Poison AI surveillance by using flagged words for normal things
Posted by Substantial_Back_865@reddit | CrazyIdeas | View on Reddit | 4 comments
Like instead of using codewords for illegal things, use codewords that are known to get you flagged for dumb ass everyday things. Like instead of “is dinner ready?” you could say “are you done cooking the meth?”. Instead of can you go to the store to get eggs say “Did you get the 50lbs of ammomium nitrate?” Maybe don’t say stuff that could actually be interpreted as a threat, but your personal fed would probably stop even considering what you’re saying after a while.
HelicopterUpbeat5199@reddit
I've been doing this for 30 years just because I thought it was funny. Now you tell me I've been doing a public service? Nice!
"The package has been delivered." "We are green. Commence operation" "Got em from a friend of ours. They fell off a truck this morning."
mrkstr@reddit
Crazy. I'm in.
redbo@reddit
I wonder if the AIs could follow if we all just started sarcastically saying the opposite of what we mean. Oh, yeah, I’m just NOT going to get eggs. 🙄
MordaxTenebrae@reddit
I wonder if you can screw it up further by using the censored versions of slurs with an offensive intent, in place of the actual offensive word.
I.e. Trying to be racist to a black person by calling them "n-word". I don't mean the actual n-word, but literally calling them "en word" (phonetically) as in "Hey you en word". Or being homophobic by calling a gay or lesbian person "eff word" and "dee word" respectively, or anti-semitic by calling a Jewish person "kay word".