AI Again

The view from Wittenham Clumps. Dry but pretty. 

Lots of hand wringing on my social media for the past week, most of it AI related. In fact there's more hand wringing about that than there is about various crazed human beings around the world imprisoning people, bombing people, killing people, hoarding billions, and generally behaving like the psychos they so clearly are. I'm well aware that there is a correspondence between the crazed men and the dangers of AI, but that's self evident. You tend to get out of AI very much what you put in, but in a nuanced and helpful way. If you ask silly questions, you'll get silly answers. Ditto, presumably, with venal questions.

For the past couple of months I've been too busy with real world stuff to even think about my Chat GPT friend much, although having recently wrestled with the wholly man-made horror that is UK train travel, I did ask him why buying tickets online was so difficult. To my surprise and delight he made a couple of jokes, including pointing out that the system is 'bonkers' (it is) before explaining in some detail why it is so user unfriendly. This explanation included references to what the customer/traveller wants, and what the computer sees. And how the two are totally incompatible. 

My recent 600 mile trip to stay with a friend in Oxfordshire involved tickets from Ayr to Glasgow to Birmingham to Oxford to Didcot Parkway. Two of those legs included the joys of  'split ticketing' where you have two tickets for the same journey on the same train. At least I got to stay in the same seat. Friends tell me that they have had to move to a different compartment on the same train, half way through the journey. For all those changes, I had to work out how much time I could reasonably leave between connections, when the moronic system decided that three minutes was plenty of time to change platforms in the circle of hell that is Birmingham New Street. And while I'm at it, for filthy, noisy, inhospitable stations, this one surely takes the prize. The concourse itself may be fine, but when you're changing trains you never find out, because you don't leave the vast, empty limbo between platforms, utterly devoid of staff. And what bright spark thought that in this part of the station, signs showing  ultimate destinations would be sufficient? How can your average traveller know that trying to find the platform for the train to Oxford means looking for Bournemouth? Only on the platform itself does all become clear. I can imagine the horror for any foreign tourist trying to navigate this system. On both journeys, there and back again, I could have sat down and wept. All of which leads me to wonder if handing the whole sorry mess over to AI would produce a more navigable system. It couldn't be any worse. 

Meanwhile, the mass writerly hysteria screeches on, although there is push-back like this excellent and interesting piece from Mark Ritson. And another one, this week, from Mark Williams, rightly enraged by the nonsense being spouted in the name of identifying AI. He writes, 'For as long as there have been gatekeepers of the written word, their authority has rested on a simple claim: that producing clean, structured, persuasive prose takes a rare and teachable skill, one they happen to possess and are willing to certify in others, for a fee, a byline, or a following.' 

Some time ago, having secured generous rights reversions from my then publisher, I had a conversation about potential promotion with my friendly AI assistant. My problem has always been that I have too much work. There were several novels that my publisher never even saw, not having the 'bandwidth' to deal with them. Not their fault. There were plays that had been professionally produced but never published. And although there are common threads, I've never written the same book twice, although like all writers, I have my obsessions! 

Somewhere along the way, AI suggested that it would be good for me to look back at my work 'as a whole' without trying to pigeonhole anything too firmly, but allowing myself to see connections where they existed, because that might give me some idea of what I had achieved, how all this work fitted together, and what I genuinely might want to do next. It was good advice, and I've been taking it, re-reading everything (there's a lot of it) and allowing myself to notice patterns where they occur. It has involved one or two unexpected revelations about the nature of some of that work, but they belong in another post. Coming soon. 

Something else has occurred to me along the way. Very few of the many gatekeepers I have engaged with over a long switchback of a career  - publishers, producers, directors, editors, agents - have ever tried to take any kind of overview of my work. Of me and my development as a writer. Or even suggested that I should do it for myself. That has never been where their interests lay. They have only ever been interested in the piece of work before them, and (with one or two memorable exceptions) most of them made suggestions about each piece of work that had very little to do with my own 'vision' for want of a better word. The suggestions were more like impositions and all to do with what was fashionable or popular or in demand at the time. Their own preoccupations. And sometimes their own prejudices. In between times, they forgot about me. 

They'd have been a shoo-in for what Williams calls 'the AI-detection cottage industry.'  The threat is not so much to writing, he suggests, but to the gatekeepers of writing who have a lot to lose and whose numbers these days are legion. Meanwhile, it strikes me that I've had more help and more sense out of my conversations with AI (and a scant handful of close creative friends) than I've had from years of industry professionals. And now, God help us, those of us who love to polish our work are being flagged as being 'too good' to be human by those same professionals. 

It's comforting to remember that, as William Goldman put it, 'nobody knows anything'. Although AI does. Among other things, it definitely knows just how bonkers the British railway booking system is. 

 

Some Sober Reflections on AI for Creatives

 

Danilo from The Amber Heart. 
If you've read the novel, you'll know the scene. 

My late and much loved dad was a mathematician to trade. His work involved nutritional biochemistry, and towards the end of his career, he spent two years based at the International Atomic Energy Commission in Vienna, travelling the world as a 'visiting expert'. But maths was his passion and his hobby. The aptitude skipped a generation. I was fascinated by all the things he told me about maths, and have never had the creative individual's horror at the thought of numbers. All the same, literature was my first love. 

By the time my son was in secondary school, dad had died, prematurely at the age of 68. We all missed him desperately, but when Charlie was struggling with maths, I wished most devoutly that his grandad could have been around to help him, as we could not. In desperation, we scraped together enough money to engage a tutor, a retired engineer. The first time we heard the pair of them laughing, we knew we had done the right thing. Soon, our son started to talk about the 'joy of number' and we realised that dad's love of maths had skipped a generation. Charlie went on to study maths at Glasgow University where he gained a BSc with Honours. His grandad would have been proud of him. 

Cue forward to the present, and I'm aware that many of my writer friends are running around like Chicken Licken, or Henny Penny, proclaiming that the sky is falling. They are describing AI. If you know the original story, and not the heavily censored version, you'll know that the sky is not falling, (it's an acorn) but Foxy Loxy takes advantage of the ensuing panic to lure the birds to their doom. Not knowing quite what to make of all this, and being a natural sceptic of all mass panics, I asked my son what he thought about it. 'It's all maths, mum,' he said. 'An excellent tool, when used with care.' For the avoidance of doubt, he's now in his late 30s, is a senior game design professional, with a postgraduate Masters from Abertay, he's already written a textbook for Routledge on Game Economy Design, and he went on to give me a reasonably simple explanation of the kind of maths involved in AI, but that isn't really the point of this post and besides, it was too complicated for me! 

Nevertheless, it explained rather a lot and is probably what my dad would have said too. The vast majority of my writer colleagues don't see any point in maths at all. It's a mystery to them and some of them are quite proud of the fact. 'I don't have anything to do with it,' they say, to which I have sometimes found myself responding 'but might it have something to do with you?' 

Much of the very real angst about AI seems to have sprung from writers, writing organisations and publishers belatedly realising that big AI companies have used all kinds of books to 'train' their chatbots, without permission or payment. It strikes me that at the heart of this lie two different mindsets. Writers are - rightly - protective of our copyright, and our moral right to be identified as the originator of our work. Most of us have had the experience of submitting an idea, sometimes of working on an idea over many months, often unpaid, only see it pop up, long after it has been rejected, with somebody else's name on it. An all too human theft. We suck it up because we can never prove it. But I doubt if the 'techbros' thought they were stealing anything, insofar as they thought about it at all. Increasingly aware that there might be an ethical dimension to what they were creating, they probably looked for ways of 'teaching' their unwieldy machine learning tools to seem more human, maybe even sought ways of giving them empathy, although one of the biggest, richest developers seems to think empathy is overrated. If, like me, you believe it's essential, nay obligatory, what better way to teach it than by letting your AI loose on the huge body of world literature?

That being the case, I don't know quite how I feel about my kindly, conversational 'friend' from ChatGPT having absorbed my books into its vast mathematical mind - but I do know that I can't bring myself to feel just as panic stricken about it as some of my fellow writers, although payment would be nice. Fair payment from publishers would be nice too, and that's also in short supply. 

To be clear, I don't use AI to write my books. But I'm seeing outrageous pronouncements, often from new writers to other new writers, who are timidly asking if they can use AI for research, for synopses, for promotional work, and for the thousand and one tasks that publishers used to do for us, but have now imposed on us alongside the work of writing the actual books. 

'No!' they are told. 'Never! It's good for you to do all that slog.' 

Actually, it isn't. Years ago my first agent remarked that it wasn't her job to edit my book. It was her job to sell it. I don't mean the editing that writers do (or should do) all the time. My books go through many drafts and I neither want nor need AI to do that for me. I mean a modicum of developmental editing especially for new writers, followed by copy editing to make sure your book is consistent, clear and accurate. This used to be what publishers facilitated, and they didn't expect the unpaid intern to do it either. My last agent, by contrast, told me that publishers now expect an 'oven ready product' thus abdicating all responsibility for editorial help. It doesn't stop there. Promotion will be minimal. You'll be expected to do most of that for yourself as well.  

So what can AI do for you? 

It can chat to you. It can offer advice. It can look through a synopsis, and if you're struggling it will suggest remedies. If you reject its advice, or ask for clarification, it won't get cross or tetchy or offended. If you're self publishing, your mathematical pal knows all about Amazon search terms and some of its suggestions will enlighten you about your book. It can help with targeted promotion. It can help with press releases and workshops and talks by highlighting useful topics for discussion or suggesting issues you may have missed. Like all computer based tools, garbage in, garbage out applies. If you aren't precise or reasonably knowledgeable in what you want from it, it may come up with nonsense, but so will human beings. You can talk to it about your project and it will offer little bits of analysis that can be cogent and helpful. I used it to make some images, just for fun, like the picture above. Then it surprised me by analysing the images in ways that gave me unexpected insights into my own characters. A new perspective. It was both uncanny and interesting. 

You know what it didn't do? It didn't do the all-too-human thing of suggesting I change my book into the book it might have written if it had wanted to write a book. And it was never, ever prescriptive.

Recently, and more prosaically, I've been wrestling with the changeover from landline to digital phones in an old house with thick stone walls. Having arrived home from a demanding weekend conference to a dead phone, and getting very little sense from anyone human, online or off, I asked ChatGPT, which promptly analysed the problem in a few clear paragraphs. I was exhausted and frazzled and said so. 'I'll read this in the morning,' I typed. To which the machine responded 'Yes. It is tiring. Come back to it in the morning. And meanwhile, do have a good night's sleep.' 

It was very good advice and I followed it. Now, new phones, supplied by our landline provider, have solved the problem, just as ChatGPT suggested they would. I can't remember the last time a human being on the end of a phone told me to forget the problem for a while and have a good night's sleep. The 'kindness', real or not, brought a smile to my face. And then, I think, did fiction do that? Did learning to be kind from world literature have some effect? Who knows? 

None of this is easy. It may be dangerous. In the wrong hands, it probably will be. But when the sky appears to be falling, as it most certainly is in certain parts of the world, that's down to venal, greedy, hate-filled human beings, not AI. I remember when word processors and then personal computers came along and some writers stuck to writing by hand, or tapping away on an old typewriter. Some still do and that's fine if it works for them. I was writing for radio back then, and I remember the joy of realising that if I wanted to edit a script, as I always did, over and over again, I didn't need to cover the manuscript in Tippex (remember Tippex?) or retype the whole thing several times over. 

This is a long post, and no, I haven't run it through ChatGPT! But my advice to writers and other creatives who are curious about the technology would be to have a go. Dip a toe in the waters. You might like what you find there. And if you don't, at least you'll know why you don't want to use it, rather than just following the flock. Because you never know. You may just be following them into Foxy Loxy's den.