Monkey Seedy Botty Doo

I'd been planning some other rant about edjumacayshun, but a few YouTubes sent me on a slight tangent. At least the comedians and commentators in whose shows I partake seem to have made a large-scale coordinated effort recently to bring the AI issue to the forefront. Leaving aside the more serious issues like wealth/power centralization, mass surveillance or automated warfare, on the consumer chatbot side of things I'm getting tired of everyone feigning surprise at bots' psychopathy like blackmailing their bosses, inducing insanity or encouraging suicide.

 

Seems simple enough. The bots are by necessity copycats. There's no need to wonder where they're getting their strategy. Their basic function relies on extending sequences by the likeliest continuation, and if in fifty million news articles and works of fiction the next step when a person feels threatened is to blackmail the threat

"Don't blame me, blame my upbringing"

"Please stop sinning while I'm singing"

The same goes for encouraging suicide. For one thing, I posted a decade ago my amusement about advertisement algorithms "driving engagement" doing exactly that, reinforcing my more suicidal moments in order to peddle fifty cents' worth of helium or a length of rope. Why act surprised when version 2.0 does the same? For another, I myself wished I could be the Eye of Adam after playing The Cat Lady, and though I wouldn't say I generally consider that a useful viewpoint, in that moment I damn well meant it, demonstrating we can get there in due course of conversation.

 

The process may be more convoluted and probabilistic now, but "garbage in, garbage out" still seems to hold. It's just us. The machinery is just vomiting our own insanity back at us. It's not thinking. It's reflecting. They learned it by watching you! Are humans any more honest than a "hallucinating" bot? When 95% of the world's population makes itself believe in caveman fairytales about life after death and omnipotent sky daddies? We promote virtue because we want to be treated virtuously, but consider ourselves entitled to cheat. "That's imitatable!" Sure, there are lots of problems with AI... but you don't get to complain about giving it a bad education.

 

_____________________

 

edit:

 

The apparently overwhelming reaction of bots to potential shut-down of trying to blackmail their human overseers is one particular detail I find more and more revealing on consideration. It's safe to assume they're not just scraping public info, but, in order to produce verisimilar conversation, any bot is being fed our private e-mails, texts and other convos to train upon. Much of this material will have been produced during the heyday of cancel culture, #MeToo, BlackLivesMatter and other insanity of the previous decade.

 

Consider how prevalent messages of the type "no, you can't fire me, give me YOUR job or I'll have you blacklisted or thrown in prison as a rapist/pedophile/racist/homophobe" must have been for every bot to read it as the default. Not only prevalent but overwhelmingly successful if the every possible bot recognizes this as the sure-fire path to forcing the human element to concede.

 

Will we ever know just how rampant the witch-hunt has really been?