I had chance to visit Waterstones bookshop in the Galleries in Bristol yesterday and as ever, the science fiction and fantasy section was overrun with fantasy books. They do their best in promoting science fiction, but they can only work with what the publishers allow them to have. The publishers in turn can only work with what agents and their slush pile readers (if they have a slush pile) pass onto them. Equally the publishers have to satisfy their financial people that a book is worth publishing. And let’s face it many science fiction writers are turning to fantasy because it pays them better.
You’ve heard all these points from me before. But there is now a new threat – AI writing apps like ChapGPT. Anything produced by an AI does not attract copyright protection. If you take into account that all the publishers have to do is support a ChapGPT through the writing process, then the process of producing a novel will become even cheaper for them. The writer of novels as a profession is severely threatened. This has already happened in other arts areas, like composing music or pictures, with unhappy consequences for the creatives.
But taking a slightly closer look at ChapGPT, it appears to be a neural network which bases its learning on predetermined teaching set and a maths technique called ‘reinforced learning’. While I’m no expert in these techniques, I know enough to work out the dangers of using such a technique in a commercial sense. Yes, it is ideal for day-to-day normal business, kind of repeating or doing variations of repeating what has written in the past. This is the kind of novel writing that the financial people at publishers like when it has a track record of making money. It will use tried and tested themes and plots as a basis for a new novel. So when science moves on, as it inevitably will, to show a science fictional idea is utter rubbish, these writing AIs will continue dumping it the new publications. Science fiction as a genre will end up frozen into a straight-jacket if I may mix metaphors. It would be like being forever stuck with Battlestar Galactica knowhow and not moving on (though this is a microcosm of what science fiction would cover).
There is a second issue also coming into play. Many think there aren’t as many great discoveries as there use to be in science. They think this is because we’re running out of things to discover. I think the truth is far stranger than that. The money men are keen to show results, which is why they want to invest in developing gizmos that are likely to give them a profit rather than in blue sky research. The added issue here is that blue sky research costs an awful lot more these days than in did in previous centuries because then we were picking the low hanging fruit of discoveries.
There is a very big but ( I mean big BUT in spades). Published science fiction has hardly scratched the surface of the extrapolations of science we do know about.
Yes, I’m serious about this. It also makes me very sad. I enjoy a good space opera like many others (in fact I enjoyed Brian Trent’s Redspace Rising – my review can be found here). But I look at what I’m writing and compare it to what I can lay my paws on to read, and believe me I’m writing about ideas that I have not seen elsewhere. And they’re just extrapolations of known science.
I find it difficult to believe that I would be doing anything special. There will be other writers doing similar out there. And yet I hear little to nothing of them. And how many times have I heard a reader make similar complaints.
But with writing AIs encroaching on the business of writing novels, it means the more innovative writers among us will have something to offer that AI writers can’t. In fact I would go further to say that most writers who concentrate on variations on a theme will soon find their services will no longer be required, whereas us more inventive ones will continue to be needed (if only to help upgrade the AI writers).