Reply to post: Re: Apples and oranges

Evil US web giants shield terrorists? Evil spies in net freedom crush plot?

Michael Wojcik Silver badge

Re: Apples and oranges

Saying that because the technology exists to identify images of child abuse and therefor we should be able to identify posts containing text that is considered as terrorism is nuts.

Particularly since reasonable human judges cannot agree on what constitutes "text that is considered [by whom?] [to be] terrorism". The SCOTUS justices can't agree on under what conditions Facebook posts might lose free-expression privileges (and their disagreements don't fall easily along party or ideological lines) - and they're the people we in the US have appointed to make precisely that determination.

Facebook can, of course, suppress whatever speech it likes on its site; Freedom of the Press applies to the owners of the press, not the people writing for it. So if Facebook wants to train some ML model to recognize "terrorism" or whatever, it's free to do so. Ultimately, though, any training mechanism, supervised or not, relies on some set of subjective evaluations by one or more humans as to what constitutes "terrorist" speech.

And, of course, what Facebook might do after identifying such speech is a rather different question. I'd don't believe anything useful is accomplished by suppressing such speech in public forums, except when it attacks individuals; and I have doubts about the ethics of automatically alerting police, though if the postings are public then the authors have no presumption of privacy.

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