sandspiel
sandspiel copied to clipboard
MaxBittker
An idea I had.
Some people have been able to delete posts, but most the time they use it to actually help. This gave me an idea: what if you choose someone, who can delete posts, so that if someone does something wrong, that person can delete the post- such as inappropriate pictures. They can delete things that others don’t want to see- if they use the power to just get people angry, then you can remove it. That way, it’s fair. Or we can just do something like a democracy, and people vote for who they want to be able to delete posts.
Sorta, except max “elects” them. They cannot ban people, but instead can delete posts.
i like this idea but it's very very hard to pick people who are trustworthy and wont' abuse the power.
it's also very hard work to moderate and it can be kind of traumatizing
so you would have to go on daily to do it yourself? if you have experience in machine learning could you try to make artificial mods?
i have enough experience in machine learning to know that this isn't a good idea. https://www.bbc.com/news/technology-49424381
how about supervised learning? as far as i know, youtubes system is just putting data in a black box but with supervised you can teach it what you want
youtube’s system is trained by human moderators. moderation has a lot of tricky decisions, machine learning might help with spammers but not with bullying and racism for instance
On Fri, Mar 12, 2021 at 3:48 PM yesyoucangivememoney.nohax.exe < @.***> wrote:
how about supervised learning? as far as i know, youtubes system is just putting data in a black box but with supervised you can teach it what you want
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/MaxBittker/sandspiel/issues/125#issuecomment-797676078, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABNEHGXJI5T2W25AB6KU5W3TDJHXJANCNFSM4X4YGFNQ .
What if you just make it generate a list of the most common words among flagged posts outside of a whitelist (of ordinary common words) so you (the owner) could then just approve or reject suggestions for word bans, and then auto-flag any posts containing verified innapropriate language from the word lists for review, auto-deleting posts from the same IP with less than 5 seconds between them and over 90% similarity to prevent spam containing profanity from causing a dDOS attack, @MaxBittker ?
