Nope. It's just not that easy. Why do you think the biggest companies on the planet struggle so mightily to reliably and un-controversially figure out what content to block.
Right, so wouldn't they be highly motivated to create models for detecting it automatically? Even if it cost millions of dollars to develop, they would both save money on human labor and boost PR by avoiding articles like this haranguing them about "torturous" working conditions.
The fact is that the state of the art for ML is simply not to the point where you could build a model for reliably identifying a category as broad and ambiguous as "violence".
I would hope HN users, even if they aren't directly familiar with ML, would appreciate that it isn't trivial.
I'm not suggesting that moderation is easy to automate. I'm saying Facebook is an advertising company that cares about content moderation exactly as far as it impacts their advertising business and no more.
They have no concern for the moderators or the content, they just want to solve this problem (as it relates to profit) as cheaply as possible. At this point in time they believe the best way to do this is by utilizing low-cost labor.
Facebook could easily devote more resources to caring for these moderators but they don't because doing so has no further positive impact on the bottom line.
The cheapest route to moderation is automation, and I would be stunned if there isn't a thick layer of it applied before this content reaches human moderators.
My point is that ML is not a magic wand and we're still in its early days. Facebook would love to have a model that accurately identifies offensive content, but that level of AI does not exist today, and no amount of money thrown at it will instantly advance the state of the art to that point.
Nope. It's just not that easy. Why do you think the biggest companies on the planet struggle so mightily to reliably and un-controversially figure out what content to block.