weev on Nostr: Used to if I got a comment from someone I didn’t recognize and it was gibberish, it ...

Used to if I got a comment from someone I didn’t recognize and it was gibberish, it meant it was spammer Markov chaining. Now in the age of LLMs if I get a nonsensical comment I know it is a real friend of mine. And if I get something in what would have been normal speech of standard structure but a decade ago, I can instantly block it as a bot. You must cultivate a circle of madmen engaging in frantic clanging, and compulsive hate speech, so that I know that everyone I am talking to is real. It is in this way that the Turing test has so rapidly shifted by the changing circumstances of humanity. In the beginning when LLMs just reached a reasonable state, all a program had to do to pass it was write better than a subcontinental. But now, to the more discerning eye, they must have that *je ne sais quoi* of the paranoid schizophrenic, bipolar, or at the very least, for the regrettably sane, you must communicate in meter. Because I haven’t seen a model that doesn’t choke when you tell it to output properly metered verse. In the war against the machine for our consciousness, one must be mad. Shove pills down your throat and let the passions take you, if you have no madness in your heart. For every utterance that is not poetry will be dismissed as mere automoton.