Love it or hate it, our phones and on-line assistants like Siri and Alexa don't just work for us. They listen to everything we say or write and pass that information along to their real masters; Amazon, Google, Facebook, etc. These marketers sift through all this data so that when we login to Gmail, Facebook or Amazon, we are barraged with ads for items we've been talking about. This ability to target ads is not the only way AI can be used but it was these megacorporations and their drive to control direct market that sparked the AI revolution. That revolution could only happen because chips, the power behind computer processing, have continued to become, smaller and faster, doubling exponentially year on year from 1965 to 2010. It was inevitable that those outside the marketing sphere would want to test the limits of this new technology. The result is that AI is no longer limited to predicting sales, It's threatening to take away (gasp), that which sets us apart from the animals. The industrial revolution eliminated mundane repetitive tasks. Computers streamlined offices, retail stores and manufacturing plants but AI isn't satisfied with taking away white collar and professional jobs. No! It wants to take over our creative jobs!

BUT before we all reach for paper bags to control our collective anxiety attack, let's take a deep breath and go back to marketing.

The more data this technology called AI, has, the more accurate its predictions become but do these predictions in turn become self-fullfilling prophecies? In other words, do we buy products because we want them or do we buy them because that's what's placed in front of us? And once we stop to peruse an ad or buy that thing, the more the algorithms, hone in on what we like  until that becomes the only thing presented to us.

This is at the heart of today's blog.

Let’s start with the basics. Publishing, like every other industry, understands the power of marketing. It would be a rare writter who doesn't offer their book on Amazon which makes understanding how the Amazon algorithms work, one of the most asked questions. It's so prevalent that a whole industry has grown up around learning how to work 'the system'. Spoiler: I can't answer that question BUT we are going to talk about another AI tool that purports to help writers hit the big time.

Technical Interlude

By now you might be asking, why is AI suddenly on everyone's lips? Partly it's marketing. Not so long ago, everyday appliances became 'smart' ones. Smart phones, smart TVs, even smart fridges. From the 1960s to 2010s, the number of transitors on a single chip doubled approximately every two years. This phenomenon, known as Moore’s Law, has led to the development of chips that now approach the size of an atom which should suggest that we're about at the limit of that technology. Proponents of AI, however, suggest that the computational power of AI, will enable us to design new paradigms, suggesting that computing power is infinite. No wonder some envision a world where AI solves all the world’s problems while others fear it will destroy the human race.

Back to our main topic, AI as a writers aid.

When does a writing tool, like Grammerly, cross the line and become an AI assistant? That’s an important question because there’s a lot of discussion about whether AI should be allowed in the arts. Grammerly, and other proofreading insect used to correct spelling and grammer which was extremely useful to writers who struggled in these areas. Only, fuddy duddies, mostly grammar teachers, fought against the use of these tools but then these tools offered to help rewrite our sentences. Does getting Grammarly or Copilot or some other tool to rewrite a sentence here and there constitute using AI? This may become an important question as more and more regulatory groups like Crafting Responsive Assessments of AI & Tech-Impacted Futures (CREAATIF) spring up.

This research project led by the  Queen Mary Centre for Creative Collaboration posted the following:

“Key findings indicate that while some participants acknowledged GenAI’s potential to improve productivity and expand learning opportunities, they also believed it was exacerbating exploitative working conditions in the creative industries. Many felt that GenAI has diminished the skill and agency of creative workers, who are increasingly tasked with reviewing AI-generated content rather than producing original work, leading to a decline in the financial value attributed to creative labour. Concerns were also raised about the lack of transparency from employers regarding whether and how content would be used in relation to AI, including for training models.”

There are two concerns presented here. One, is that AI insect will take away jobs from creative and two that the existing work of creative will be used by AI to train it to take away those jobs. AI in the creative world, offers predictive software that authors can use to correct spelling, grammer, sentence structure and even suggest alternate words (like a thesaurus at your fingertips). To function thos software requires data found in published works such as dictionaries, thesauruses, and books on grammar (which can vary)

This software does not generate text so much as it highlights potential errors. To write text, writers need the Generative software.

Let's start with predictive AI that goes beyond the basics. Let's talk about software that claims it can turn your book into a best-seller.

Since I was doing this research, Google decided to help and suggested I check out a tool called IRIS. It wasn’t cheap but I forked out the money - all in the name of science or, in this case, curiosity.

I entered my credit card number, uploaded my manuscript, and in 24 hours, I had a detailed report on whether or not my book was likely to appear on the NY Times best-seller list. The book I uploaded has already been published so I already knew that it was nothing in that league. But where had it let me down?

The first page of the report gave me the most relevant bit of information, my books sellability score. While not destined to be a best seller, it was still considered a good book that was rated as likely to sell a decent amount. (We shall see about that). The sellability score was based on a combination of analytics. Those analytics were determined by feeding the AI program with best selling books in each of a number of genres and tracking things like sentence length, types of words used, etc. The AI software read my manuscript and compared my book's analytics with comparable best sellers and gave me an expected star rating. An acceptable 3.9 but the software didn't stop there. It was happy to tell me where I had fallen short of the coveted 5 stars.

Since the benchmarks were based on genre, it was important to properly identify the genre I thought I was writing for. This was difficult because I don't see myself as a science fiction writer but my stories have an element of out of time or in another place. Without magic, I didn't qualify for fantasy either. I settled for women’s literature and dystopian. My manuscript received its highest scores when rated against books in the crime and speculative fiction genres. Drilling down on crime, my ms received high marks as a thriller, and the next highest score was for action. Drilling down on speculative fiction, there were high marks for dystopian and low marks for science fiction. (No whizz-bang technology, no aliens, no smart women with big busts and short skirts)

An interesting feature was a graph of the age group each part of my manuscript was likely to appeal to. My book follows the activities of characters from different ages, backgrounds and genders. There’s no one protagonist, so the graph was quite colourful as the individual characters appealed, analytically speaking, to the age of the protagonist in that chapter. Personalky, I thought my manuscript would appeal to adults but the analytics suggested a three-way split between, young adult, new adult and adult. That  split lowered my over all genre score.

Next came style fit. The software suggested that my manuscript was harder to read than other books in women’s literature. The software suggested I shorten my sentences and that I use less complicated words. (Am I the only one to find this sexist?) The analytics were based on a standard readability score. My readability score wasn’t that low (low means hard to understand) so I think the software was confused about the reading level I was hoping to attract. Apparently, even 7th graders should be able to read, if not understand, what I'd written.

Next came Emotion and Pace. The software measures the number of words and sentences that carry emotional sentiment. A best-selling Romance novel, for example, would have a high number of emotional sentiment. While my novel didn’t reached the heights of a best-seller, it met the metrics for my genre. The software further provided a break-down of my sentences by emotion; fear, happiness, disgust, surprise, anger and sadness. It also provided a diagram, in colour, of the emotion each chapter elicited.

I found this quite useful because one of the things I do as I revise is decide what emotion I’m hoping to convey and then re-read and re-work until I think I’ve achieved the desired effect. In this particular book, I wanted to run the gamut of emotions with the tension increasing as the book progressed.

Pacing, the report pointed out, is the main reason for a publisher rejecting a manuscript. This too will vary by genre. Comedies move faster than dramas, adventures faster than thrillers. It then provided suggestions on how to pick up the pace in those sections of my book where it dragged. I should mention that with all the graphs, I was able to drill down to specific sentences that the software deemed sub-par. I must say, I was pleased when it came to pacing because there, the program thought I’d reached best-seller level.

So what did I think of the program? As a writer, I was impressed with the detail on emotion and pacing but like any tool you have to know when its advice makes sense to you as the writer. You don't rewrite a sentence because the software says it slows the pace. Sometimes you want to give the reader a chance to take a breath. On the other hand, It's interesting to see a pacing or emotion graph. That bird's eye view is something I often miss when I've been working at the chapter level.

The other thing to keep in mind, is that the software can't tell if the plot is interesting, or If it was developed in a consistent or even, entertaining way. It can't tell you if you've left major questions unanswered or if there were holes so big that you could to drive a bus into them. Nor could it say anything constructive about characters. How relateable they were, how witty or stilted their dialogue was. It can't graph character arcs. Nor could it determine if my ending was satisfying. Which leads me to my next question:

If publishing companies use this software to reduce the number of manuscripts in their slush pile, is there a chance that a really good novel with an experimental format or a new genre, will be rejected because it doesn’t conform to the standards for a best-seller? The publishing industry has already been accused of being too conservative, focusing primarily on authors that are established brands, leaving little room for emarging artists. The other problem I foresee with predictive AI is that writers will try to give the software what it wants, forgetting that they have to develop their own voice. Which begs the big question. Why do we write? As I’ve mentioned in previous blogs, most authors don't earn a living wage so if you’re expecting to become a best-seller, with or without the help of AI, then you may wind up disappointed. On the other hand, if you are compelled to write, then predictivec AI might help you to spot flaws. I remember reading that Daphne de Maurier plotted the emotional content of her chapters over and over until she got the tension to peak at just the right moment. Would Iris have saved her months of labour? Quite likely. In other words, if you know what you want AI to tell you, you will be better equipped to use that information wisely but ultimately, writing is still a black art. How readers rate a book is as much about word of mouth expectations while what they is more about marketing.

But what if we got AI to do all the writing for us? If we define the characters, the plot, the genre but letcAI construct the sentences for us, are we still writers or are we something else.? As I pondered this question, I thought of a friend who had a great story about his early life. Not only wasn’t he a writer, English wasn’t even his first language. Fortunately, a publisher heard his story and teamed him up with a ghost writer. The finished book bears his name as the author but the book was written by someone else. No one in the industry raises an eyebrow when a book is ghost written but when AI is the ghost writer, everyone gets upset. And that brings us to the topic of Generative AI. So tune into the next instalment when I trial software to write for me.

DISCLAIMER: No AI insect were used in the writing of this material which explains my spelling, grammar and all.my side remarks.