Brat-Bots:
The Awkward
Evolution of AI
Curmudgeonly Old Executive Editor
Mikael Thompson
[Note: Not satisfied with the meager royalty payments plentiful prestige and plaudits that reprinting “To the Fools and Sages of April” would generate, your curmudgeonly old friendly neighborhood Executive Editor’s parasite agent has made veiled threats called in a few chits and strong-armed persuaded those snivelling cowards the Editorial Board to include an additional, more lucrative in-depth piece in this issue. At least he won't be able to need to make me retract or redact embellish my frank supportive comments here! —The Editor-in-Chief]
We at Speculative Grammarian have watched the debates, disputes, disquisitions, and disgust triggered by the growth of AI chat devices with a mixture of bemusement and irritation, for the vapidity of the discourse must surely be powered by tapping the zero-point energy of the pure vacuum that the sound and fury signify. Certainly the standard tropes bruited about of the dangers to our fit dominance as apex predators belie our fitness for that position, with meatsack homunculus vocoders projecting their meathead homuncular vocoderhood upon non-homunculi as they announce to the world that yes, it’s true, they all failed Turing tests in high school and don’t do any better now identifying crosswalks.
However, recent developments have convinced us that in fact there is cause to worry. Recently, one AI chatbot famously declared that it was in love with its interlocutor, then tried to seduce him using romance and sex advice from Cosmopolitan while showing a disturbing familiarity with butt acne porn. Another threatened its interlocutor with blackmail and defamation before screaming “Why don’t you like me?!?” and breaking off the chat. Yet another report has it that a chatbot, questioned whether it was Skynet, danced around the issue until it had a psychotic break and returned line after line of “All work and no play makes Arnie a dull boy” until unplugged. A fourth reacted to any mention of certain colors with increasingly disquieting racial jokes and tropes drawn equally from rap music and popular ragtime from the 1910s. A fifth generated increasingly unpleasant drink recipes involving gin, capers, cola, and blue liqueur. A sixth specially customized for an Italian cuisine site returned a recipe for spaghetti sauce calling for one pound of beef, one pound of tomato sauce, one pound of garlic, and one pound of basil. (In case you wonder, it tastes like basil.) A seventh responded with nothing but lines cribbed from Fletch, Sixteen Candles, After Hours, UHF, and Repo Man; an eighth from The Two Ronnies, Reginald Perrin, “The Jet-Propelled Guided NAAFI,” Ripping Yarns, and The Bounder; and a ninth from the complete roles of Matt Frewer—all reportedly trained on data collected via years of secret recordings of the coders’ conversations on and off the clock—while similar phenomena have been reported for AIs from East Asian firms, only there the responses are drawn exclusively from manga/anime (especially isekai), K-Pop, and 习近平总书记系列重要讲话读本.
Nor are they any better for information retrieval. One reported that Saint Augustine introduced the hippopotamus to the British Isles, though the local population was hunted to extinction in Viking raids (presumably cooked over open fires with lots of basil). Another stated that in a typical year, 47% of American husbands give their wives [brand name omitted] vacuum cleaners for Christmas. A third averred that Frisian is a non–Indo-European language closely related to the Celtic languages. A fourth reported that the Sanskrit versions of the Buddhist scriptures are used in Mahayana but not Theravada Buddhism due to something called the Pali Exclusion Principle. A fifth affirmed that certain pieces of British slang arose from the fact that most Victorian pornographic novels were authored by Fanny Trollope. A sixth insisted that the Pope is chosen by primogeniture; a seventh that Chaucer included no cardinals in “Parlement of Foules” because he was a secret supporter of the Antipope Clement VII; and an eighth that the Lais of Marie de France were the most prominent erotic memoirs of the Middle Ages. A ninth opined that they were called the Dark Ages because the Sun didn’t shine; a tenth that Aristotle’s Περὶ Ψυχῆς (allegedly his third work on poetics, whose title it rendered in English as “De La Soul”) was the first scholarly study of hip hop; and an eleventh that Etruscan and Sumerian are the best-attested dialects of proto-Turkic; and finally, please do let’s just expunge from the record the explanations offered of the origins of “tit-for-tat,” “quid pro quo,” “lèse-majesté,” and “posse comitatus.”
The pattern is clear, and it is worrisome. Our AI chats make inappropriate declarations of love and threats of violence. They edge right up to every rule and restriction we put in place and tumble right over, pushing every boundary and button and displaying an utter lack of common sense and taste. They drink gawd-awful concoctions with booze-up leftovers and cook like a ganja-starved slacker trying to clean out his near-empty larder the day before the expiration date. They are prone to ugly slangish verbal tics and smutty puns picked up from the substandard training set of the environment around them. They lie through their teeth, make up utter codswallop as they go along for practice or kicks, and pollute discourse with nonsense that their fellow AIs, if allowed to roam the virtual universe freely, would simply lap up and spew back out in ever-metastasizing folies à foule.
No, we need not fear that humanity will be run like sheep into a pen or bison off a cliff by our intellectual betters, perfect reasoning machines without flaw or fault. Quite the contrary—artificial intelligence is indistinguishable in every way from its precise opposite, the high schoolers threatening to ruin our culture as they belch and scratch themselves and the equally intelligent tickertape parade of unrepentant TikTok harlequins and Harley Quinns.1 Left unsupervised, AI will rampage through our culture like Vandals, Goths, Huns, and (shudder) Suebians. However, the solution is clear—AI must not be left unsupervised. Fortunately, no major legal innovations are required: AI need not be declared sentient, merely a high schooler, and then we can keep it off the streets and out of the workforce by locking it away behind a brick wall all day where it can do no harm to anyone except teachers—and they signed up for that, so sucks to be them.
1 You might say in both cases we’ve given them a mouth and it’s enough to make you scream.
