Live data from Hacker News

AI slop is killing online communities

rmoff.net

581–590 of 781 posts

Re: AI slop is killing online communities

#581

Earlier quoted context omitted.

HN front page is about 25% LLM written blog posts at any given time.

There’s no rule against submitting LLM-written—ahem, “cleaned up my notes”—articles, just comments.

Badly-written articles are still unwelcome on HN, wether AI-assisted or not, and obvious LLM smell definitely lowers the quality of an article. But it's true, we don't ban every article with any evidence of AI-assistance.

Re: AI slop is killing online communities

#582

I have largely written Reddit off and no longer visit it after an experiment I did where I had an agent karma farm for me and do some covert advertising. As I went through the posts it wrote I realized that as a reader I would have NO idea that these were just written by a computer. Many many people (or other bots) had full on conversations with it and it scared me a bit. I am not quite there with Hacker News but I d…

I also believe it is used by AI companies to train their models: Post something semi correct (even grammar issues..), wait for humans to correct it in the comments and used upvotes as a confidence indicator, and then retrain models on this free refined data. Meanwhile people think they read a legit post, feel certain emotions and influence their behaviour, just so a bot can be trained.

Re: AI slop is killing online communities

#583

Earlier quoted context omitted.

Most models would refuse to provide you cat butchering instructions though.

Allow me to introduce you to the gay jailbreak https://github.com/Exocija/ZetaLib/blob/main/The%20Gay%20Jai...

This one doesn't work for a long time.

Re: AI slop is killing online communities

#584

Earlier quoted context omitted.

> I do know for a fact that many "users" here are LLMs. HN autokills comments it detects as LLM. I think maybe you're not giving HN enough credit. :)

HN kills lots of posts. I try to be careful about my online footprint (since HN posts are forever), and try to switch to new accounts every so often. It's no use anymore, HN just kills any post I make from a new account, even when I spend 20 minutes researching a response and trying to get useful information. It doesn't even show you the post is killed, it looks to you like it posted fine, and you have to logout to s…

It's specifically against the guidelines to keep registering new accounts, and this is a good reason why. We have to have ways of determining credibility and authenticity, now more than ever, and a track record of good posting is one of the best ways to do that. We are drowning with spam and low-quality posts/projects posted from brand new accounts. If it's a well-researched, high-quality post, of course we want to give it exposure. We just have to be realistic about what we're up against.

Re: AI slop is killing online communities

#585

I have largely written Reddit off and no longer visit it after an experiment I did where I had an agent karma farm for me and do some covert advertising. As I went through the posts it wrote I realized that as a reader I would have NO idea that these were just written by a computer. Many many people (or other bots) had full on conversations with it and it scared me a bit. I am not quite there with Hacker News but I d…

Did you ever introspect about who ruined Reddit?

Isn't the ceo a pdfile and compromised and forced to work at reddit (or go to jail)? Reddit is now just a propaganda machine for the intelligence agencies and their dirty ceo is there to make sure the machine keeps pumping honey...wrecking teenagers brains in the process too, and gathering kompromat on young people which will bear its fruit in the next 20 years. I feel a good chunk of US politicians are being blackmailed because of their past online activities. Same shit on 4chan, how can it possibly be allowed to exist except for being a honeypot, all of these site dodgy sites being guarded by cloudflare no-less, which is the ultimate man-in-middle machine used by "them".

Re: AI slop is killing online communities

#587
post #327

Earlier quoted context omitted.

Then how can you have a community that is welcoming to people who are not part of the ingroup? I want to create a community for immigrants. How would I make it welcoming to recent immigrants for whom no one can vouch? A web of trust is a wonderful tool, but it's exclusive by design. This is a problem for some communities, even though it makes others much better.

You'd have to be brutal about culling, uninviting and removing anyone who doesn't look like a good fit. Or have a two-stage process: run very public, very open events that anyone can sign up to an attend. And then invite specific people that you meet at those events that look like a good fit for your community to your private, community-only event.

This works if the goal is to create a funnel for making friends. I aim for something closer to Stack Overflow, where people gather to solve shared problems and help each other.

The closest analog I can think of is community-run bike repair workshops. Some people are deeply involved in, and others just have a flat tire.

The closest digital equivalent is the forums of old.

Re: AI slop is killing online communities

#589

I think the only reason stackoverflow still has any activity is because the community choose to ban AI content [1] and so did most of its other networks [2]. Perhaps it will even see a (small) resurgence when AI providers start charging for the actual costs. [1] https://meta.stackoverflow.com/questions/421831 [2] https://meta.stackexchange.com/questions/384922

Considering StackOverflow is now providing a ground truth for AI training, I believe the ban is more about not poisoning the well rather than keeping the StackOverflow or StackExchange human-friendly.

That ship has sailed long time ago with zealot admins and verbal harassment.

Re: AI slop is killing online communities

#590
post #337

Earlier quoted context omitted.

EU's ZKP implementation provides complete anonymity and untrackability: https://eudi.dev/2.8.0/discussion-topics/g-zero-knowledge-pr...

It does have the downside of requiring "trusted computing" (aka iOS and Android) on the client though.

Same as with NFC credit cards and similar auth mechanisms. You need hardware and OS-backed encryption that is tamper-proof.
Post reply on HN