Anyone else seeing a drastic drop in community activity for Support forums?

Gemeenschap
103

If someone can’t discern between something they have written themselves and used an agent to proof it, and giving an agent an “essence” and having it wrap it with verbose text and structure that does not improve the message they are trying to convey, then discussion is lacking a foundation.

Tools exist to improve something. Where that something is simply eradicating the time and effort it takes to construct a thoughtfully considered topic then why should anyone else invest their time and effort to thoughtfully reply and help guide?

You may as well just ask an AI, which is of course what they already did but failed to get the response they wanted.

Engagement is central to any community. When you write 3 pages instead of a clear, concise paragraph then it is an instant turn off to members who may otherwise have engaged.

It is why we have clear guidelines.

Now, we accept that language can be an issue so exceptions are necessary. This is also why we have just implemented Discourse’s full automatic translation stack (driven by gpt as the llm) - it has removed language as a barrier for collaborating.

We have set the bar that the site is for humans and inter-human discussion. Sprinklings of AI to improve that dynamic is not a bad thing, but a whole helping is going to be automatically bounced by the agent that is looking for that content.

We have gone to great lengths to explain why the position exists and to guide anyone, constructively, if they bump into one of the guardrails. For the most part this works well and people appreciate why. But as the saying goes, you can’t please everyone.

Taking exception to the house rules is like being invited to dinner at someone’s house then criticising the decor and food. Attendance isn’t obligatory and house rules exist for the betterment of the whole community, not to accommodate the specific whims of an individual.

This type of content affects the whole tone of a site and has a direct knock-on to who will choose to visit and stay. Each community is different and what works in one may not work in another. If a community has become disconnected from what its membership wants, that becomes obvious pretty quickly, a barometer we certainly wouldn’t ignore because of individual preference and perspectives.

1 like
104

So, it IS indeed an issue of FORM, how it looks and sounds (“tone”). That’s up to each forum to decide if the posting of “AI outputs” is acceptable or not. Maybe a human written summary/introduction with a link to the output (and/or the whole AI conversation) would be preferable.

Or a “click-to-expand“ seems ideal to me.

AI output shown there after a human presentation. Everybody is free to click. Or not :+1:

Or, as mentioned, you seem very focused on “who wrote” the words. If enough back-and-forth, context, guiding, disagreeing, follow up questions, etc. have been done with an AI model, the result has been “produced” by the human using the tool. +proof-reading and finding the result interesting and good enough to post has value.

The issue I see is discriminating any “AI output” as if it was the result of only a single sentence prompt (and of course it will tend to infuriate the user). This “written themselves+proofing” vs “must necessarily be garbage” seems a fairly narrow way of viewing things. Using (correctly) AI as a tool/partner to refine one’s own ideas, research, expand, double-check, is very different from a 5 words prompt. And it will give great results if done properly.

Unless you stay stuck on content “having to” be 100% human generated (if the user “rewrites” himself the AI output, would it then be acceptable, or cheating?)

This is getting quite besides the initial topic and is more about “AI usage in forums” (which is probably a subject).

1 like
105

Or, you know, you can take exception to and criticize the rules of a house where you weren’t invited and didn’t even enter. :joy:

106

Are you talking about me, Matt? I’m NOT criticizing the rules of anyone (and I made it clear: That’s up to each forum to decide if the posting of “AI outputs” is acceptable or not). I was talking about the posting of AI outputs in general. I guess the question I would have for @nickdb about having a “foundation” on the matter is: From his POV, can there be good AI outputs to post? I’m not denying there also are (plenty of) bad ones.

107

As far as I know, a human wouldn’t be allowed to claim copyright in that sort of thing.

108

Whether it’s inherently copyrightable or not, all the commercial providers explicitly assign such ownership and rights to the human that prompted it. E.g., from OpenAI’s TOS:

Ownership of content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output.

And yes, Mevo, obviously, that was about your disagreement with Nick’s stance in his community… but this is all quite off-topic now.

2 likes
109

It is entirely about the words, we would moderate a poor human-written topic as quickly.

The reality is that is a rare occurrence.

What isn’t a rare occurrence is when someone has chosen to be a proxy for an interactive, generative search bot. These provide a very consistent product. The output of which falls short of the quality vs quantity test, adding more noise that obscures valuable content.

Getting back to topic, AI is not solely diverting people from seeking out a good community forum, it is entering the community using a backdoor via its human proxies, making a forum a less attractive and engaging place to be.

Imho.

2 likes
110

The original title is ‘Anyone else seeing a drastic drop in community activity for support forums’

So Matt’s forum is an example of a ‘support forum’. It has other content but it is also a support forum. Obviously Helga’s forum is also a support forum.

Although this may be true, I’m not sure that is what would cause a ‘drastic’ drop in community activity. My reasoning is… there have always been annoyances in using forums. Trolls, spammers, know-it-all users, users guarding their corner of the web, poor moderation, favoritism, you name it

However, with the availability of AI recently (it is built in to google) I see AI as direct competition for a support forum.

I can only speak for myself but if I used to need a syntax example or help finding a bug I would go search stack exchange. If searching proved fruitless, I would put on my flame suit and… ask a question in a forum.

Then I’d wait

With AI, there is no wait and the search is done for you

If I ran a support forum, I’d be thinking how can I address this direct competition. How can I provide a better product. How can I use this to help me and how do I distinguish my product from AI

jmho

3 likes
111

Fair point.

For clarity, we are also a “support” forum.

It is intentionally quoted as it is a bit more grey than that, but that is still one of the driving factors bringing people to us.

1 like
How do you guys deal with users that use AI to automatically reply to threads so they can gain trust level?
112

By the way Nick, I was making a general statement not directly addressing your comment (even though I quoted you)

And I get it. YouTube is littered with AI generated content. It is a MAJOR problem for that space. You see a video that sounds really interesting, you click to watch, it is really really well produced but then… uhg… it is AI

It annoying and you click away

I’m sure the same thing in forum format is just as annoying

1 like
113

Ah, right. “Obviously” if it wasn’t criticizing, it was at least a “disagreement”. And it was also about his community specifically. Everybody should always take it very personally whenever possible, of course. What’s for sure is that it wasn’t me trying to explain the view of people he had “irate conversations with” he was referring to. People he didn’t seem to believe, or even really tried to understand. But who cares about understanding each other, right? Forums and conversations aren’t about communicating with one another, obviously.

My apologies. I was very wrong, as I now know these instances are in any case far too rare to even consider. But some of the reactions of the concerned users are good enough to mention “repeatedly” (his words) to make fun of. I mean, these people “seem to have a fundamentally different grasp on reality and the related concepts” (his words, again. And “the stance”, as the quote was clearly indicating).

My bad. Understanding is overrated. At least it is for some people. Thank you so much for the clarification :heart: :+1: (even if it was, well, all obvious). Now I do remember why I wasn’t reading/posting here since a while and I am going to do that again right now :+1:

114

Wow, I’m really sorry, too. I found this back-and-forth to be the perfect encapsulation of my comment above – that describing the why (in terms of the content itself) has been helpful for us to ensure folks don’t overly object or take it personally.

115

I think this is the problem…

I think the snake will eat itself because of this behaviour. If everyone blocks the updated, accurate, human content from refreshing the models we can’t expect the quality of the responses to be good.

Thoughts?

116

I think they already have everything from us, and the plan is to isolate human interaction to the core. The models started to be trained without human intervention: RLAIF, RLVR, RLVF, DLMA or Inverse-Q.

That could explain why the language and the overall communication methods of the frontier models seem to be great but are really weird strange.

Over the last few years, I got out of social media and almost all chats. Then I bought a lot of books, some musical instruments and returned to analog music players or radios. I plan to start crate digging and get into turntables as soon as possible and stop using the Internet someday.

Well, I think there will always be a safe place to be, but not on the open web.

117

How can they? We have new ideas every second.

118

Oh yes, but outside the box. From a data perspective, they manage almost infinite resources versus a tiny fraction that a human can keep or know.

119

Right, but that resource comes from us. If we close our forums down, we are not regenerating the data source.

1 like
120

Well, I will push a little deep; please split it if you or the team consider this off-topic. I like the OP and understand that my point can be off-side the technical discussion about support forums activity and the impact about the use of IA for this specific websites activity/group.

I get your point and it is a very fair one. I always value your motus, and I really hope for a future where it will like you think and describe. Your interactions remind me closest people in those who I trust.

I think it started as is: humans feeding generating source for a tool used by everyone. But if the tool was made from another people, not the ones like you, I think they can abuse (like spam bots) and motivate more subtle ways to feed re-generate their commercial data-sources.

Not from interactions anymore, because they are being minified and it was part of the original idea. Use a lot of people to train models completing captchas, then make them to pay for refine the tool until they start to be useless because the tool can be trained by itself like it is happening now.

On some point it is cheaper, and better. There are papers on it, I linked one but the acronyms I named before are the most accurate actual training methods out there.

This is social and has different variants, tests, and experimentation. I remember a lot of people isolating some years ago. I really like silence just as much as I like conversation. And I remember that it used to be much simpler to talk to others; today it happens less often, and in increasingly trivial ways.

If someone mentions something outside the context of a particular story, it leaves an uncomfortable feeling. When that’s exactly what gives us the most feedback and overall value. It is okay to be weird.

Perhaps the new forms of training can be realized directly through the body. Big tech patents talk about mining doing exercise. Almost everyone would do it for a UBI; they scanned their iris some years ago, and after a climate collapse, that can be the new honorable thing to do.

With the respect I have for all of you, I think and understand that analyzing this perspective could be painful because all that you are doing here depends on people using the internet like in the old days.

The future is not in our hands, but ours, individually and as a community. We need to allow ourselves to see the truth; the compute capacity of a couple of corporations is incalculable.

When we are talking about this, they are processing all the data in the world from almost everyone who uses the internet. Public, private, everything. And we have no idea what they can do with their private models and prompts.

The iceberg is below. And it will will be always beautiful to watch the stars and the moon. From my humble perspective mantaining or recovering the simplicity is the first resistance rule.