by Peadar O Guilin
Lightning strikes, and everything catches fire. “Eureka, eureka!” You’ve just sat in the bath and sent water spilling over the side. The apple above finally lets go, knocking ideas loose as it bounces from your wig and rolls off into the grass. The universe is changed in an instant.
But the Muses don’t need to work at speed. They may also share their gifts through the medium of slow, persistent revelation. Like when you have an image stuck in your brain, of children finding wolf cubs in the snow, and you know there’s something there, something beautiful and special just waiting for the right circumstances to bloom. Dig here, the image says. Just keep digging and you’ll find it.
But barring belief in actual Muses, where do these ideas come from? Is inspiration a purely human experience?
It would appear so, because AI writing sucks. That’s the consensus. It sucks hard.
This doesn’t stop the curious from trying to squeeze brilliance out of Microsoft’s new Bing search. Nor does it prevent imbeciles from using ChatGPT to generate mountains of toilet-worthy crap that they then submit to actual editors. Even as I write this — yes, it *is* me writing this and not a piece of silicone — a major SF market, Clarkesworld Magazine, has closed itself to submissions until it can figure out a way to filter out the digital turds clogging up its u-bend.
What I’m hearing from the online furore this has caused, is that while the AI is well able to combine the work of real-live sweating authors into recognizable plots, not being human, our poor robot friend can never find the magic little twist of inspiration that launches great writing into the stratosphere. It’s like we finally have the famous room full of monkeys and typewriters, but no amount of time will get them to the level of a Dan Brown, let alone a Shakespeare.
I’ve seen some of these stories and my own feeling is that the consensus is right, but that this situation may only be temporary. Even as we humans waste our time eating, sleeping and keeping clean, more powerful monkeys are being bred in underground laboratories with ever more sophisticated typewriters, so that by some not so distant Christmas, you’ll be able to order up a new Dr. Tachyon story where you, yes *you*, star as the villain. Maybe Dr. Tachyon’s creator, Melinda Snodgrass, will get a royalty when that happens, and maybe not.
But we’re not there yet. I’ll get to some of the reasons for that in a little while, but first let’s look at some of the problems with AI “creativity” I’ve seen so far.
A few days ago, I finally managed to log in to ChatGPT and I tested its knowledge on several Wild Cards topics. It did pretty well, churning out answers that appeared convincing until I looked more closely.
What, I asked it, are the main differences between the Wild Cards universe and the Marvel universe? The AI got a lot of the basics right — the viral origin of the superpowers; the alternate timeline that began with Jetboy; the ability level of some of the characters and the fact that our lot don’t always have to be overpowered. Well done, ChatGPT! I’d be handing you a gold star for all of that, if you didn’t so spectacularly mess up on the *one* thing that makes Wild Cards as a shared universe, so much better than anything else.
Let me quote the relevant part of its answer:
“Interconnectivity: The Marvel universe is known for its shared continuity, with characters frequently crossing over and teaming up in various comics and media. The Wild Cards universe also features a shared continuity, but it is less interconnected and more focused on individual storylines and characters.”
As somebody who’s had the privilege, and let’s face it, the absolute *joy* of writing for Wild Cards, cooperation between authors and characters is *the* most fundamental of all the rules. You must bring in and refer to the creations of other writers and you are obliged to respect the integrity of those characters and all the lore associated with them.
It’s not the only thing that must be respected. The timeline too, is sacrosanct. Heroes age. Immortal characters are a near impossible sell to the editors. Reboots are forbidden. Unlike some other shared worlds out there, our readers can be sure that when they sit down with Khan or Bloat, that the rug will never be whisked out from under them. They’re standing on a foundation so solid it reaches all the way down to the bedrock.
In another question, I asked for a list of twenty new superpowers. As before, the answer seemed plausible enough. Yet, it soon became clear that pretty much everything on the list had already appeared in Wild Cards. The powers were “new” insofar as the AI eschewed the more obvious things like super strength, lightning speed and the ability to fly. But still. Number six on the list was basically, Hiram Worchester’s power of gravity manipulation; number fifteen was a far less original version of what The Amazing Bubbles could do and so on.
I don’t fault ChatGPT for this. Lord knows, when submitting my own new ideas for ace powers to George or Melinda, the refrain of “we have one of these already” was pretty constant. It’s no easy thing to come up with the unexpected and it comes as something of a relief to realize that a megamind with billions of pages of content to draw upon, can’t do any better than I can!
Again and again, when you look at the output of ChatGPT or other AI systems, the answers are plausible, but just not very good or completely correct.
Famously, $100bn was wiped off the Google share price when their own ChatGPT rival, known as “Bard”, presented a great sounding, but factually incorrect answer to the question, “What new discoveries from the James Webb space telescope (JWST) can I tell my nine-year old about?” Bard claimed that the JWST had produced the first picture of an exo-planet. Nope. Nay. And no, sir. It did not! But had *I* been the questioner, my ignorance of recent scientific discoveries would have allowed it to mislead me. In the same way, some poor reader posing the same questions to ChatGPT that I asked earlier, could end up thinking, “Oh, Wild Cards isn’t for me then, because I do love a good team up…”
You might insist that when it comes to story-telling, however, that small inaccuracies such as these wouldn’t be an issue. I mean, the whole point of fiction is to, you know, make stuff up. Even fake stuff!
That’s true. And on the surface, there are a lot of similarities between how a Wild Cards author and an AI operate.
As when working with an AI, the development of new WC stories often begins with a prompt, though in this case, it is the editors who tell us authors what they want. “We’re making a book about jokers on the Moon! Submit your ideas if you’d like to be included.” Writers scurry to their burrows to plot. Like ChatGPT, they respond to the initial question by trawling through everything they have experienced up to this point in time. But unlike the current generation of AIs, a good human author still has some capacity to recognise mediocrity and to steer away from it. The WC editors certainly know crap when they see it and aren’t always diplomatic about saying so.
And this is where we finally start to see the reasons why AI has so far failed to be a good story-teller. The clue lies with the editors…
Listen, once upon a time, shortly after reading my first Wild Cards book — *the* first Wild Cards book — I was a student at a university in Ireland and a very enthusiastic member of its amateur drama society. There was a guy there, known to be a good director, and I happened to see him rehearsing his cast one day, and I thought to myself, “He doesn’t seem to know what he’s doing!”
The director kept making the actors repeat the same scene, and when they asked, why? What are we doing wrong? Well, this lad didn’t have an answer for them. Still, he kept them at it, until suddenly, when they were all tired and exasperated, he said, “That’s it! Yes, I love that!”
Theatrical scenes, like any form of art, are only as good as the effect they have on a human being. That is the entire purpose of art, to affect us, to change or influence us. The main superpower of an editor, like a great director, is recognition. Ordinary readers have this power too, though it usually resides at a more subconscious level: we might feel tears in the corners of our eyes, or laughter may rock our bodies. Or maybe we find ourselves at work thinking about the book we’d read the day before, unable to let it go. Our emotions and our intellect respond to the art.
AIs possess a full list of all the works ever written that affected any human being, but when an AI encounters something brand new, the only data it has on the quality of what it’s seeing is a measure of how closely this new thing resembles whatever went before.
“Oh,” it thinks. “Yet another group of adventurers going on a quest? This must be as good as Lord of the Rings.” Or, “Wow! An immortal superhero? It’s so original. Like the Sleeper all over again.” Um, no.
Being the same as something that was original, is not, sadly, the definition of creativity.
Leonard Cohen explained the value of originality in a perfect little three-line poem. I’m probably not allowed to quote it here, but I can paraphrase it. He talks about writing a song for God. But why bother when God already owns everything that has ever existed? Well, yes, He does, but not this song. Not yet. You can’t just buy your deity a gift — He owns the whole store. You have to make it.
Yet, even LC couldn’t magic something completely out of thin air. Innovation, like atom bombs, is about smashing things together — things that already exist! — until they generate far more energy than we put into them.
When a new author writes for Wild Cards they bring their own obsessions to the table. These are thoroughly mixed with pre-existing WC lore and… Shazam! Something new.
I got my invite to submit in the middle of a visit to Australia to promote my novel, The Call, a book heavily dependent on the more terrifying parts of Irish Mythology. I pitched a dozen ideas for characters and superpowers to the editors, but unsurprisingly, the one that resonated, the only one accepted at that time, was tied into the exact same material I had been mining for my own work. In other words, the parts of my proposal that most affected the editor, the parts he thought might result in a story that would move (or, let’s be honest, horrify) fans of Wild Cards, were the very same bits and pieces that had been speeding my own pulse for years.
People pitching new movie ideas will often sell them with a line like, “It’s Avatar meets Alien with a bit of The Parent Trap thrown in…” Unfortunately, most of these concept combinations result in a story that, if it affects a human at all, it is to produce a shrug of the shoulders, or that well-known passage of sound through the lips that we generally refer to as “meh”.
But sometimes, the magic really does happen. An idea hits us that we just know is right, that makes us want to go running through the city shouting, “I found it!”
Our poor AI will never go running through this, or any other city.
Yes, it can combine any number of things, but lacking the ability to feel excitement, or indeed, anything at all, it fails to recognise which of those combinations has produced brilliance. This is why it’s unlikely to be our next Wild Cards author. The editors, George and Melinda, would have to do the recognition for it. They’d have to wade through hundreds, if not thousands of its awful attempts before finding something original.
But that’s not even the worst part. A good story has to get hundreds of little things right, from the beginning that lures you into reading the rest of it, through a middle that deepens the mystery or heightens the tension, to an ending that stops you from flinging the book at the nearest wall, and while from what I’ve seen, ChatGPT can do a decent approximation of many of these things, my prevailing reaction, as a reader, is a mixture of mild puzzlement and indifference.
So, for the time being, as long as we want humans to keep reading, we’ll have to rely on other humans — with the help of rising bathwater and whatever apples may fall on them — to do the writing for us.