Building an AI is expensive. It also means FB is on the hook for false classifications.
This moderator labor is (relatively) cheap. It also allows FB to point to a third party if something goes wrong.
Look at their statements in the article. It's easier for them to distance themselves and point to "bad actor" contractors than it is if these were direct FB employees.
From the descriptions of the work in this article it sounds like Facebook is actively choosing to have the same moderator re-moderate the same content over and over. At that point it almost seems like intentional malice from Facebook rather than a "developer oversight". Surely the first time a video has been flagged it should be trivial to identify further uploads of that video?
The only reasonable explanation I can imagine is that Facebook is doing everything they can to avoid having to implement a "Content ID" system like YouTube. Now why exactly they don't want to do that can only be speculated.
