CCTV on trains

[O L Leigh]

I think the key is that the AI is just used to notify traincrew, not to do anything directly, then a human can look and see if there's really an issue or not. But then there's scope creep...

But it needs to be able to reliably identify instances of trouble if it’s going to be worth the, er, trouble, otherwise there’s going to be no faith or value in the system. At present it looks like the people spending the most money on this are still struggling to make it work properly. That’s what I’m flagging up.

 

[jon0844]

I was reading a BBC news article this morning where an astronomer was complaining that she had been locked out of her Twitter account because the AI moderator had flagged a video she’d posted of a meteorite falling as “explicit”. The article went on to mention that AI moderation on social media had similarly flagged a photo of some grazing cows and of a combine harvester working a field as “inappropriate”. https://www.bbc.co.uk/news/technology-63626769

I’d like to think that AI could somehow manage to spot the early signs of trouble, but I don’t feel quite so confident.

I saw this and it was flagged as being a video of something of a sexual nature without permission of the person in the video (such as revenge porn) and what shocked me was that if it had been, Twitter saw the suitable punishment as a 12 hour suspension!