The advent of LLMs really opens the door to shunting off these “community members “ who’d rather contribute in misanthropic ways for the lulz than either leave or not contribute at all. They can take part in an interactive echo chamber that gives just as well as they can. You don’t even need a powerful model so the overall costs to the community are probably lower than the alternatives of trying to coexist with community-arsonists.
I spent years trying to find ways to bring people productively “into the fold” but eventually realized that it is futile in some cases because there’s zero value to the individual or the community to find a middle ground. They want to see things burn, and the community simply wants them out.
You're talking about using a machine to detect social undesirables then quarantining them in the matrix.
I’m specifically referring to people who have seemingly made it their sole purpose to create as much indiscriminate damage as possible.
You can ban them, block routes for them to attempt to Sybil themselves back to having accounts, etc. but even with great moderation tools and systems, it’s extremely difficult to set up a strong enough set of controls which don’t adversely impact everyone else who you want to have participate in the community.
Yes, a shadow environment is dystopian. It’s not my nature to want to even consider using one.
But we’re talking about privately run communities which also deserve to exist to serve their purposes.
So given the choice between anarchy which drives away people who contribute to make the community what it is and a shadow option for those actively working against its interests, I’ll consider the community first.
You may have misinterpreted my comment. I’m not suggesting you use LLMs as moderators. I’m talking about using LLMs as participant “members” of this shadow board to interact with someone whose account was flagged by a human moderator.
There is no perfect tech solution to a human problem. But in my opinion, having access to a partial mitigation is better than no mitigation.
---
Can you please edit out swipes like "Do you hear yourself?". This is in the site guidelines: https://news.ycombinator.com/newsguidelines.html.
If you wouldn't mind reviewing the other guidelines too, we'd appreciate it. Note this one, for example:
"Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith."
If you do that, let me know and I'll unflag the post and collapse the moderation bits ;)
Further it may still make sense to use human reports to gate some automation even if it slows response.
If you ban words, people will introduce typos / different spelling. If you ban concepts people will change the concepts (see the "unalive" thing that's popular now). If people want to be miserable, they will be miserable. And they'll invent new words, or use euphemisms to get their point across.
Use a cheap purpose-built LLM like OpenAI's free moderation endpoint to classify the text and send the original text plus the classification to clients, and let clients choose what to do with it, with opinionated defaults appropriate to the app.
Maybe you still need to identify persistent bad actors rather than acting only on content. But still, allow clients to decide what to do with that information.
I suppose my thinking is that strong default automatic moderation that's invisible to offenders is a requirement for a project like this to be able to offer a welcoming experience to users, but putting the power in an LLM and fixed filter lists feels very wrong. So my thought is to use those things to give the client power. But maybe that makes no difference if nobody changes settings away from defaults anyway.
i would think they should have the freedom on the site they build and host to choose the impression they give. and sure, if they choose to let their square be filled with noise rather than signal, that’s absolutely their choice. but they also may choose for it to be filled with signal rather than noise. the key thing is the site owner should have the choice to give whatever impression they want for their creation.
again, if they want that impression to be hijacked by noisy trolls, they can choose that too.
if we limited it to client side only, the creator of the site has lost their choice of how they prefer their site to come across.
and for sure, some people may absolutely prefer to send an experience full of noise that overshadows the rest of their site.
Was planning to add github oauth to get a known identity and persistent messaging so visitors can chat with each other across sites.
Instead of a webmaster adding script to their site, it was a browser extension.
The intent was two folds:
1. Get to know other people having similar interests,
2. Try something on the lines of a decentralized chat/messaging system.
Site-local chat was meant to be the default way. But, was not restricted to that. People could keep whitelisted list of sites that they are open to get pings/DM's from.
I was still not able to figure out the privacy-focused, and local, interest matching part since it was meant to be p2p (without any server storage), and local storage can easily be tweaked.
If this was solved, my plan then was to automatically suggest people who may have similar interests depending on the sites they browse, for how long, and the messages they generally send.
Kinda like tinder, but, you don't have to sign-up and install no app.
I have similar moderation concerns in my browser game/engine but I only ban offensive slurs not swears, but I give no visual affordance that the word is not allowed
The only surface where players see the input content is in a share card, and if they finish a game and get to the share card they will find the offensive word has been REDACTED lol
So it’s a long feedback loop just to find out your hijinks lead nowhere
Having my game aesthetic + slur on a sharing artefact is a great way for people to think my game encourages that crap and to “cancel” the game.
Not that I agree with, or care about, cancel culture - but I don’t want people in general associating my game with slurs.
I’m all for swearing and saying funny things that are not pc, but I draw the line at harmful offensive content.
Do what you wish but do no harm unto others.
Is this an appeal to the law or something? It's not a very good argument. Many things which are harmful are not illegal. Many things which are illegal are not harmful.
Also, that's not even how divorce works!
> no one is claiming words are physically harmful
"speech is violence" is a position that some people hold.
> if words had no power, we wouldn’t fight so hard to protect them from government abuse.
People fight hard for all kinds of nonsense. Historically, people have fought hard for things that didn't matter at all.
>> if words had no power, we wouldn’t fight so hard to protect them from government abuse.
>People fight hard for all kinds of nonsense.
you're wrong, moderation is needed in ventures like this
They have the privilege they forfeit the moment they try to hurt others. If they don't behave like cunts I don't care, but if they do, I'll use any tool I have at my disposal to bar them from the space they don't respect. They can talk to their kind in the nasty spaces anyway, so it's not like they're in the solitary confinement.
Being a racist asshole carries some inherent risks too, like people choosing not to let you enter their spaces.
i’m so sick of people just adding dipshit noise to every single place they can and making it impossible to have normal ass conversation.
at the end of the day, these people need to realize the simple shit we realize in like 1st grade: if you’re an asshole to everyone around you, no one will want to be around you.
it really is that simple. but for some reason some of these people struggle to understand basic ass things little kids learn easily.
1. Theres a near endless supply of assholes
2. If youre loud enough and an asshole enough you'll find other assholes.
3. Your new found group of assholes can then go off and terrorize other people while attracting more assholes.
Lo and behold, the EXACT person I had in mind when making the rule, moaned about it.
Works as intended.
I contemplated implementing it on my site for a while but decided I didn’t want to add the JavaScript. Still find it a really cute concept.
A hade a lot of very funny and meaningful interaction with people visiting the sites
I deeply appreciate the honesty here. I’ll pass based on the vibecodedness of it but perhaps it won’t be a dealbreaker for others.
Also, in this context, security.
Also, it seemed to be disconnecting quite often, maybe a hug of death thing.
I wouldn't add it to any of my sites due to the unmoderated nature of it - seeing some fairly unsavory things in your demo - but that's just a little tweak, I'm sure!
On the details like benches and trees, you can also customize it. YOu can add as many props as you want, wherever you want, and style them accordingly. :)
I know it might be too much to ask for, but it would be great to have the ability to add your own props, like an icon and how to animate it. I would love to add maybe cats or drones icons that also move around!
edit: downvoters either know something I don't about its unsuitability, or have outdated info on it and think it's against terms to use it for reasons unrelated to GPT. it's not against terms.
Having read some of the comments, I'd happily use if there was a way to blast (prefarably with a Doom shotgun) some of the miscreants from appearing, but only on my site, and maybe some filters (slurs, etc) that auto-ban them that I can set. Other people could moderate as they will but I'm kind of tired of the toxic people ruining everything.
But I'm also adding more moderation functionalities. Feel free to send me an email with more ideas :D
Nevertheless, in general, the sites using TownSquare are much much quieter then the landing page.
People are flooding the channel with messages, causing the widget to use too many resources on ios; and the website is being endlessly reloaded.
Any idea how to fix this? Bonus points for user friendly non technical solutions, ie is there a way to design an online social space where people want to collaborate and their first thought isn’t trolling?
Whatever it is, if something hit the crowd, you will end up with people trolling. No matter how much moderation you add.
You can check it now and things are much calmer and friendlier. The same if you go to other people's websites using TownSquare.
Regaring the mobile resources usage, I'll need to fix it. I'll be working on it next.
Is there a way to throttle iframe cpu usage ?
You could probably queue messages, cull old messages, cap concurrent connections. That may be the issue is too many peers
Edit: a super smash brothers-like would be fun too. Maybe the page itself could be the platforms.
Internal link to blog gives 403 https://townsquare.cauenapier.com/blog/
Townsquare is probably not the best name - there are tons of things already with that name
Regarding the name, I'm exploring other ideas, but that's not hard.
I did the "go to next town" It worked, but then I got to (https://emilesilvis.com/) and got stuck. The Town Square is in 1/2 the page, so you can't get to the two ends to be able to go to the next town.
I was toying with the idea of making a little crowd representation sticky at the bottom, like watching a screen together :)
Regardless, I love concepts like this. Thanks!
I don’t know all the answers, but joining people together, and reminders there are real people still there as we’ve moved from the physical world to the virtual world is an important part of the progression. In 50-100 years we will either have this epiphany or we will fail miserably.
I could see it being used as some avenue for saying toxic stuff, so I'm suddenly feeling like the only allowed actions should be ones that can never be harmful (like text interactions are just choosing from a pre-set list)..
It already has block/ban and block word list, but unfortunatelly I haven't added them in the settings for the landing page before HN made it front page.. Opsy
However, there is still lots of room for improvement.
But I get your point.