3 ms·
Seems like it could be interesting if the robot could learn from its mistakes, specifically that it would probably learn to be subversive (how to tell jokes tha
by throwaway894345 4y ago
Seems like it could be interesting if the robot could learn from its mistakes, specifically that it would probably learn to be subversive (how to tell jokes that bypass the censors' policies but still communicate the wrongthink) rather than learning to toe the party line.
- genewitch 4y agolast year "unalive" was used to bypass tiktok censorship, right now there are a ton of videos about "mascara" which is - and i already don't like this word - algospeak to talk about sexual experiences, apparently good or bad. Usually bad, though. I imagine a sufficiently advanced AI or even human could create a convoluted word or phrase and breadcrumb enough to walk it back to whatever would have been censored, a la "for want of a nail" style.