Ban Banspeedrunners instantly using an AI.
Discuss if you think that's a bad idea.

AI makes mistakes. Human do, too, but they make less mistakes. And the ST is pretty fast at banning ban speedrunners. So, no support.

FWGcats wrote:

AI makes mistakes. Human do, too, but they make less mistakes. And the ST is pretty fast at banning ban speedrunners. So, no support.

Detect bad words in their username.

eisjo wrote:

Detect bad words in their username.

that's not ai

eisjo wrote:

Detect bad words in their username.

The username filter when creating an account already does that - and if something gets through, the account can always be reported.

Other than filtering inappropriate usernames (which is already done without AI), what would you suggest the AI actually does to detect these speedrunners? I feel as though it's quite easy for a human moderator already to see someone intentionally posting inappropriate content and getting a lot of reports and realise action needs to be taken.

8xa wrote:

Other than filtering inappropriate usernames (which is already done without AI), what would you suggest the AI actually does to detect these speedrunners? I feel as though it's quite easy for a human moderator already to see someone intentionally posting inappropriate content and getting a lot of reports and realise action needs to be taken.

Triggering the filter within one hour of joining can be a sign, if it is done more than thrice, its a serious red flag.

https://scratch.mit.edu/discuss/topic/795097/?page=3#post-8461266

Cheddargirl wrote:

If someone is going to suggest use of AI in moderation, they're going to have to bring forth a working example in action instead of just saying “add AI!”, otherwise I would have to close it on the basis of being vague and not a good point for discussion.

Also, based on that link above as well, I'm sure this would be overstepping how much they want AI to help moderate, so it might also be rejected. It's up to ST though

I agree that it's way too easy to say “let AI do it!” without specifying just how it would do it - or showing a working example on some other site. So I'll reject this at this time. If a more actionable suggestion or a working example appears - you can request that this be opened back up.

The Scratch-team can block the users, until the Scratch-team decided what happens with the user (nothing, warning, temporary ban, permanent ban, termination, ip ban).

Closing because literally rejected now

eisjo wrote:

The Scratch-team can block the users, until the Scratch-team decided what happens with the user (nothing, warning, temporary ban, permanent ban, termination, ip ban).

Yes, but how exactly is AI supposed to know who is or isn't a ban speedrunner? How would the AI know if a new user is trying to ban speedrun, or if the user is simply trying to use Scratch?

I makes mistakes, humans do but less, so it's unlikely to occur.

medians wrote:

eisjo wrote:

The Scratch-team can block the users, until the Scratch-team decided what happens with the user (nothing, warning, temporary ban, permanent ban, termination, ip ban).

Yes, but how exactly is AI supposed to know who is or isn't a ban speedrunner? How would the AI know if a new user is trying to ban speedrun, or if the user is simply trying to use Scratch?

Break many rules in the first day.

Paddle2See wrote:

I agree that it's way too easy to say “let AI do it!” without specifying just how it would do it - or showing a working example on some other site. So I'll reject this at this time. If a more actionable suggestion or a working example appears - you can request that this be opened back up.

Did you intend to leave the topic open? It's still open.

M1mikyu wrote:

Did you intend to leave the topic open? It's still open.

The owner probably asked to reopen it

everwinner64 wrote:

M1mikyu wrote:

Did you intend to leave the topic open? It's still open.

The owner probably asked to reopen it

Yes, they did, I saw this project when it was closed, so the owner must've requested a reopen

M1mikyu wrote:

Paddle2See wrote:

I agree that it's way too easy to say “let AI do it!” without specifying just how it would do it - or showing a working example on some other site. So I'll reject this at this time. If a more actionable suggestion or a working example appears - you can request that this be opened back up.

Did you intend to leave the topic open? It's still open.

This topic is reopened.

Paddle2See wrote:

I agree that it's way too easy to say “let AI do it!” without specifying just how it would do it - or showing a working example on some other site. So I'll reject this at this time. If a more actionable suggestion or a working example appears - you can request that this be opened back up.

As “a working example on some other site”, the Clyde platform has recently announced an AI-powered automod feature that purple servers can add to their existing library of human-powered moderation bots, and so far the AI automod seems to be performing good enough for most servers to not mind

“Make a sandwich with AI”
for both cases AI is extremely unnecessary, no support