Has AI Grown Up yet? Debbie Bennett

There are many things in life which polarise people. Off the top of my head, I can think of politics, religion, climate change. Opinions on these things can make or break friendships and other relationships. How many times have I read online: If you vote X, then unfriend me immediately and similar statements. I have friends of all persuasions and their opinions on most things are all part of what makes friendships rich and interesting both online and in real life. 

In the online literary world – well, social media anyway – the controversy of the month seems to be artificial intelligence. People come at this from all angles. Some hate any use of AI in any shape or form – it’s stealing, it uses too much water, it’s cheating, and they can go on and on about saving the planet and the livelihoods of every creative persn living on it. Others see it as the future and why shouldn’t they use all the help they can get, they don’t have any money and don’t see any problems. 

The reality, I suspect it is way more nuanced. Assistive AI is used by most of us without thinking about it and has been for years. Spellchecks and grammar checks, whether built into software like Microsoft Word or as standalone software. Apps that tell you how readable your prose is and whether you should rephrase a sentence – the key words here being whether you should. Assistive AI makes suggestions, but leave you, the author, to make the decisions. The vast majority of authors have no issue with any of this.

Generative AI is different. It creates stuff. Like the post I wrote back in July 2007 where I asked ChatGPT to write something in style of DJ Bennett. Or the post on AI art from 2023. But AI has grown up over the last few years. You see it more than ever now – everywhere you look. On social media, on at least 75% of adverts: It’s not X, it’s not Y. But it’s Z. Lots of short paragraphs with short clippy prose repeating the same thing. See enough of it and the patterns become obvious. But it creates stuff from its databanks, from all the gazillions of words its been fed and digested (and the illegal feeding of books without the copyright owners' consent is a whole subject on its own). So it's not creating anything new - just learning how words connect to other words in different contexts. And every time somebody feeds the machine on the pretext of detecting whether AI has been used or not, the monster consumes yet more text; so the more words it eats, the more a piece of text will look like AI as the monster gradually becomes more human.

Scary isn't it?

On social media, posts are awash with books created by – or with – AI. At first it was 'colouring books'. Create the smallest book you can get away with and stuff a load of AI-generated pictures in it. Upload it to Amazon and, hey presto, sit back and watch the millions roll in ... Except that Amazon is catching on and suddenly the authors are online complaining that their accounts are being terminated and they don't know why.

(c) Mikko Paananen, CC BY-SA 4.0
https://creativecommons.org/licenses/by-sa/4.0,
via Wikimedia Commons

Now it's 'children's books'. Vacant-eyed, round-faced cartoon kids with identical wide smiles. The animals look the same too and the stories are always teaching kids about emotions and dealing with reality - semi-educational, I guess. The covers often have a yellow wash, which appears typical of an AI cover. But look more closely at these covers and covers in other genres and find the devil in the detail. It's not as bad as the picture here from 2023, but it's still there.

Limbs are still a giveaway, particularly when there is more than one person standing close together. AI will often put a hand on the wrong way, or one arm will blend into another. Hair will blend into clothing or accessories. Genre books can often have a kind of gritty sparkle across the whole cover.

And trust me, people do notice. Readers do judge books by their covers. And people do think that if you used AI to do your cover, you may well have used it to write the book too. And really, what is the point of that?

And what happens when the next generation of AI is so good that we really can't tell the difference. When Frankenstein's monster comes fully to life?