CrazyMyraApr 17, 2024
Quite amazing how completely techbros can fuck up without really trying.
Take BlueSky's new labelling system, released late last month. It allows users to create labels for accounts and individual posts.
So you can warn your pals/subscribers about, for example, transphobic accounts or content.
Sounds good, right?
There's two little catches.
Anyone with a few tech skills can set up a labeller. So the transphobes can target trans people, too. And if you feel you've been mislabelled, you have to subscribe to the labelling account to even report it.
But the real lulu is that you can't block a labeller from seeing - and labelling - your account or your posts. A block merely means YOU can't see them on BlueSky.
Way to go to open a whole new harassment vector, dipshits.
Show threadCrazyMyraApr 17, 2024
Quite ridiculously, the devs are listening to a group of idiots who want to make it impossible to "hide" certain labelled content.
The racists, other assorted fash, misogynists, and antisemites are rubbing their hands in glee.
Show threadCrazyMyra
Oh, and you can't even see the labels on your account without going offsite to an app where the single sane person in this meshugas allows you to check.
Apr 17, 2024 at 9:50pmWeb
Show threadCrazyMyraApr 17, 2024
The details only get worse.
The BlueSky account representing a labeller does not have to show its follows or followers. It indicates only "likes". Those who subscribe to its service do not have to like anything to see all the labels.
(It's this account you can "block" - not the labeller itself.)
Show threadBenApr 18, 2024

@OutOnTheMoors I guess it just takes a while to see all those problems with first principle thinking ;)

Also weird though: isn’t this exactly the kind of task AI should be good at? Let people subscribe to different ones if they need to.

Show threadCrazyMyraApr 18, 2024
@b3n The site uses AI in determining its own labels (such as nudity and gore), and it works reasonably well to determine visual issues.
What's a whole lot less easy to determine are when it comes to context and behaviour. AI is pretty useless here.
Any subtlety of analysis runs the risk of an entire account being branded as supporting a cause. This is already happening wrt Palestine.