New research has shown just how bad AI is at dealing with online trolls.
Such systems struggle to automatically flag nudity and violence, don’t understand text well enough to shoot down fake news and aren’t effective at detecting abusive comments from trolls hiding behind their keyboards.
A group of researchers from Aalto University and the University of Padua found this out when they tested seven state-of-the-art models used to detect hate speech. All of them failed to recognize foul language when subtle changes were made, according to a paper [PDF] on arXiv.
Adversarial examples can be created automatically by using algorithms to misspell certain words, swap characters for numbers or add random spaces between words or attach innocuous words such as ‘love’ in sentences.
The models failed to pick up on adversarial examples and successfully evaded detection. These tricks wouldn’t fool humans, but machine learning models are easily blindsighted. They can’t readily adapt to new information beyond what’s been spoonfed to them during the training process.
(Score: 2) by jcross on Saturday September 01 2018, @02:31PM
In the case of hate speech and offensive content, it's not even easy for people. We can't seem to agree on solid definitions for what constitutes either one. I mean we don't even have agreement among members of the same culture let alone across cultural boundaries. Even in a targetted community like SN, we can't always agree on mods, although I think you're right that the system does about as well as can be expected. It boggles my mind that people expect AI to do better somehow, when we can't even define what "better" would look like, and the current state of art in training machines is akin to digital Skinnerism.