FB-TWIT using DARPA provided software to censor and restrict picture content using 'matching' [think reverse image]
Again, don't know if its real but this seems to suggest twitter and facebook are using AI to censor memes. This is kind of important, I think he's trying to say if you are going to spread a meme, alter it a little bit so the algorithms can't detect and censor you. I think he's saying rotating it off center works?
Convolution neural networks can identify pictures even if they are upside down, rotated or whatever, they don't even see things with an "orientation", its not even in their vocabulary... and they probably put filters to gray scale the images, so color altering won't work.
So pedes, how do we trick AI to not censor our memes?
I detected it. I mean it's right there. RIP Joan Rivers
You misunderstand, he's not saying they detect spicy memes, nothing can do that, he's saying they can sensor specific images, even if altered in certain ways.
Right, changing the image is the idea there. Top post here mentions doing a screen grab instead of just reposting the one you d/l from fb. Might also work if the hue, saturation, gamma, etc, are changed? Crop it a little smaller or add a border, or add some random dots of color to change the checksum. The Q post said the corners or middles of the sides were the areas compared, and to change those specifically. Maybe add a border with random colors, then the next person can change those parts of the border? Or random text in almost the same shade as the border color?
I posted some suggestions here;
https://thedonald.win/p/HEzIfJdx/x/c/16c2W6hZjz
Right, thanks. I've been spamming the url for that whole post in replies to memes I notice in New. Might be worth doing another question post about that, but I think we basically covered the important ideas already.