and why I’m not mad about it
There are some things that AI is very good at and there are some things that it is not. Moderating social media posts falls squarely in the latter category.
A friend recently posted a meme on Facebook, describing various things that many people think are in the Bible yet are not. Things like the Rapture, the Seven Deadly Sins, and a preference for a particular political party. I commented on the post and quoted a statement made by the US President about foreigners and added, “also not in the Bible.” I thought I was being clever and witty. Facebook banned me (for three days) for hate speech. AI sucks at understanding irony. Alas, most humans do as well.
Setting aside the absurdity that a direct quote of the president’s words will get you banned but the person who said it is still active on Facebook, I want to talk about the process.
Facebook uses AI to scan new posts for content that violates their “community standards” (a euphemism for “rules of conduct” — I’m not quite clear on why they feel the need for a euphemism here). These rules are designed to protect the company from litigation related to content posted by members. Presumably this works because Facebook is still here.
It’s possible that Facebook uses AI for this important function because humans would be too expensive. Facebook’s FY 2018 net income after expenses was $22.1 billion. I don’t know how many posts are scanned per hour, but FB could hire 20,000 humans at $100,000/year and still spend just $2 billion. Maybe 2.5 with benefits. But who am I kidding? They would classify them as independent contractors.
My point is simple: AI has no subtlety. Humans would at least have the capacity to see some context. That said, I get that this would be mind-numbing work that no self-respecting human would want to do. And that would rule out anyone capable of understanding context.
I support the effort to moderate the forum. Hate speech is bad. But it’s no fun being collateral damage. And, alas, I have no better solution to offer.