|
|
|
|
|
|
by vineyardmike
461 days ago
|
|
|
> it likely applies to any content in less common languages, yet this issue rarely comes up in discussions about their moderation efforts It comes up constantly, just not by politicians and lobbyists with ulterior motives. It's certainly a huge reason they're investing in LLMs too. Politicians typically don't care about the reality and limitations of the policy they propose, and view moderation either as censorship or a jobs program. This is also a widely understood concept, especially today with the rise of TikTok. There is an entirely new vernacular of English forming in response to automated moderation efforts (eg. "un-alive" vs "killed"/"suicide"). Internet users generally understand and are capable of learning to evade moderation by changing language usage. This is also popular in high-moderation environments like China too, where there are plenty of subtle euphemisms. This was also one of the core issues in the "genocide caused by Facebook" in Myanmar. It was reported that during the relevant time periods, Facebook often had between 0 and 1 full time employees capable of understanding the local language and customs, but didn't want to invest in hiring moderators with knowledge of the language. |
|
|
Thanks for your useful comments, I hadn't heard much discussion of the issue but I'm pleased someone is talking about it.
Wrt LLMs filling the gap, given non-techs rather wide-eyed view of what they can, or will, do, I'm concerned that the discourse will become one of "don't worry we've got ai on the job".