The gloomy narrative that AI writes better than humans needs a swift, human redraft
As the slurry of AI slop deepens, we’re realising that – at bottom – we don’t want more vacant content, we want connection. Computers, it seems, can’t do it all.
First things first: the fear of AI writing replacing that of humans is a big, over-cooked, red herring. A new study published in the Journal of Judgement and Decision Making in the US is being picked up by the panicky press to suggest that humans can’t always tell the difference between human-penned and machine-made prose. Shock horror. Can you hear the novelists and the journalists quaking in their boots yet? I can – but I’m not sure they should.
Beyond the headlines, if I’m permitted to venture so far, the “Bot or not” study suggests some caveats. True, the people polled can’t always tell the difference between work by ChatGPT and a human author but think about it. Are you surprised? Not all human authors are very good. Even fewer are unanimously popular. The study’s main conclusion could as easily have been summed up with a sigh and the phrase “there’s no accounting for taste”.
There’s also a perverse logic at play in the panic. Fretting that AI can write better than (some) humans is like pitting a powerful calculator against mental arithmetic, a hammer against a soft human hand or a Range Rover against a sprinter. Praising the tool misses the point. Human endeavour – in sport, literature, art, mountain climbing, whatever – is only interesting because it’s done by weak, fleshy, imperfect, irrational, flighty and benighted people like you (and me, I suppose).
With the complete works of Shakespeare, Baudelaire and Chekhov at its robot-writing desk, it stands to reason that a trained machine could scrabble together the rudiments of a pretty good yarn. There are rules to follow, text to copy and ideas to ape, a Moby Dick to dive into and a Frankenstein to borrow from.
The study suggests that stories composed by computers lack nuance, complexity or ambiguity. They mush rather than meld the human world that they misunderstand. What people like about them isn’t the awesome power of their insight; it is – the study says – their simplicity. But since when are the most rewarding or interesting things always just the easiest to read? How can we solve or discuss the issues we face if everyone is skimming the un-fact-checked dummy’s guide version of reality? Truth, beauty, irony, anger, art, justice, war – not simple.
Which brings me swiftly onto journalism and the renewed importance of the human editor (I would say that, of course). Much has been made of the capacity of AI to aggregate news but let’s not forget who’s gathering it. Without a source, a person on the ground to bear witness and be present, there’s no search business for the tech titans to tantalise us with or sell their advertising against. Human judgement, an experienced reporter, genuine insights and fresh ideas are becoming more, not less vital. They are valuable as the reservoir of AI slop and slurry deepens. Publications and people who deserve trust are at a premium, something to cling to in a sea of uncertainty.
Pleasingly, the study also showed that people preferred things they were told were written by humans. Even the naïfs who didn’t spot a computer pretending to be human were rooting for their own species. In the end, we want connections, not “content”.
And last but not least, as cleverer, more amusing people than me have noted long before this little rant, computers can’t do it all. They’re not funny. They lack a sense of irony, timing, whimsy and the absurd. They don’t get it.
Like most herrings, red or otherwise, the idea that human experience, insight or creativity can be replaced by a mirthless mimic machine is best enjoyed with a pinch of salt.
Josh Fehnert is inferior to a machine by almost any measure. But he is – for now – Monocle’s editor. Subscribe today to support our independent, human-made journalism.
