Reply to post:

Boffins build AI that can detect cyber-abuse – and if you don't believe us, YOU CAN *%**#* *&**%* #** OFF

SW10

There is a serious point here about how the neural network is trained.

If it’s a group of (say) white blokes tagging what is ‘normal’, ‘bullying’, ‘aggressive’ etc., they could well overlook or miscategorise things that are perceived as abusive (or banter) by others.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

SUBSCRIBE TO OUR WEEKLY TECH NEWSLETTER

Biting the hand that feeds IT © 1998–2019