To AI or not to AI
Before I begin, let me make several things really really clear. I am no tech wizard: I get by, and I consider myself pretty good with Word but for most things I shuffle my way along and then slip into ‘this is easy' because of the beauty of repetition, or ‘I don't understand' caused by unfamiliarity. I have to leave notes for things that I encounter rarely so I don't have to keep relearning stuff.
Next up is I don't really have a drum to beat against AI. I mean, I don't like that it scrapes artists' words from the internet and uses them without recognising where they came from. But it's here to stay. If you could put AI in a box and present it to some people, they would stab it or beat until that box was a pile of splinters and whatever was inside it was mush. They hate it quite vehemently with a passion because they see it as a threat* and they see it as a virus.
You might remember buses in the UK back in the 70s and 80s all had conductors who’d take your money and crank out a ticket. Well, when their role was made redundant by tech the driver had at his fingertips, there was uproar. But no one nowadays can remember those conductors let alone gives a toss about them. Like I said, it's here, get used to it. And you probably already have encountered it without even knowing.
Next is AI itself. I believe there are two kinds: that which helps you create things, and that which creates things for you. This post is about that which helps you – not that which does all the work for you.
Lastly, and this is the most important bit. I LOVE WRITING. I never have and never will allow AI to replace me and flog it off to you as my own. I write my own books and they are wholly my own work – mistakes and all. Why would I get some machine to do the things I love? I wouldn't. Some people reckon a piece is written by AI if they see too many em-dashes in it. I think that's just bollocks – I use em-dashes frequently, so I can say without reservation that that statement is bollocks.
I understand that some people use AI to sidestep difficult to research topics or to create dissertations, and I've seen videos of lecturers losing their shit because they've found out some or most of their students have used AI to write their papers for them. I think life will eventually find out those people and turf them out: there are some things you cannot skip; there are some things you need to do to make you the professional you aimed to be when you took whatever course it is you're trying to cheat at. Life will get you even if the lecturer does not. You are a cheat. There - that's my little rant over.
But anyway, with those things out of the way, we can finally get on with the topic in hand.
I recently looked into producing an audio book of A Random Kill. It would cost me somewhere in the region of £2000 to £4000 to have a human narrate the story. Let me make this plain: I cannot afford that, especially when you consider that the narrator would also like royalty share as well. And also, consider that the main audiobook platform around these days wants the rights to your audiobook for seven very long years, meaning you can't offer it anywhere else. All in all, I think it's a pretty rough deal for a small-time author.
Readers ask about the chances of audiobooks all the time. Yes, all the time.
So I began looking for an alternative and found one – costing around $200. I get to keep my rights, am able to sell on multiple platforms, keeping all the royalties, too. Bargain!
I lost around five days messing about searching for the right voice and exploring all the options available to me. These five days were not solid days, you understand, they were blocks of time I stole from family- and work life each day.
There were lots of voices available – those that were sampled from your own voice, those that were created from prompts that you type into a suggestion box, and those you picked off the shelf. I picked one off the shelf that was basically a real live person who had donated a sample of their voice for people like me to use. Understand this: they put their voice up there voluntarily and they get paid for me using it.
So the AI I was using to create my book was not morally corrupt, and it certainly was not putting anyone out of work because if it hadn't been for this techy option, I still wouldn't have chosen the human narrator, as I said above. Another plus was the cost: if it was cheaper for me to produce, and I got to keep more of the royalties, listeners got it cheaper too. Isn't that a good thing all round?
You can feel a but coming on, can't you?
You're right! And here it is: the AI narrator is not as good as a human would be at reading emphasis or dialogue and drawing inference from those words. Most of the time I wasted was trying to get the silly thing to leave a pause between paragraphs. It could manage that perfectly well between sentences, but it would rush into the next para immediately after the last one finished. It sounded hurried; it didn't have a natural break. And only sometimes would it listen to me when I asked it to pause.
Sometimes – well, often, actually – it just wouldn't get the emphasis on certain words I needed it to have. I might have been making a sarcastic comment or a harsh comment, and often it would miss it, and I’d have to try and force it to accept the emphasis... but it still didn't sound right.
But the biggest problem I had was inconsistency between paragraphs, particularly those with dialogue in them or consisting wholly of dialogue. The voice would change inexplicably, sometimes the accent would change, but certainly the tone would, and I had no idea how to change it back (remember the first paragraph about me not being tech-savvy?). Sometimes it was horrendously obvious, like a slap in the face, but mostly it would be subtle, but I could still hear it, and it irked me enormously. In chapter 12 as we’re introducing Bradshaw, it was so horribly wrong that in fact, it was the final straw, it was the thing that caused me to pull the plug on the whole project. He would speak with a neutral English accent and then slip into broad Yorkshire. How the hell do you tell a machine not to do that? I couldn’t fathom it. Added tot he already annoying problems of delivery speed and no pauses, you can see why I just abandoned the project. On top of that, I was under a time constraint: if I didn’t get a move on, I would hit the monthly renew date - another $99 please!
Let me straighten things up before we leave it like that. The voice and the way it spoke would have you utterly convinced that it was human – no doubt at all. But I think I'll have to wait another couple of years until the disparity between paragraphs is finally sorted out before I tread that path again.
In the meantime?
In the meantime, I shall continue my search for a real person who can narrate my stories and read the words as I heard them while writing them – no mean feat! Say this next bit in whisper: But I need a person who enjoys reading the various characters and who won't charge me more than I can afford to pay. A rare beast indeed.
*Remember my saying some people see it as a threat? Well, in one sense they are right. How many times have you been doom scrolling only to be shocked by something you later learn is a construct of AI ‘written' or ‘videoed' by some lazy bastard who can't be bothered to research and write it himself, and who lets AI scrape shit together and present it to the public as fact? This is more than scary to me – this is the kind of material that makes me angry because it's a lie, because it sets people against people, community against community, and it's very dangerous. So in that respect, I'm right there alongside those who would smash that box of AI to bits – even though it's the lazy bastard at fault, not the thing inside the box.