Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI slop is killing online communities

rmoff.net

Post reply on HN
Print view View original

711–720 of 781 posts

Re: AI slop is killing online communities

#711

Post by mh- »

mrkstu wrote: post #702
RankingMember wrote: post #683

Earlier quoted context omitted.

Hackernews' approach is hellbans. FWIW that seems to work pretty well here at least.

The combo of quick bans and incrementally earned rights seems to be a pretty effective combo.

I think user-driven flagging is starting to show its cracks here, though. I'm starting to see stuff flagged to dead status, from otherwise reputable posters, because they went against the current in a thread.

I really think HN should consider giving out limited mod points, like Slashdot did (does? haven't used it in 25 years.)

This would probably be the ideal* system for somewhere like Reddit, too.

* Defining "ideal" as what drives high quality discourse, not what drives engagement.

Re: AI slop is killing online communities

#712

Post by Terr_ »

AnthonyMouse wrote: post #362
fidotron wrote: post #77

Earlier quoted context omitted.

> I think it's going to effectively kill public chat communities without either proof of identity or attestation through a web of trust. This seems self evident to me too. It's another factor in why I think the tech community needs to get ahead of governments on the whole "prove your ID on the Internet" thing by having some sort of standard way to do it that doesn't necessarily involve madness in the loop.

> It's another factor in why I think the tech community needs to get ahead of governments on the whole "prove your ID on the Internet" thing by having some sort of standard way to do it that doesn't necessarily involve madness in the loop. The problem here is that the premise is the error. "Prove your ID" is the thing to be prevented . It's the privacy invasion. What people actually want are a disjoint set of only ma…

Exactly, "ID" is a solution masquerading as a requirement, the real requirements are far more granular, and the more we can narrow it down then the better our chances are for a solution that isn't evil/abusable.

To recycle parts of an old comment [0]:

> If I had my 'druthers, there would be a kind of physical vending machine installed at local city hall or whatever, which leverages physical controls and (dis-)economies of scale.

> The trusted machine would test your ID (or sometimes accept cash) and dispense single-use tokens to help prove stuff. For example, to prove (A) you are a Real Human, or (B) Real and Over Age X, or (C) you Donated $Y On Some Charity To Show Skin In The Game.

> [...] The black-market in resold tokens would be impaired (not wholly prevented, that's impossible) by factors like: [...] scaling the physical portion of the work [...and...] There's no way to test if a token has already been used, except to spend it.

[0] https://news.ycombinator.com/item?id=45523550

Re: AI slop is killing online communities

#713

Post by 81AB24FB »

Seattle3503 wrote: post #569
CrzyLngPwd wrote: post #150

Earlier quoted context omitted.

It was a surprise to us how vehemently some folk defended AI content and assumed it was their right to post it within our community. We had no problems with people using it and posting elsewhere, it was the demands that we must allow it that were problematic and made us question whether we were doing the right thing. No regrets now, though, as we see competitors being flooded with AI slop and they are too invested in…

I have a similar problem in a community I'm a a part of? How are you reliably detecting AI?

It's not about perfectly identifying AI content. There's a relevant XKCD: https://xkcd.com/810/

When posts fall within "acceptable" then it does not actually matter where it comes from. Logorrhea, massively offtopic, and/or shitposting are bad when humans do it. Those should suffer the same fate.

Historically it was tolerable, but has become the highest priority today because machines have cranked up the volume. If we mis-identify human garbage as robot nonsense it does not matter.

Re: AI slop is killing online communities

#714

Post by erpellan »

intended wrote: post #570
erpellan wrote: post #295

Invert the economics. Right now there is value in posting LLM generated content that is more than the cost of using the model. If platforms had a subscription model that you had to pay for in order to do more than just read comments, there’d be a lot less LLM content. There would also be a lot less of all content. But maybe that’s the price you pay (literally) to get rid of AI slop.

Good idea, but models are only going to get cheaper. Unless governments add an information environment pollution tax. Oh hey, now thats an idea.

I meant the social media platforms. If HN charged a fee for posting and commenting it doesn’t matter how cheap the AI models are, it’d be punishingly expensive to sling slop.

Re: AI slop is killing online communities

#715

Post by 81AB24FB »

intended wrote: post #545
layer8 wrote: post #186

Earlier quoted context omitted.

If it can no longer be distinguished from real, why would it make people leave?

Underrated take. Yeah, the “blast radius” for social media AI slop is 80%-99% of humanity. There’s many times even I cannot make out if something is slop. Hell, AI slop is going to be even better than reality for a portion of humanity, so it’s More likely they will stay online.

A sufficient number of people are satisfied with human slop. That the new slop comes from LLMs makes no difference.

In the same direction, if you can't cell the AI slop from the low-quality human output, does the source matter?

The problem is quantity. Has been for a while, the robots are only making it more obvious.

Re: AI slop is killing online communities

#716

Post by armchairhacker »

joe_mamba wrote: post #695
armchairhacker wrote: post #651

Earlier quoted context omitted.

Ban them. I also recommend making it explicit: your last rule should be “moderators have final discretion”. The hard part is figuring out when to apply rules leniently or strictly. For example, I think users should always be warned, then given shorter bans, before a long or perma ban (unless they break an obvious egregious rule, or they’re a suspected alt of a banned user). At the same time, you do need to kick out p…

>Ban them. I also recommend making it explicit: your last rule should be “moderators have final discretion”.[...] >[...] Many communities (especially on Reddit) have become echo chambers mainly because the mods are too happy to ban people for unspoken and/or vague rules To me, it seems you're contradicting yourself here. In the first paragraph you suggest the solution to vague rules (mods having full authority), but…

The point is to have discretion but use it rarely. Reddit echo chambers use it way too often.

It’s not contradictory but unfortunately impossible to fully explain, otherwise it would be explicit rules. When someone acts borderline, you must intuit when they’re acting in good or bad faith. Hence I think a first warning, then short ban, is necessary for anyone not obviously breaking a rule; because the former will be careful, but the latter will keep acting borderline again and again, avoiding exactly what you warned about while still causing malice.

Re: AI slop is killing online communities

#717

Post by 81AB24FB »

colechristensen wrote: post #225
mrhottakes wrote: post #133

Earlier quoted context omitted.

What do you think happens to the least prolific organisms that lose the survival of the fittest?

To be clear I'm talking metaphorical survival of the fittest that takes the form of prestige/popularity/status/etc. of people and organizations.

In the same way that monasteries preserved ancient greek philosophy during the middle ages, hopefully some groups will maintain our modern culture through these dark ages.

Re: AI slop is killing online communities

#718

Post by 81AB24FB »

michaelt wrote: post #378
pjc50 wrote: post #234

Earlier quoted context omitted.

The IRL social network is actually the important part of the trust structure. The only one of these I've seen that really worked was the Debian developer version: you had to meet another Debian developer IRL, prove your identity, and only then could you get the key signed and join the club.

> The IRL social network is actually the important part of the trust structure. For Debian-style applications that are 100% about openness and 0% about secrecy, sure. But if you want to secure communications between pro-democracy activists in China, or you're a Snowden-like whistleblower wanting to securely communicate with journalists - y'all probably don't want to be vouching for one another's keys.

I participate in developing anti-censorship tools. Chinese are a significant population, and it has some overlap with activists there. In practice, identity settles at "who controls this email address".

Self-signed PGP is very occasionally used to prove continuity across channels or addresses. Cross-signed basically never.

Re: AI slop is killing online communities

#719

Post by chongli »

senordevnyc wrote: post #700
chongli wrote: post #495

Earlier quoted context omitted.

It's a market for lemons [1]. The issue is that if AI slop can't be readily distinguished from real human content, the real human stuff will get less and less attention over time. With less attention, people lose interest in writing, and eventually abandon the community altogether. As genuine human writers leave the community, the concentration of AI slop increases, and readers begin to realize that there isn't anyth…

As genuine human writers leave the community, the concentration of AI slop increases, and readers begin to realize that there isn't anything of value left to read, so they depart as well. How will they realize if AI content is indistinguishable?

Because it’s not 100% indistinguishable. It’s practically indistinguishable for the purposes of moderation (or at least moderation at scale), but recognizable because it provides no value to the reader. It’s empty calories for readers in search of content with real substance.

Re: AI slop is killing online communities

#720

Post by 81AB24FB »

mschuster91 wrote: post #568
thin_carapace wrote: post #439

Earlier quoted context omitted.

wouldnt it be more productive to talk about the systemic framework leading to this inflamed state of affairs, and ways that we can tackle the issue on the ground level? perhaps inhabitants of the west would prefer pseudo masculinity to another few decades of migrant influx without corresponding upgrades to social infrastructure. this sort of internal struggle would provide a ripe substrate for foreign agents to perfo…

> in other words, extremists are violent regardless as to their political persuasions No matter where you look, the far-right kills and maims substantially much more people than the far-left does.

That may be true where you live, but some of us have to deal with 打着红旗反红旗.
Post reply on HN