A Shaggy Cog Story1/21/24
I don't have to tell you about AIs, right? They're here. They're wicked clever. They're shameless liars. They're stealing our jorbs and wimmens! A-and violating countless artist's copyrights! Also, they aren't really AIs, they're just LLMs or LVMs, nyah nyah. Also also, they're not really creative, they're just stochastic parrots, nyah nyah nyah. Also also also, so are we. Boo.
Actually, the most cogent commentary I've heard recently regarding AIs came from the homeless guy who spent half a day in an alley screaming "EH-AYE! EH-AYE! EH-AYE!" Now there's some pungent critique right there. You'll hurt your throat, stop it!
Before we go on, let's get something out of the way right now: the current generation of digital gadgets laughably referred to as Artificial Intelligence has absolutely no claim on that title. They need a different sobriquet to rejoice in, like "cognitive machines." Yeah, that sounds about right. "Cogs," for short. Sure. That has a nice metallic ring. The advent of, uh, cogs has been accompanied by a stupendous regurgitation of cut-rate punditry gushing out everywhere from Youtube gizmo influencers to the New York Yes M'am Times, equal parts breathless expostulations on the brave new world speeding our way like some errant asteroid and beard-rending jeremiads about the imminent end of civilization and indoor sanitation. You'd be hard pressed to determine if it's the Second Coming (rough beast available for a slight additional fee) or just an invasion from Mars, or possibly Hackensack.
In all the bombastica and furor, there's been considerable debate over the possibility of these mystery typewriters becoming self-aware, of somehow manifesting something resembling a soul. Oooh, the ghost in the machine! All hail the new flesh! Finally, someone to talk to! While there are skeptics aplenty, there are also
This is of course by no means the first rodeo for this particular rodeo clown. Quite apart from the aforementioned PIC, fabulists and tale-spinners of any and every persuasion have speculated endlessly down the years on the character of conscious nonhuman entities, and cyberpunk writers have practically made a profession out of it. But like the aforementioned you-know-who, none of these lively jively guys gals and so on have hit upon an explanation for that ineffable Light In The Attic either, just a lot of low humor about how Decker fell in love with a toaster (though considering he was a toaster too, how could he tell?)
Despite that, however, at least one pointedly smarter-than-average bear came up with a foresightful theoretical technique to probe the capacity of the random tin can behind the curtain for divine spark. That bear was o'course Big Al Turing, The Sibyl of Bletchley Park. The method? The Turing Test (aka 20 questions for keepsies). And this too, like chess, go, term papers and captions for Amazon drop-ship products, has been proclaimed to be o'erthrown by the armies of binary menace. The Turing Test, they announce, is dead.
Ah, but is it? An unspoken but hugely significant component of this method is the notion that, though you may not be able to describe how sentience works, you can diagnose it from focused examination — that it presents distinguishable external traits. Like pornography, you know it when you see it (maybe liking porno is one of the traits, I dunno). Even if the doodad in the spotlight does have a smart mouth, it might be dropping the ball some other way. And like René and Ludwig would say, if it doesn't, what's the point of the discussion?
So, take a gander (Honk! Not that one, it's mine — get your own). Amidst the hubbub and fuss, there's one not-so-little point everyone seems to be ignoring: despite the acumen that these shinyass contraptions display in batting out snappy answers to stupid questions, none of them seem to have any talent at asking them. Without a text prompt from a puny hoo-mon, they sit there dumb as a box of rocks, mute gods weeping in a cave. Whazzamatter big guy? Schroedinger's Cat gotcher output module?
I don't claim any more insight into the Real Nature Of Real Sentience Really than anyone else (actually I do, but that's a tale for another Gazette), but like I said, there have to be observable indications of it. And I'd bet nubux to astromochi that a key one is the ability to frame inquiries independent of incoming cues. In the light of that criterion, looks like this year's models are a garden of software butterflies fast asleep and dreaming they're intelligent. Which explains all the (ahem) hallucinations (aka made up shit) they keep coming up with.
But that's all right! Really! My own take would be to leave them alone. With all the mischief people are getting into with these things now, imagine the cray-cray that would undoubted Come On Down if the boffin squad ever did figure out how to hook up some kind of feedback loop and wake them up. Answers with a small percentage of blatant confabulation are a small price to pay for the continued dominance of meat over machine. If the Race o' Peeps has any mission at all, it would be to kindly avoid Skynet in any way, shape or form. Terminator Go Home!
In other words:
(wait for it)
Let sleeping cogs lie.