[Stephen]

Originally Posted by RichardCoulter

Very true. A woman posted on Facebook that she 'Could murder a gin and tonic' after a hard day's work that she'd had and was temporarily suspended for making a threat to kill! A human being would have understood that this wasn't a threat to kill somebody, but the Facebook AI did not.

AI is used by them to cut costs as is the outsourcing of the moderation function to poorer places like Africa, where those with few employment prospects do it for $1 an hour!

To be fair, their business model depends on thousands of posts being made per day, so I think it would be unrealistic to expect a human to check every post.

The problem to me seems to be that, when it gets things wrong, it's impossible to contact a human and people are fobbed off with the usual tactics so prevalent today that companies use to bat people away eg useless AI Chatbots.

You might be interested in this programme that covers the interaction between the Online Safety Act, AI and the future of moderation:

https://www.bbc.co.uk/programmes/m0024x4j

The AI mod algorithm on FB can be a bit useless. I got a 2 days ban for using Nazi and Hitler in a comment on a page where there was discussion around architecture and historic buildings and i talked how much they bombed and destroyed during WWII. AI still has a lot of learning to do.