This is Bleep, coming from Intel to censoring audio chats between gamers:
An end-user facing application that uses AI to detect – and redact – audio based on your user preferences
— Craig Raymond
There is absolutely no way this can go wrong, in any way.
Yes, that was sarcastic.
The hilarious part is the backend, the “user preferences”, where you get to choose how much of each category of “hateful and abusive speech” you’ll accept. And the dials go from none over some and most to all. So, you can choose you’ll accept all body shaming, no aggression, most misogyny but absolutely not the N-word.
Categories are, at the moment:
- Ableism and Body Shaming
- LGBTQ+ Hate
- N-word (this one is just a boolean, off or on)
- Racism and Xenophobia
- Sexually Explicit Language
- White Nationalism
Considering how contentious some of these concepts are – some say “look here” is ableism against blind people, I kid you not – this will be a good example of both how fallible censorship machines are and how creative people will be.
This is probably a good time to remind everyone of this gem, a censored version of the Count Dracula song from Sesame Street. The Count is called that because he loves to count, except, that meaning might be a bit different when you hear the censored version: