The Black Dog
A most unwelcome hound indeed.
The hound that sits upon my back, weighing me down and immobilising initiative, is something I have lived with for many years. What I am actually talking about, euphemisms aside, is often called ‘high-functioning depression’ (although sometimes I doubt the functioning part). As many can attest who also relate to this mental health state, it just comes and washes over you. “Cheer up!” is often the last thing you want to hear, because it assumes that it is just a mood that you can control or snap out of, or that it is a result of ‘something’, and that if you fix that ‘something’, everything will be better again. It doesn’t.
That is not to say that when the black dog pounces and pins you down (sorry, euphemisms again), genuine issues become matters that do make you sad. And well may you ask: did the depression make the issue something to become sad about, or did the issue bring about genuine sadness, with depression descending as a result? Chicken or egg – you decide. I’m too unmotivated to get out of bed, much less worry about causes and effects.
How people deal with this varies. Some turn to medication (prescribed or otherwise), and friends who have, swear by it. Others just let it win and hope that the fog will lift tomorrow. And then there are others – like me – who choose escape: a retreat to a happy place where they can build worlds, create stories, build characters, and then write it all down. So yes, writing for me is as much therapy as it is work.
But something of late has punctured that happy place and let the black dog track me down – and that something is artificial intelligence. As writers, we feel this intrusion viscerally. It has become the topic that is sure to raise the temperature of a room as soon as it is mentioned. The latest National Writers’ Conference is a good example: whenever the topic was broached, it was mentioned with trepidation, and those who let loose fell just short of having steam gushing from their ears and a complexion that could best be described as crimson. For me, it just made me sad.
You would have to be living under a rock not to have encountered A.I. slop in all our social feeds: the same tired script, the same structure, the same A.I.-generated character whose mouth just doesn’t feel right when it talks. Even now I can see the structure of what I have just typed: the statement, the colon, the three examples. But this IS my voice, these ARE my words, this is NOT A.I. content. But when you hear it on every post while scrolling, you flick your thumb and a little bit of bile comes up to accompany the disappearing intrusion. So I agree with the critics – A.I. slop is vile – but, and there is a but, I would not throw the baby out with the bathwater either.
To hear an opposing view, I encourage writers to seek out Lara Helmling from the YouTube channel The Guerrilla Publisher. Her take on A.I.’s use is compelling and not without merit. Her analogy (because I love a good one) likens the argument to the Dr Seuss book ‘The Sneetches and Other Stories’, where the decision to use or not use is only dividing the community, with declarations of usage likened to Sneetches with or without stars. I know the issue is much more nuanced than a simple Dr Seuss book, and I’m sure Lara would concur, but it is something to consider. Lara also raised a good point about the merits of having a declaration attached to work at all. Why are we writers being forced to declare which tools we use? Does a gravedigger (sorry, blame the black dog for this analogy) declare that he used an excavator rather than a spade? Or did writers of yesteryear have to declare they used a typewriter over quill and parchment? I know these are simple arguments around tool use rather than content provenance, but still – should the proof not be in the pudding rather than in a written declaration by the chef?
Having written over many years, I know A.I. is not as new as many believe. For decades we’ve had access to word processors; then came the red squiggly underline correcting our spelling, the blue squiggly underline for grammatical errors, the right-click thesaurus at hand, and, with tools such as ProWritingAid, AutoCrit, yellow squiggly underlines to rephrase passive sentences or make the content clearer and more succinct.
A.I. is a tool and it can be used and misused, like any other tool. Will it end humanity or cure cancer? We don’t know. I’m sure we will find out soon enough, but I would suspect the answer lies somewhere between those two extremes.
As a former I.T. specialist, of course I use it. It has designed my websites, designed my business cards, created my book covers and illustrations, and become my editor – line, structural and developmental. Having written a book long before A.I. was around and having finished one using it, I can honestly say the process was not only more efficient and more rewarding, but has produced work far superior to my previous efforts. What I made sure of before I started was that it was not to be a co-author; I strictly banned it from creating or suggesting content. The words were mine – in fact, the experiences in the book are based on my own lived experiences. Swap the hammam in El Jadida, Morocco, for one in Damascus, and you have the basic outline. Swap my self-erasure attempt in the 90s (along with the interactions with the medical establishment) with my character’s attempt, and you have the foundation for my writing. Even the muses I had as an inspiration for my characters are my muses (you know who you are).
A.I. not only fixed my punctuation; it was great at tracking plants and payoffs across the 100,000-word manuscript. It was good at detecting point-of-view drift, tense changes, and where first-person paragraphs were actually required as opposed to my usual third person limited. Structurally, it even challenged some parts of the manuscript that I had written. I remember a response where it said, “Blair, we need to talk. Your writing here is lazy and you are relying on a stereotype – you can do better.” I then went on to have a moral argument with a machine, and the machine was winning. I was lazy. And worse, I used racial objectification (in a complimentary way), but it was still objectification – and I rewrote the passage using attributes rather than race.
Because A.I. is contextually aware, it could alert me when I was drifting too far from the character profile I had built. The visual imagery I had produced using A.I., and used to create my Miro storyboard, helped ground me visually as I wrote. When I needed feedback on choices I had made, it was a tool to turn to that offered detailed and nuanced evaluations, turning a very lonely process into one that was genuinely rewarding and enriching. So when the purists espouse ‘Thou shalt not…’ even sign up for an account, all I can think of is the baby crying somewhere around the drain.
So why did A.I. bring about my current ‘funk’?
Publishing has gatekeepers, and with the deluge of slop, they are using tools to detect A.I.-generated content. And as someone who has spent an entire year writing my latest novel, I wanted to see how these tools viewed my work. I was curious as to whether allowing an A.I. to massage and correct my text would have an impact on the detectors. I would rather know before being told – nay, accused – that my work has been deemed unworthy. Chapter by chapter I had them scanned, and the results were mixed, and it did not make sense how those assessments went. Many chapters came out 100% human, while other chapters were stamped red – all A.I. Same writing process, same author, but completely different results. Then I uploaded my manuscript from 2016–2017, before A.I. existed: same verdict – 80% A.I. content.
I had been told not to trust these detectors, and logically I can see that. (Even Shakespeare gets flagged as A.I.-generated.) Yet it still depressed me that a machine can label all my work, even work written a decade ago, as slop. The effort to write – a year of labour, pretty much full time – labelled red, with a percentage figure stamped on it as supposed evidence. ‘Chin up, don’t let it get to you’ – easier said than done.
So where to from here?
Given my manuscript had widely varied results, I looked at the text in those passages to see what about it made the detector scream foul. The clues were there. These passages had relied on euphemisms and analogies, because they were a hell of a lot better than using cheap adverbs. I am now faced with the decision to leave my original words intact and have detectors blare sirens, or to rewrite my manuscript to try to satisfy them. I am still sitting in that quandary – release by my deadline, or do a fourth draft that could well be flatter and less compelling in the end and take another 2-3 months to produce. In short: I both love and hate technology, and that dichotomy has left me wondering about the wisdom of getting out of bed today.