed then (as it could be considered machine generated), but I also think this is fine.
This seems like a practical distinction and might be worth stating explicitly.
I also don’t understand the difference between a term of service and community guideline?
I think this should be a community guideline as it seems like be less ‘legal’ and more about behaviour.
Unless the person speaks zero English, maybe it will no be that hard to communicate.
Why not a grammar checker and an anti-AI checker? It will help the cops.
I love when natives switch you’re and your. It’s fun for us even after a decade living here
I can speak on my personal experience. I’m not a native speaker. I used AI on my first topic here in the forum. It was about a project I deeply cared about (my first harmonizer for SC), I was nervous about sharing it with other people, I was unsure about my English and whether or not I would be able to communicate my thoughts properly, so I wrote a draft, asked GPT to correct it and published the corrected version.
I didn’t do it out of malice of dishonesty, but simply because participating in a forum where there’s a lot of technically skilled people was something that was new to me, and that led me to feel insecure.
The same was true for the README file to that project, for the same reasons (not the code though. I’m perfectly capable of writing my own code).
Only after have I came to realize the ethical issues surrounding authorship the usage of AI. I was familiar with the issue of AI generated code, but AI generated text was not something I’d really thought about. All my posts since then have been thoroughly written by myself.
I just wanted to share this experience as an example of what may motivate a person to go astray and use AI hehe. May the SC gods forgive me!
That’s a good point to clarify. The thing that this thread is objecting to is related to effort vs benefit. In your case, you put effort into a project, and there’s a concrete benefit to the community (an extension providing a new feature), and we all respect that! If an LLM polishes some text, it doesn’t negate the benefit to the community in that case.
But the post that triggered this thread inverts the effort vs benefit: low effort to generate the post (e.g. the lack of research into legal consequences of reverse engineering proprietary algorithms), plus there’s no benefit to the community unless others put in the effort and take on the cost, without the OP materially contributing. Using an LLM to puff up this type of unbalanced dynamic is not what I’d like to see.
So yes, I think, case by case basis. +1 for touchups when clearly effort has been put in. -1 for “please debug my Claude code” questions. -100 for massive engineering proposals for others to implement, that use screenfuls of text to hide the fact that they haven’t been thought through.
hjh
I totally support this!
The whole ethics on AI is its own big topic, but one thing that I find peculiar and really annoying around AI is that it is often used to fake authorship.
AI can be helpful and an enabler, but I think it would be nicer if people are transparent about its usage when writing a post or code - assuming they are fine with the ethical implications of using such systems.
I think it is totally fine for people to use translators or try some things out via LLMs as an enabler, but I also encourage everyone to be transparent about their “shortcomings” - it will most likely be appreciated more than generic text or code and this is how learning and developing a style works within a community. Embrace the error is a very good credo of the livecoding community which I think also suits the SC community well.
BTW - what exactly are the community guidelines? Are we talking about the CoC or does the forum has its own rules?
I’d propose to add something like the following to the CoC, feel free to enhance/edit via https://hedgedoc.musikinformatik.net/sc-coc-ai?both
AI slop and AI based low effort contributions are not tolerated and will be judged by moderators discretion on a case by case basis.
Since almost all LLM providers also train on their input data it is not allowed to paste other users content or code into LLMs without their consent.
Do not use LLMs for arguing or discussion! Another person is communicating with you, so you as a person should reply and not forward this task to a chat bot.
AI tools such as translators or even LLMs may be used at your discretion to enable participation or contribution with the community.
At the same time: Please always try to participate without any such assistance, it should be used as a last resort.
You can assume that the other side is always preferring your own words, code and style over some generic AI generated content. The SC community appreciates the error over generic content.
The moment you use any kind of AI tool to generate text or code you need to be transparent about it and add a disclaimer.
PS: I also proposed to guard this forum with Anubis to protect it from AI bots and AI scrapers, see Guard this forum with Anubis - I think the fight against AI slop needs to be fought on at least two frontiers.
@Siriil thank you for sharing!! Your project has clearly has a great deal of care and thought put into it and is the kind of project we all want to see more of on here. — FYI your English is great!
participating in a forum where there’s a lot of technically skilled people was something that was new to me, and that led me to feel insecure.
This is a feeling that I suspect many people have that perhaps we as a community could do something about… I don’t know what, but this could be discussed in a separate forum post.
I often hang out on here because I like helping people with supercollider, music and technology… keyword being ‘people’. If I can’t trust that I’m talking to a person, I’m not interested in helping. If others feel the same way, that means over time there will be less people helping others, which is community decline.
BTW - what exactly are the community guidelines? Are we talking about the CoC or does the forum has its own rules?
Yes this is confusing…
https://scsynth.org/guidelines
https://github.com/supercollider/supercollider/blob/develop/CODE_OF_CONDUCT.md
So there are three things, guidelines (scsynth) (which is a copy paste of the code of conduct), terms of service (scsynth, seems more legal), and the code of conduct (github/supercollider as a whole).
I think your suggestion should go in the code of conduct over on github, and be copied back over on this forum.
Would you like to turn your suggestion into a PR, it reads well to me!
I’m a bit unsure what the review process would look like though as this isn’t a ‘dev’ thing and we want to make sure everyone has a chance to see the change as it is community wide.
@jamshark70 with your moderator hat on, do you have any thoughts on what happens next in terms of process?
This will probably be a very unpopular opinion, but I think that LLM’s being able to scrape and understand SC actually helps adoption for new users. The documentation is very dense, there are many undocumented features that only exist on this forum and the old messageboards, and there are so few active users that it is hard to find advice when you need an answer to a specific programming problem sometimes.
I will occasionally use LLMs to help with SC programming problems when I can’t find a solution, as they have got markedly better in the last year (although they are still unreliable for most issues). Maybe it’d be nice if there were a community driven project for a self-hosted model based on the sc docs/forums…
I somewhat agree with this actually and seems like an important thing to discuss, particularly if it helps makes the docs/forums more navigable… but lets keep this thread specifically about users posting machine generated / ‘low-effort’ content, and avoid ai as an information retrieval tool. @asher Would you mind re-posting this over here instead Guard this forum with Anubis?
I don’t think that anything suggested in @dscheiba’s proposal would impact this use case and for ease of implementing this change, it would be good to keep it that way, as agreeing on a guideline for AI as a whole might be really difficult, but user generate forum content seems pretty unanimous (so far!).
Code of Conduct and Community Guidelines seem to be in the overlap. I would say that Code of Conduct should contain community guidelines - which I consider an expose of expectations and advice how to behave in the community - while Code of Conduct can expand on this with explanations of what is a total no-no and will get you banned without discussion, etc.
Well, I think it’s more or less semantics, but Code of Conduct feels to me like the community takes Trust & Safety seriously, while guidelines can be more or less anything. So I propose (and vote) for replacing Community Guidelines with Code of Conduct and that’s the only document what explains all the protocols, hard rules, and also expectations (soft rules, advice, etc).
While Code of Conduct is a dynamic/fluid document which can change over time with community itself and consent, I think terms of service are a legal document that protects the project (community platform like this Discourse space) for being legaly reponsible for certain things.
I think the community has to make an effort to be kind and welcoming to anyone, especially those who make mistakes, are clumsy with english as their non-native language, or with projects that are in alpha state, experiments with clumsy code, etc. SC is actually big and complex and has a steep learning curve for beginners, especially those not coming from CS background, for example. By putting something like Code of Conduct in a visible space, and enforcing it properly (not saying it isn’t being done) we maybe support existance of a culture of helpfulness where newcomers don’t feel afraid to say “something wrong” to reach for LLMs to articulate their thoughts. That said, I totally agree with @jamshark70 and their “scoring table” wrt how LLMs are used.
I disagree that LLM chatbots are useful in adoption of new users. In my opinion they are useless most of the time. They get things wrong, and we should rather put an effort that people feel safe asking questions on the forum regarding their projects or any confusion when trying to make something work. In that regard, everybody in the community should be discouraged making snarky-funny comments and articulations that can appear as condescending or patronizing. These micro-tones can quickly give wrong impression to those new to this community, or those who are from disadvantaged groups.
EDIT: I realised I have touchd the topic of this thread too tangentialy. To that effect - I think the articulation proposed above is pretty on the point, machine-generated content and “over-use” of generative AI in SC spaces shouldn’t be allowed - but can also be looked at on case-by-case basis. - and I agree that it is added to Code of Conduct (or Community Guidelines)
I tried to clean up this text by removing some of the text that felt redundant and also transformed into passive voice (i hope that’s the right term, “it is expected”, instead of “you must”):
AI slop and AI based low effort contributions are not tolerated and will be judged by moderators discretion on a case by case basis.
Since almost all LLM providers also train on their input data it is not allowed to paste other users content or code into LLMs without their consent.
Using LLMs for arguing or discussion is not acceptable!
AI tools such as translators or even LLMs may be used at your discretion to help with participation or contribution to the community, but should be used in moderation, or as a last resort. “Imperfection” and “your own words” is prefered over AI-generated or polished content. The SC community appreciates the error over generic content. If AI tools are used, community expects complete transparency about it.
Just to add, as I understand it, although the LLM hypers have a lot of funds to burn on lawyers, the legal status of ingesting all the free/open source software and stripping it of licenses is still not settled.
Also supercollider’s license defines source code as “the preferred form of the work for making modifications to it”. That wouldn’t include e.g. minified code, and perhaps doesn’t include LLM generated code either, if it’s not written by human and not designed to be read and modified by one. In that case mixing handmade code with low effort LLM-generated code breaks the license when shared.
Yea, that’s a broader and even harder problem. (And it is a really complicated problem, we would need some lawyers among us. It is really hard to write effective English for this stuff in the legal sphere. It is not by chance that the GPL license is not translated).
I second your concern, we just need to be aware of the “rules of engagement” on this battle in the real world. I critized (and pardom my sarcasm before) when it is done in a way that is not effective about… anything. really. and it is not impossible to see something written down officially being corrupted in practice.
EDIT: @jamshark70’s effort/benefit framing is sharp. Maybe that the core of what make some of us feel “trolled” here. I personally feel that as the important ethical question for us, inside this community. I, not a lawyer nor native English-speaker, and try to focus on “genuine effort” as an ethical guide. Something like “Contributions should reflect genuine effort” and supplement that with James’ rationale behind it. I feel it would reflect our ethical understanding (effectively) without much debate.
EDIT2: A serious discussion about some other important problems (the AI and grammar checker controversies, some a bit annoying sometimes) can become a habit hole. Maybe save this to another thread.