Coming over here, taking our jobs …
As so often, Orwell got there first.
In Nineteen Eighty-Four, Winston Smith’s girlfriend Julia has “some mechanical job on one of the novel-writing machines”. The Party churns out a whole slew of literature, music, drama and entertainment designed to keep the masses happy, including “sentimental songs which were composed entirely by mechanical means on a special kind of kaleidoscope known as a versificator.” (Full text is available at http://gutenberg.net.au/ebooks01/0100021.txt.)
The idea of machines creating art is not new, but it took Orwell to give it the wonderfully sneering label, ‘prolefeed’. The topic has been back in the news recently. AI technology is being used to put dead actors in modern movies, or finish symphonies by dead composers. There are AI services like Midjourney, which can create any art you like, and – of particular interest to me, as a writer – ChatGPT. This is a service which, in response to a basic instruction from you (“Write a publicity blurb for this widget”; “Apply for this job on my behalf”; “Write a detective story in the style of Jane Austen”) will within seconds produce some impeccable prose apparently doing exactly what you ask.
Does this mean that writers will soon be replaced by AI? Because it must be said that so far, for better or worse, that does seem to be the case. A good friend recently lost a major income stream as the website she was working for has decided to use ChatGPT instead. It makes no difference that the product is flat, insincere, dry and devoid of any human feeling. It’s also cheap. And it’s not just the bean counters who are turning to it. A recent article by Neil Clarke, editor of the online science fiction magazine Clarkesworld, says there has been a surge of AI-generated submissions to the magazine by authors desperate to get into print but not actually prepared to do anything so old fashioned as get good at writing first.
I decided to give ChatGTP a run for its money, and asked, “Is James Bond sexist?”
ChatGTP: “The James Bond character, as portrayed in the books and films, has been criticized for being sexist and misogynistic. Bond is often seen objectifying and disrespecting women, and treating them as little more than objects to be used for his own pleasure. This behavior is not acceptable and is not in line with contemporary values. It’s important to remember, however, that Bond is a fictional character and his views and actions do not reflect those of the creators or actors who have portrayed him.”
A very worthy and right-on answer, I thought, showing as much actual engagement with the question as a politician toeing the party line on the Today programme. Still, one thing ChatGPT has over politicians on the Today programme is an ability and willingness to learn. The more it is used, the more feedback it gets, the more polished it will become. That is how things work. I think it is an unfortunate fact that AI will come to take over more and more tasks currently performed by humans. Interesting times lie ahead.
And yet…
As mentioned above, a couple of years ago, an AI lab proudly reported that it had completed Beethoven’s 10th Symphony. They had taken the few sketches Beethoven left before his death, taught the AI all of Beethoven’s other works, and pressed the START button. And yes, what came out does sound a lot like Beethoven. Every now and then you hear a familiar snippet from one of his other symphonies emerge, subtly altered.
But here’s the thing. Ludwig – Beethoven 1.0, as we may have to learn to call him – never produced two symphonies that were alike. Therefore, precisely because you can recognise bits, the one thing ‘Beethoven’s 10th’ does not sound at all like is a symphony by Beethoven.
At the other end of the musical spectrum, musician Nick Cave was profoundly unimpressed when a fan sent him a sample of ChatGPT’s ‘song in the style of Nick Cave’. (“Nick Cave says ChatGPT’s AI attempt to write Nick Cave lyrics ‘sucks‘”) Says Cave, “Songs arise out of suffering, by which I mean they are predicated upon the complex, internal human struggle of creation and, well, as far as I know, algorithms don’t feel. Data doesn’t suffer. ChatGPT has no inner being. It has been nowhere. It has endured nothing. It has not had the audacity to reach beyond its limitations, and hence it doesn’t have the capacity for a shared transcendent experience, as it has no limitations from which to transcend.”
Put another way, ChatGPT will never write “The Mercy Seat“.
When fantasy author Juliet E McKenna was starting her career in the early 90s, she was told by an editor that her manuscript was a perfectly competent fantasy novel, “but there’s nothing to distinguish it from the six other perfectly competent fantasy novels that have crossed my desk this week”. It was then left up to her to make it different, which she did. Cue a successful, award-winning career, now in its third decade.
Original inspiration is one thing that ChatGPT or any other AI seems unable to provide. It can copy, yes. It can pastiche, and it can do it very well. It cannot originate, whereas most works of any kind of art are sparked in the creator’s mind by thinking, “What if…?” They then pursue the idea, often developing as they go. My debut novel “His Majesty’s Starship” was ground zero for a whole host of influences swarming in from all over the field of science fiction: some of which inspired me, some of which I wanted to deliver my own take on, some of which actively irritated me and needed putting right. Stories so often come from an idea a writer wants to explore, or improve on; and on top of that, the human author then adds their own take on ethics, values, feelings… the things that can never be listed on a spreadsheet but are the residue of our entire human experience.
AI does none of that. It just copies. And any copy is always inferior to the original.
So, while AI continues to learn from human-created examples, it will at best produce perfectly competent work with nothing to distinguish it from any other perfectly competent work. Eliminate human input and it will only be able to learn off poorer and poorer AI-generated examples, spiralling ever further down into bland mediocre sludge that makes Orwell’s prolefeed look highbrow. To many outlets requiring text – like the website my friend worked for – that won’t matter. To anyone requiring something new and different – otherwise known as thinking readers – it will.
Phew, our jobs are safe, in the long term. That’s the good news. The bad news is there’s no way of knowing how long it will take for the bean counters to realise – or care – that humans do it better; and that that, in the long run, is what imparts lasting value.
But before Skynet starts sending robotic assassins back in time to silence me, maybe I should be nice about it. AI is not the game changer for creativity often claimed – but it is a game tweaker.
I can see how it could be used, say, to test the logic of a plot. To check the continuity, so that the brown-eyed brunette of Chapter 3 doesn’t become a blue-eyed blonde in Chapter 5 (we’ve all been there). To run your text through the Bechdel test or any other number of filters you might care to name.
In the Beethoven link above, one of the team says, “When it comes to the arts, I see AI not as a replacement, but as a tool – one that opens doors for artists to express themselves in new ways.” Well, quite. And meanwhile, when you come to one of us, you will get the services of a sympathetic, empathic, intelligent, informative, creative and 100% human brain. Accept no substitutes.