I’m not sure if the baristas are just doing an excellent job (as usual) or we haven’t been targeted yet, but its looking like keeping spam out of digital communities is getting more and more difficult.
Anyway, thank-you for your efforts in moderating and such
This is one of the main reasons why I’ve been so thoroughly disinterested in socializing on the Internet lately. I also don’t feel comfortable socializing in real life for various reasons, so I’ve just been waiting around hoping that the Internet will adapt somehow. I’m gonna guess that CAPTCHAs have been largely defeated at this point since I don’t see anyone mentioning them in the article or the comments.
I’m shocked that some people in the comments are suggesting that it would be okay to let some chatbots into forums as long as they behave. I really don’t see the point of letting chatbots participate in communities with people even if their posts are pretty much indistinguishable from human posts. It’s like saying that AI art should be accepted as long as it looks just as good as something a person would make.
One of the comments says “It’s occurred to me that we need to move to a default-private, curated-public model to change the incentive model,” which I totally agree with on general principle. I think removing as much incentive for this behavior as possible (no monetization, no algorithms, no like/follow counts, require manual approval for posts, etc.) along with some kind of utterly brutal proof-of-work system could curb the problem to an acceptable level.
It seems like the proof-of-work stuff people have been doing to deter scrapers has been at least somewhat effective and I’ve been wondering if severely throttling upload speeds could be an effective AI content gatekeeping system. Imagine if the upload speed for text data was throttled to just a few bytes per second; it would take over a minute to post something even the length of a tweet and over 10 minutes for longer posts while also requiring the user to stay connected to the site the entire time. It would be a huge pain for spam bots, but real people could just write the next post while they wait. I feel like that could work on any site that accepts user uploads and it’s something that wouldn’t have to constantly scale and adapt to the increasing power of computers or waste energy doing meaningless calculations.
the piece brings up a lot of important points and pitfalls, but i think it comes off as pretty resoundingly defeatist. there are many more methods that communities can use to fend off this sort of stuff, and those methods usually also come with the added benefit of filtering out human poseurs. some commenters brought up the two big ones:
$50/yr is definitely too high lmao, but if there was even a one-time $5 membership activation fee in the vein of something awful, i feel like that alone would solve the problem completely. leather / kink seems to me to be a pretty expensive hobby and i think their community could tank that without much incident, and it would immediately raise the cost of the kind of spam the author is talking about high enough that it simply would not be worth it. again, this policy comes with added benefits: it raises money to maintain the community’s digital infrastructure, and it also heavily discourages human spam, trolling, and brigading.
aside from that, there’s this:
i remember back in the day that enforced lurking used to be pretty common once you initially joined a forum - posting rights were granted to you gradually over time as you demonstrated that you weren’t cramping the vibe and that you could be trusted to meaningfully contribute. the commenter thinks this is bad - i think it was great, and i think incidents like the one the author is talking about make it clear that we need to go back to that paradigm.
in any case, incidents like this are yet more proof that the dark forest theory is correct - the future of the internet is default-private, curated-public, as @Spots put it. i don’t see any other path forward that doesn’t leave the door open for the sort of issues the author is talking about.
I have often been falsely flagged as a spambot in multiple sites. I don’t know if it’s my username being a chip or if it’s that my writing sometimes feels a bit like an LLM when I write long texts even though I don’t use them at all…likely the later.
I’ve been flagged as spam before too… but I know it’s because wordpress doesn’t like my email address because if I post the exact same post ten seconds later using my gmail address it goes through fine.
The weird part is, I have a wordpress account, which I am logged into while posting, using that email address.
I know that a lot of places just blanket ban emails which are from any domain that isn’t Gmail, Microsoft, Gov etc.
The explanation is that just so much spam comes from self run email servers that “it isn’t worth it”
We have a trust level system here that iirc functions relatively similarly. You can’t see some sections until you’ve hit a certain level of activity, some things aren’t visible without an account, etc. This is to maintain user safety as well as limit spam.
I agree also that the article feels defeatist. Spam has always existed, even robotic spam. Being able to answer those ‘screening’ questions is new, sure, but it’s not an unbeatable problem or something that you need to restructure your system to overcome. As long as there’s an active group of moderators, spam can be dealt with quietly and quickly. This is an example of ONE (!) group using mastodon to spread spam using LLMs. It’s going to get harder to combat spam in a reliable way but as long as you’re not relying entirely on automated systems to keep spam out i can’t imagine it becomes much of a problem?
Also, this person runs a mastodon instance, not a forum.
Yeah lol I thought that one was pretty ridiculous. The whole point of the post is that spambots don’t behave.
I think this is a good system for sites like this. Not just as protection against spam, but also as some measure of protection for potentially more sensitive topics or threads, as well.
