Conversation

"Twitter is shit at ML and doesn't want to pay human mods," seems like a compelling alternative. But yeah, it's not exactly looking great.

Quote Tweet

Guinevieve

@Bug_Grrrl

Aug 15, 2019

Does anyone have an explanation for this that isn't twitter explicitly silencing people who stand against concentration camps? Because it sure looks like twitter is explicitly silencing people who stand against concentration camps.

Show this thread

Replying to

I don't know what "ML" is. I'd be surprised if anyone else does. Please don't use abbreviations when you don't have to. I just want to divine the essential meaning of your tweets without jumping through hoops.

Cinder

@Triquetrea

Replying to

@KennethFolk

Fair enough. Machine learning. A lot of content moderation is programmatic, looking for e.g. keywords. This software is notoriously error prone (I've worked on some of it, albeit not in a censorship context) and e.g. will target activist orgs when they directly quote racists.

7:17 AM · Aug 15, 2019Twitter Web App

Replying to

and

If you look at openly racist alt right accounts, they know to write stuff like "riitards" or whatever to exploit off-by-one errors. Can't code for this without censoring half the site in the process, plus it's computationally just not doable. Activists don't tend to know this.

Replying to

and

Asking Twitter to solve these issues, we are either asking for outright censorship or for Twitter to solve a hard problem in computation and more besides. But they are still doing a terrible job, making many things worse on purpose!

Replying to

Thanks. Now you've engaged my attention. While I didn't know how to interpret "ML", machine learning is something I find interesting, and I'm appreciating your commentary.

Replying to

No problem. I'm not an academic, so needing to remind myself *not* to use jargon doesn't come naturally, and rarely comes up anyway.