‘He’s more machine than man now’, complains Obi-Wan Kenobi of his notoriously fallen apprentice Darth Vader in Star Wars. The same thought crossed my mind last week in the wake of the worst betrayal I have suffered as an English tutor. Something is wrong when your favourite pupil uses AI to generate the two-line reference they had offered to write for you, making you sound even blander than feared. ‘Ottillie,’ – not her real name – I blurted, ‘how could you?’ The reply was endearing but terribly ominous: ‘I wanted it to be perfect.’
ChatGPT is already shockingly good, much better than most people admit. Get it to rewrite itself to sound less like a robot, remove anything preposterous, and you might be done. I know journalists writing AI articles and barristers writing AI submissions; it conquered business months ago, where, to be honest, everyone already sounded like that. And having spent the spring preparing children for their exams, I can confirm that only the seriously dull ones are still writing any of their homework themselves.
Rational friends tell me to give up writing because AI will write all the good books we want
Some of them ought to start reading the machine stuff over before they hand it in. I wasn’t persuaded by a summary of Casino Royale in which Bond gets up after having his balls flogged, drives Le Chiffre back to the casino for another hand and seduces Oddjob. But such obvious nonsense has been rarer this year as the algorithm has improved, and it’s terrifyingly easy to make the opposite error. I recently accused several entirely innocent teenagers of cheating, only to find that they really had thought of something so ridiculous it didn’t sound like human work. An especially charming example appeared in a coursework essay on Macbeth, in which Shakespeare warns us that treacherous characters like to ‘win us with honest trifles’, i.e. reel us in with small truths before deceiving us in larger concerns. My tutee really did write a paragraph praising the ‘trifle’ as a confectionary metaphor, expressing the complex and layered psychology of the play.
The young man responsible for this analysis was, of course, unusually bright, but the hushed-up fact in education at the moment is a stunning collapse in teenage sentience. I have loved all my students, but many have come across as stranded, alien souls, unable to express residual human thoughts. ‘Yeah, aargh, it’s good! Isn’t it?’ began one recent lesson. After minutes of consternation: ‘Sorry, aargh: it’s good to see you! That’s what you say, isn’t it?’ Yes, it was a common greeting, once upon a time.
To think is to connect ideas, hard to practise if you can ask a robot for your next one. Some people think this is the evolution of the mind, and that we should raise children as human-android hybrids from the start. Elon Musk has a school which turns lessons into video games, since these are ‘like crack’ for children. Perhaps a cyborg generation will save us in time, reared on individualised mini-tasks and expert in commanding virtual research slaves. In the meantime, too much crack sounds bad and my bet is that the first cohort of such kids will be suicidal.
The broader problem with removing youngsters from the internet is that the adult world is so full of nonsense. The message of the GCSE English courses – which need some work, to say the least – is that to be literate means primarily to be inhumanly pedantic and inane. I teach students to defeat the English language exam by making the highest absolute number of points in each answer, and by ruining their writing with words like ‘cerulean’ and a colon in every sentence. If they struggle with plot ideas, I have them memorise my short story about a donkey, which can be used to answer an astonishing variety of creative writing questions. Meanwhile, it takes a seriously expensive school to avoid choosing the shortest books on the English literature course each time.
The impact of AI on young minds is worrying for the same reason that outstanding machine writing has been elusive. Genius entails uniqueness as well as intelligence, and good writers are at their best when most unmistakable, connecting infinity to a particular time and place. Having already sorted young people into a set of inane subcultures, the internet will now ensure they converge to acquire the character of an average, semi-robotic child. Meanwhile in the adult world, we will enter a Valley of Dissociation: a long period in which AI writing is useful enough to be everywhere, but not good enough to be healthy to read.
Yet the company of the young always provides hope. I find them surreally self-aware about their cyborg selves and appropriately embarrassed by their unauthoritative prose. They remain curious and visibly delighted to have ideas. They have excellent natural taste: even my least able charges prefer good writing they can’t understand to accessible nonsense. My rugby boys have been fond of Keats’s ‘La Belle Dame sans Merci’, with its progression from ethereal shagging in the meads to heartbreak on the hill.
Rational friends tell me to give up writing because AI will write all the good books we want. Once AI can do that, we won’t be talking to each other. The timeline over which that happens depends on an astonishing question which our generation will live to see answered: the nature of the soul. Judging by the sheer malaise of the young in their AI-drafted world, I think some form of soul-like residue may prove to be a strangely robust illusion, and that the masses may demonstrate a little more craving for the spark of reality than feared. We are on the brink of a tide of nonsense, but also resistance.
Comments