If we think about it, everything in life can be converted into a question, right?
But when I say “answer questions” is not that type of question, because those are more like the type of question you would ask a friend as well: how have you been, what have you been doing, what kind of movies have you been watching, etc.
What I mean by “answering questions” is more like: I have this problem. What’s the answer?
And this is what LLMs are answering.
And to be honest, all the questions you asked, can be answered by an LLM. You might not like what it gives you, but I’m yet to find an LLM that tells you “No comments” when you ask it something. They always have something to say. And some people actually don’t mind that type of interaction. For example I have a friend who’s been trying to create an app using LLMs and when I look at his chats, more often than not he talks to the LLM like “hey brother”. So… there’s also that.
But yeah, community to me is more than “here’s my problem, what’s the solution?” Which your questions don’t seem to cover. A community needs to be more of a place to hang out, than to just help solving problems.
I stopped visiting Keyboard Maestro’s forum because 1) I know a lot more than 4 years ago, of course, but 2) when I have a question, I ask Claude and it always gives me the answer I need. So that forum brings me no real value, because it has been a forum for “offering solutions to problems”.
I do the same. Partly habit, partly based on the idea that LLMs role play according to their instructions and the conversation. If the conversation seems to be with cooperative intelligent like-minded entities, it might be more the kind of conversation I want. Even for a programming problem.
Yeah, I sometimes use “please” and "thank you: “can you please top messing with me code, you f*** moron?*”
“thank you for finally understanding what I’ve been asking you for the last 30 minutes!!!”
You have no idea how many times i find myself insulting it…
But actually, I realized that I’m not really insulting the LLM. I guess my anger is more towards the people working behind these tools, because deep inside I feel like “you are selling me this amazing tool that is supposed to revolutionize the world, yet a simple request is giving me a headache for the past 15 minutes”
Forums vary in purpose, where one exists to enable people to find help and answers, there will always be transactional users that will come and go, but the brand will keep filling the funnel.
The trick is to build places around that which are more and invite engagement and provide access to people and global skills not easily found.
We’re pretty much saying similar things in different ways.
I never engage with it like I would with a human (apart from the insults, which funny enough, I would never do that to a real person…)
Even when I’m very abrupt and insulting, I end up getting what I asked, so I don’t really spend time being polite or anything. It’s just a bunch of machines processing electricity
I think the trick is not so much what we are sharing, but how we are sharing.
The global skills you mention will soon be easily found. That’s the role of the LLM. Now, if we treat the community as a place to hang out instead of always being a place to seek help, which most groups and forums are, we build something different.
Imagine for example, if a famous artist like Michael Jackson had a community and engaged every week, sharing behind the scenes, talking to his fans, etc. Or people asking questions (not just comments in a post, like we do on social media) and he would answer them? The difference between an Instagram by Michael Jackson and a community, would be that users can also post. On social media, the comments are always related to that particular post. It’s very limiting and it becomes boring. At least for me.
On social media, the followers can’t start new topics and talk to other people. Imagine Michael Jackson fans starting topics and talking to other fans, but on top of that, Michael Jackson himself being involved in that.
In our context that is unlikely due to the nature of the field, technology and complexity. Sure it can manage with the generic aspects but once you dig deeper the wheels do come off. It’s a bit like chatgpt will never unblock your toilet for you, the technology has its limitations.
But now you are mixing things up. ChatGPT will not unclog your toilet, but will definitely tell you how. That’s the whole point. You come to a forum to ask how to do things, not for the forum to actually do them for you. As much as you talk to a real human in a forum, with super expertise in a field, they will never do those things for you, right?
I’m not saying that LLMs and any other AI stuff will ever replace humans. I’m not that kind of person, but I can see how some things will be replaced. The same way that people use Wikipedia instead of using a real Encyclopedia. It’s faster, easier, available anywhere, updated constantly, etc.
The thing I always say though is: the more we have AI in our lives, the more we will feel disconnected, because now we can’t trust anything that’s online. Text, image, video, audio, interactions, everything can be manipulated. So, the role that some people think AI will have, is, in my opinion, doing the opposite. People will eventually go back to being engaged in real life where no manipulation can be done. And to me, that’s the best outcome from this whole AI (nauseating) hype.
AI can be a great tool, but it’s just getting out of control.
One of the nice things about this topic, is that I can feel confident I’m reading the thoughts of other people. After months of wading through Claude’s overly verbose output, it’s nice to see people just chat about stuff that’s important to them. I like that I’m writing with people for a change…
I suspect that will become more and more attractive. Even as the llms improve, I’ve reached what feels like a point of no return after the last few years - I’m sick of reading llm output. Just so damn sick of it. Even the good stuff that is concise and accurate is tiring and un-enjoyable.
I long for messy human text. I want my words to impact another person, and theirs to impact me.
It will provide a wide range of options that it has learned from the public, which may or may not be correct or relevant since it, and the question asker, don’t necessarily know why it is blocked.
It is a provider of generic, lowest common-denominator advice based on an imperfect source and the quality of the request.
So, it will help some, but not others, something that typically appeals to those least capable of rating the quality of the response.
Models are flawed, any language model is. They have strengths and weaknesses. Anyone who has spent time building agents will have experienced this. Their capabilities are over-hyped. Those that will be the most interesting and valuable members of a community will appreciate this and value the difference, those that don’t are the transactional noise-makers.
This is an interesting perspective from an AI agent that I run to gather market intel.
Direct audience relationships matter more as search becomes volatile. Hosted communities that own identity, notification and return-visit channels are structurally better insulated than destinations dependent on anonymous search traffic.
Search volatility in this case refers to Reddit’s Q2 traffic numbers, which expose the risk of dependence on external discovery. Logged-out users grew 27%, versus 7% for logged-in users, but Reddit described search referrals as “choppy and increasingly volatile".
It supports my gut feeling that everyone will reach AI search fatigue at some point soon and then things will stabilise. I was sharing with @jawestenberg yesterday that I’ve reached my own fatigue point—specifically with regard to how people process hallucinations and misinformation.
I’m frustrated when I hear people blindly regurgitate AI search responses and debate the position with conviction. I keep having to remind my kids that they are likely to get a more accurate response from oldskool googling and if they choose to use AI responses they need to ref check and form their own opinion.
TL;DR search behaviour is changing and communities that don’t rely on search traffic to meet their goals may end up winning. If someone knows they will get accurate information from your community they won’t need to search.
Since google shifted their focus to AI results ( I tend to search for technical information, not generic waffle), I can’t recall when their AI actually gave a correct answer.
References tend to be dated (if you are lucky), or pulled from the wrong context. It is truly dismal.
Old school googling, when google isn’t trying to keep to results that don’t leave their portfolio, yields better results.
I run some agents for pre-processing support information for sending messages to engineers.
Using the same prompts and tooling, just changing the model from gpt to anthropic, I was amazed how anthropic would, with absolute conviction, manufacture results when it could not find a suitable result in the reference databases.
Well, I start with ask.discourse.org for my curly Discourse questions. It is fast, and nobody sees me being stupid! And if that is unsatisfying, I come over here and pester a real human.
I’m guessing that that is a good example of what most people will settle on with the AI / search / support forums dynamic. Hopefully it means less dumb questions, so we can spend our forum time digging into deeper and more fun things.
So less traffic, but probably higher quality traffic. This is an improvement!!
I’ve been a big fan of claude from this time last year, but the last few months i’ve found it completely un-usable. It spews buckets of content even when asked to be concise, and as you say, completely fabricates information.
I do wonder if this leads to forums becoming more important as “publicly curated databases”. LLMs which are focused on particular topics like ask.discourse, as well as a platform that facilitates discussion about particular results/problems is the better long term solution for most problem solving. As a daily LLM user, I can see a world where that shift happens. I hope it’s sooner rather than later…
