Readers, let’s try an experiment.
Say you are a sentient being (human, midforties, white, male) who writes a regular arts column for SA’s best daily news and business publication. Say the year is 2023, soon after the introduction of ChatGPT to a world that is wide-eyed with a combination of excitement and terror at this latest leap forward in artificial intelligence (AI).
Your human readers — cosmopolitan, well-informed, up to date — have probably already come across various examples of AI-generated artworks. They have given some thought to the ethical, aesthetic and philosophical questions raised.
How do we define creativity? How much do we value it? What makes the human mind distinct from a computer? (In which ways is it superior? In which ways inferior?) If AI is a form of collective human consciousness, what does this mean for individuals? Is human creativity as we have long understood it soon to become redundant?
Desperately hoping that the answer to that last question is “no”, you and your readers alike took some comfort from the early glitches: evidence, it appeared, that robots can approximate art, or imitate it, but they can’t quite get it.
Then the art-making machines started learning. They don’t produce as many bloopers these days.
The same thing has happened with robots writing. At first, they generated texts that were either generic and boring, or that just felt “off”. They couldn’t quite produce sci-fi film scripts in which the characters were realistic; short stories seemed formulaic; nonfiction was just dull. Then we had to admit that humans, too, place outlandish film characters in unbelievable settings; that most human writers work within genre conventions; and that every day, thousands of people craft press releases, newspaper reports and academic articles that are tedious in the extreme.
Moreover, the textbots kept on improving. ChatGPT’s learning curve has been especially steep. Expose AI to vast portions of the internet, and you get a new iteration of the infinite monkey theorem. Instead of the random results of monkeys at typewriters, we now get targeted, commissioned and bespoke pieces of text.
So, back to the hypothesis: say you are this hapless arts columnist. It occurs to you that, however many human readers you have, they are far outnumbered by your machine readers. Your columns may be behind a (very modest) paywall, but they are readily accessible to an autodidactic AI that is ever-hungry for more virtual text to spur its learning process.
In that case, how long would it be before a bot could write “like you” just as well, or perhaps even better, than you can? And when that day comes, who will be able to tell? Your readers? (That is, your human readers — to your machine readers, there is no point in asking this question).
If there’s little to no difference, however, and if AI takes seconds while you take hours ... well, you can see where this is going. Say you and your editors agree on a new arrangement; now you get paid per word of the instructions that you give the bot generating your column (but, to complete the utopian picture, you get paid much more per word). Writing AI prompts, we are assured by vocational prognosticators, will be a critical — perhaps even lucrative — job in the near future. The role requires a person who is part novelist, part engineer, part psychologist and part teacher.
Still, at the risk of appearing to be a Luddite: as interesting as it all sounds, please count me out of this future. ChatGPT and other AI writers could certainly have the effect of improving the overall quality of the countless texts in circulation. They could hardly do worse than most of the writing produced by humans, which does not have much to recommend it. But when the day comes that the best writerly feat a human can achieve is feeding a bot to write something better ... that no longer feels utopian to me.
Perhaps that day has come already? I’m not sure. Which is why, dear reader, I am interested to know whether you think this week’s column was written by Chris Thurman or by ChatGPT. We’ll run the experiment again next week.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.