Please note, I normally avoid AI because I'm clueless about it, but for the purposes of this article, all images are AI generated unless otherwise stated.
I’m not comfortable with AI. I don’t really understand it, and that makes me reluctant to use it. I typed in a question about motivation once, and got some good advice, which I promptly ignored, but otherwise I’ve watched with caution. You can find previous posts on my blog here and here.
The internet has never been a place of unrelenting truth. Right from the start, there have been emails and posts and so-called advice that have been unfounded, fictitious or downright malevolent. There are even new words to describe some of the more predatory online practices, like phishing or catfishing. I’ve always taken the view that I can’t trust the internet when it comes to health, wealth, or religion. I can access sources, and some will be more reliable than others, but if I’m in doubt, I need to speak to someone qualified.
These days, that’s not such a guarantee. I know that there are accounts of AI being invaluable in all sorts of ways, but the professionals who use it are human too. There have been cases, including one quoted in the Law Gazette here, where the AI has had a dream. An AI has been asked to provide documents that includes supporting cases. I guess that the instructions should have included real cases that have previously been heard in a real court. As it was, the documents were plausible, persuasive, and absolutely undermined when it was shown that the supporting cases that had been quoted didn’t exist but had been dreamed up by the AI. The qualified legal professionals who were getting the documents together didn’t check. I can imagine the absolute fits that the judge had at that lack of diligence.
To me, dealing with AI is like dealing with the Fair Folk. You have to ask your question so carefully or the answer that you get is like fairy gold that turns to dead leaves in the light of day and is useless. I’m okay asking for help with a route if I’m driving to an unfamiliar place. I know that I’m going to go over it on the map anyway and I’m quite happy to adjust it to suit myself. In theory I could ask for recipe suggestions, like Weary Wolf on their YouTube channel who asked for 21 meals for one to cover a week on a budget of £20, and got an okay, if potentially uninspiring, result, here, if you’re interested.
It reminded me again of dealing with fairies. These weren’t malevolent fairies. They wanted to be helpful and useful. But when there was excess money, the AI suggested that it was all spent on cheese, because more cheese makes everything better. Obviously, the AI had seen a lot of the ‘cheese makes it better’ memes and took it literally, like a brownie or elf trying to understand mortals and not quite getting all that they saw. Thinking about the recipes, I would have suggested more fruit and veg, or perhaps more seasonings, but the AI was going from its instructions and insisted on more cheese.
One of my friends, posting as Supernovauk, is using AI to make music. They’ve got it on YouTube, here, if you’re interested. It’s light and pop and fun, and it reminds me of the days when I was young enough to dance the night away. I know how hard they’ve worked on that music. The lyrics are all from them, and they’ve spent a lot of time and effort getting the music exactly where they want it. It’s absolutely not some slop thrown together, but a real creation using AI because the access to musical instruments and the time to master them isn’t available to them.
I feel like it’s a deal with the Fair Folk, making a deal with fairies, which is a dangerous thing to do. My friend has put in so much effort and care, but because it’s AI, they can’t copyright it. It could just slip through their fingers. Other AI efforts, particular in the romance book area, are like dreams or perhaps like fairy gifts. They lack an indefinable something, like fairy cheese made without salt or lumpen bread made without an iron oven. Or, frankly, are plain awful.
Authors have a few different problems. The publishing world is getting overwhelmed by some extremely poor AI books, churned out with little regard for a reader. Romance has been particularly vulnerable to this. As a voracious reader of romance novels, I have lost count of the number of books that I’ve picked up, read the first page and then rejected because the AI writing is appalling. However, if a determined AI user is willing, they can publish dozens of books in a day, bung it on Kindle Unlimited, and if enough people like me read the first page, they’ll get plenty of single pennies from the Kindle Unlimited accounts and make a nice living from it.
Those AI collections of words have sparked a backlash and there is a witch-hunt against anyone suspected of using AI. There have been a lot of accusations thrown around, and not all of them have been grounded in reality. It’s not always obvious whether it’s AI or bad writing, or even if the author’s writing style is one of those that has been heavily copied by AI models. I’ve been acting as a beta reader for my friend Holly White and unfortunately their writing style is one of those that has been copied. And it's not AI! They’re writing a series of fan fiction romances, based on Quidditch teams in the Harry Potter universe, and I’m loving them. I’ve been with them every step of the way and those stories have been hammered out on their keyboard, tap by tap, in evenings after work. As they’re caught up in the joy of writing in a genre new to them, they are racing ahead and so books are coming out quickly. Not only that, the writing is, in my opinion as their trusty beta reader, awesome and not comparable to the AI autochurn that’s been thrown together and dumped on Amazon (and you can find Holly’s work on Wattpad here, if you’re curious). I’m not sure how that will protect them, though, if accusations are launched at them.
I had just booked a cover from GetCovers for An Apprenticeship of Ghosts, which will be coming shortly, knowing that they don’t use AI, and I wondered what I could do with Canva, AI, and a little time. They have an AI function, and I clicked on it, nervously, and put in a prompt book cover with a raven flying holding a ring on a ribbon. I got this.
I didn’t give a title. I certainly didn’t give it a subtitle or tagline. I absolutely definitely didn’t invent the name for the author. I was unnerved enough to check, and at time of typing, there wasn’t a book or author with those names on Google. AI had dreamed a story to go with those twelve words that I used as a prompt.
I came away with two thoughts. One is that I want to write that story. The other was that I gave the seed to the AI for this dream. I wonder what unbidden dreams haunt AI.
No comments:
Post a Comment