dorian@doriantaylor·Nov 1, 2017Is the president ḅüÿïṅġ Ṫẅïẗẗëṛ ḟöḷḷöẅëṛṡ?Quote TweetTim Karr (@TimKarr@mastodon.social)@TimKarr·Nov 1, 2017Almost half of Trump's Twitter followers appear to be fake http://newsweek.com/donald-trump-twitter-followers-fake-617873…32
dorian@doriantaylor·Nov 1, 2017Replying to @vgrhaven't you seen those spambots that obfuscate the phrase so they don't get caught by sweeps? was that joke too esoteric?22
Venkatesh Rao@vgr·Nov 1, 2017Replying to @doriantaylorInterestingly cf our humaniform ai convo, an image-recognition text processor would be less fooled than a symbol9c string processor1
Venkatesh Rao@vgr·Nov 1, 2017Replying to @vgr and @doriantaylorDue to visual similarity an image-based reader would put higher weight on the correct reading and then confirm by dictionary1
dorian@doriantaylor·Nov 1, 2017Replying to @vgrthat strategy is also orders of magnitude more processor-intensive11
dorian@doriantaylor·Nov 1, 2017Replying to @doriantaylor and @vgr…I remember >10 years ago a spam tactic that involved rendering text to an image and then spanning confetti shreds of it across GIF89 frames12
dorian@doriantaylor·Nov 1, 2017Replying to @doriantaylor and @vgr…because on an animated GIF you can take the frame delay down to nothing and set a disposal policy to "leave it" —confused OCR detectors.12
Daniel Lemire@lemire·Nov 1, 2017Replying to @doriantaylor and @vgrAre you guys speaking technobabble on purpose?22
Venkatesh Rao@vgrReplying to @lemire and @doriantaylorDorian technobabbles. I just regular babble7:53 PM · Nov 1, 2017