外刊精讀 | 第一篇｜原文

If bosses fail to check AI’s onward march, their own jobs will soon be written out of the script

Machines have already taken over the drudge work. Now they’re coming for the fun stuff. This may focus minds.

If there’s one thing Hollywood screenwriters know how to deliver, it’s a snappy one-liner.

“Pay your writers, or we’ll spoil Succession,” read one of the?placards?paraded outside movie studios in Los Angeles this week, as thousands of film and television writers went on strike. “Pencils down, middle fingers up,” said another. Closer to the bone, however, was a placard reading: “Wrote ChatGPT This.” For the plot twist is that this strike isn’t just over money. The Writers Guild of America also wants to establish some ground rules preventing studios from using artificial intelligence to generate scripts in ways that cut humans out of their own creative process.

The union has been understandably spooked by the rapid progress of ChatGPT-4, the chatbot capable of generating uncannily convincing knock-offs of any written genre, from rap lyrics to?Jane Austen. What it produces is hollow pastiche rather than art, piggybacking shamelessly on centuries of human endeavour (it learns by scanning samples of existing writing). But how long before it’s capable of generating a mediocre but acceptable TV sitcom, or the umpteenth movie in the Fast & Furious franchise? After all, studios already use algorithms to analyse box office data and predict which combinations of actors or storylines will seemingly get bums on cinema seats. The logical next step is to make the software write its winning formula up into a?screenplay, maybe hiring a human to give it one final polish.

If this is one of the first AI-related strikes, it won’t be the last, and in future they may be much, much angrier. Almost half of Britons think a machine will probably be able to do their job better than them within a decade, according to new research for Jimmy’s Jobs, a podcast on the future of work set up by the former No 10 adviser?Jimmy McLoughlin, and 63% felt government should intervene in this process somehow. McLoughlin identifies finance, media, advertising and education jobs as particularly vulnerable to disruption, though the technology is evolving so fast that its effects are hard to predict. This week?IBM?sent shivers down white-collar spines by announcing plans to freeze recruitment in back office roles such as HR, on the grounds that many of these jobs may soon be automated.

But as a separate report from the centre-right thinktank?Onward?this week warned, what differentiates this wave of automation from previous ones in human history is its ability to take on creative, cognitive tasks, from writing to photography and graphic design. Once upon a time, humans could be persuaded that getting machines to do the drudge work would free them up for more interesting tasks. And for the lucky ones, sometimes, that was true. But AI is coming now for the dream jobs: well-paid, absorbing work done by people who love what they do and won’t let go easily. It’s coming not just for our ability to pay the rent, but for the things that make us happy.

Imagine a world, Onward suggests, where it’s possible effortlessly to churn out an unlimited number of Tom Cruise movies or Taylor Swift tracks every year. (AI can already copy voices with spookily convincing accuracy, helpfully for fraudsters now employing it in increasingly sophisticated scams, and could easily be trained on an artist’s back catalogue to produce songs that sound recognisably “them”). Good news for Taylor Swift, maybe, but would new talent ever get a break?

And the argument doesn’t stop there. If movie studios use AI to create storylines, why couldn’t publishers use it to sift manuscripts and even to draft them, especially at the more formulaic end of the book market? True, they’d miss out on ground-breaking new writers who might have caught a human editor’s imagination. But there might be fewer of the expensive flops that inevitably come with taking creative risks, too.

The net result could be a more lucrative industry – at least for the limited number of remaining humans in it – but a horribly stale, bland, homogeneous culture based on endlessly rehashing last year’s mass-market hits rather than discovering something new, plus the socially explosive prospect of a generation who have already made it pulling up the ladder behind them. It’s older workers who often struggle to adapt to rapid technological change, but this revolution could be tough on young people too, if the first casualties are the entry-level roles in which they once got their breaks.

Too apocalyptic? Maybe. AI will certainly create plenty of new jobs, even whole new industries, and it isn’t going to gobble up everything we know. Jobs requiring empathy, emotional intelligence or relationships of trust – such as nursing, classroom teaching or caring for the elderly – may prove more AI-proof than most, though perhaps only if we’re willing in future to pay more for properly human public services instead of replacing them with chatbots. (Talking of funding public services, Onward suggests the Treasury should move away from taxing labour towards taxing capital, which might sound like a win for the left if it weren’t driven by concerns that in future there may not be as much labour to tax.)

Too often, warnings like this are greeted with a fatalistic shrug, as if there’s nothing humanity could do about our own invention. There is now active political debate over regulation – whether the tech industry should in future be allowed to create God-like intelligence it doesn’t understand or control – but far less about the ways in which existing AI is already disrupting jobs and lives. Yet there are huge moral choices to be made here and they can’t be left to the market or to the consciences of CEOs.

Years ago, I sat through a Conservative party conference fringe meeting on the tech industry, which has stuck in my mind because of the unanswerable question posed by a middle-aged man in the audience. He ran a mid-sized company, and reckoned that in the near future he might be able to replace hundreds of his staff with an emerging technological process. What he was asking was whether, morally, he should. Instinctively the idea of firing loyal workers troubled him, but if his competitors all cut their costs by using this technology and he didn’t, he might go bust and the jobs would be lost anyway. Nobody on the panel had a good answer to offer him, but his question feels even more urgent today.

For now, it’s writers out on the street waving placards. But as one of those placards pointed out, the logical next step after eliminating them is to automate away studio executives’ jobs, too. Do employers really want to live in the world they may be about to create?