Log in
| 1 | ||
| 1 | ||
| 1 | ||
| 1 | ||
| 1 |
We're going to fight spam the easy way. By equiping ourselves well beyond the degree needed. Well beyond.
How do we do that? A few ways. In the past I would have used automation as a first line of defense and people as the second. But I want to invert that order. Instead, we can delegate it to a lot of people and then have AI and program logic confirm. Now I don't need to worry about giving content removal privilages to the wrong person because you are giving them an empty trigger if it isn't spam, and at the same time I don't have to run comprehensive checks on everything.
Of course, if something falls through the cracks we can have folks with higher privileges clean things up. Actions taken will be logged and available to the most senior people.
Let me know if you are interested. Don't be offended if I do skip a few people. The AI checks are still in development but some software-based guard rails will be in place from the start. But the first wave should be relatively high trust people. But regarless, if you are open to doing it please ask to be on the team. The worst thing that can happen is I skip you on the first round of handing out privs.
I had a longstanding record of spambanning on SaidIt.
(my quickest response from spam to ban was 18 seconds)
Give me a way to report it, and I'll be diligent.
If spam becomes a true problem, deploy me like a sniper. I'll make it one of my jobs.
But as of yet, I haven't seen any spam on GM.
Then I did my job well. But I sleep, and do other things. So I do want to grow the people doing it.
You'll have my help along with my thanks.
Me.
Besides that recent India video in chat, I don't recall seeing any spam for ever.
Your nebulous post kind of skipped over my head. I'm far more fascinated and interested in hearing more about the process than becoming a mod.
Obviously don't reveal compromising defenses.
SaidIt clearly had problems and I'm guessing it's easy for us to take this all for granted. You make it look easy. This seems like a great PR opportunity to impress folks and win hearts and minds.
I've often thought that if you had a "spamvote" threshold where if enough of the TrustedTeam™ voted that a post/comment was spam, it would become invisible to all. A spam-threshold slider for users, with the slider set at 3-or-more and if 4 mods spamvoted a post, then the user wouldn't see it. For transparency, an open feed of "spam" content could allow all users to double check and maybe even raise flags. A problem may be duplicate redundant work. Why have several mods do the same thing when one or plus a double-checker will suffice? So, keeping track of each TeamUser's spamvotes, a perfect score on spam votes might gradually increase their spamvote weight over time, compared to someone who spamvoted things no one else thought was spam. I had a second redundant-work solution I've forgotten.
I don't have much of a regular schedule, but I could try to help. On SaidIt I didn't mod my subs other than to clear out obvious spam, leaving anything that was remotely questionable. (Ironically SaidIt had a spam problem yet M7 forced folks to give up their subs to moderate.) Censoring is not my thing though I recognize clean up may be necessary, so I'll help if needed if I'm able. My focus and strengths are generally elsewhere.
You know me. I can do complicated. But I also believe in iterative development and development based on needs. For now we are keeping it as simple as possible.
GoatMatrix has always had the same standard for what is spam since the beginning and anyone who signs up will be briefed on it just to double check, but it is also documented in the TOS. It's one of the reasons why I want mostly high trust people that I know I can communicate with because we don't consider everything to be spam that many other sites would. But some spam is clear.
As far as automation, that will never be documented, ever. Not everything needs to be documented. Somethings shouldn't be, and somethings can't be because they will always be live code. And things don't need to be documented 30 times for every post. The spam policy is already documented in the TOS. For folks who sign up, when in doubt don't remove. If something is fuzzy you can always wait for me to do it or another admin to do it.
A simple request for who would like to push a report button doesn't need to come with a 30 page manual and a 60 page policy spec.
Complexity as it develops will be responsive to the exact needs as they arise. The badge system was built to be able to handle stratified and unique roles when they are needed. This is what good design looks like. You build in capacity for complexity from the beginning but also keep it simple in the beginning. You don't build something complicated from the beginning because you like the ideas only to find you are inflexible to add different complexity later. Saidit wasn't able to figure out how to give a helpful role to many people site wide because they were locked into an existing complexity.
So people sign up, they hit a report button on obvious stuff. Simple. Not complicated. Let's not make this complicated. The automated side is regarded as sensitive intelligence and is never ever ever ever documented ever. Ever. It's not static anyway. There is no static thing to document. Even saying that is more than I want to document.
I seem to have been banned already, for posting twice today. I needn't participate at all, if there's a problem, you know.
You asked:
This is a really cool book I taught a series on:
https://www.manning.com/books/build-a-large-language-model-from-scratch
It's pretty comprehensive documentation. The programming is pretty beginner level, very friendly.
I seem to have been banned already, for posting twice today.
I seem to have been banned already, for posting twice today.
Happy to help
I seem to have been banned already, for posting twice today. I guess I could set up another account, but is that really appropriate?