By Steve Donoghue
A publisher friend of mine many decades ago used to fetch up in my book room every week for wine and calzones and a good deal of complaining. He’d always apologize ahead of time – “You don’t need to hear my woes” – and then launch into some first-class kvetching for the next few hours, and it was wonderful.
I’d hear all about the ins and outs of the business, back when the margins were much smaller and the interactions with authors were much more direct and personal. My publisher friend sometimes bailed out his authors, sometimes picked them up off barroom floors, often floated them some much-needed cash against their royalties, and called most of ‘his’ authors friends. But oh, how he complained about them! And his complaints almost always came down to money.
Time and again, he’d put one of his flabby hands on the domed head of one of my beagles and sigh, “I wish YOU could write spy thrillers! You’d stay loyal, and you’d come cheap.”
I recalled those moments this week while I was seeing the latest news about AI – artificial intelligence.
That term always sounds like pure science fiction, but of course it’s been an integral part of all our lives for a decade or more. We all use Google’s search function; we follow the recommendations of the mysterious algorithms on sites like Amazon or YouTube; we rely on in-home surveillance devices like Alexa, or in-car surveillance devices using GPS; when we’re traveling, we increasingly consult translation apps on our phones. All of these things are AI, and the extent is far wider than that; your latest gas bill or meter reading – to say nothing of your gallbladder diagnosis – are very likely generated by AI.
But most of those applications are one kind of AI: you feed it many separate pieces of data, and it gives you what amounts to a calculation. If you gave that same data to a different AI, you’d get the same calculation, and if you yourself assembled that data (and kept alert, and made no mistakes), you’d probably get the same calculation.
“Generative” AI is different. You can see hints of it in Alexa or Siri, but its current outgrowths are much more advanced than a database recognizing spoken key words. Tools like ChatGPT take all their pieces of data and make something new with it.
It was only around a decade ago that machine learning algorithms became so sophisticated that suddenly the “something new” was in many cases startlingly, almost alarmingly new. Suddenly AI programs like ChatGPT weren’t just assembling data – they were creating prose and images at a level of skill often indistinguishable from what a creative human might produce. The two differences were price and time. If you give a human a commission to write a horror novel set in 1960s San Francisco, that human will agonize and procrastinate and get you a very rough first draft in about a year – and expect a few thousand dollars for their trouble. If you give the same commission to a generative AI, it’ll get you a more-or-less finished novel in five minutes – and it won’t ask for a penny, because it’s not a person, it’s a tool. Likewise if you’re a publisher and you need a book cover for a new fantasy novel, you can either send your ideas to your art department, get their first responses, send your appraisals, wait until they all troop back in from lunch, get their second responses, and eventually, next Christmas, get from them three or four possibilities, all of them ugly. Or you can give the same ideas to a generative art AI and get back three or four dozen possibilities, all beautiful, in about five minutes – no “just a reminder” internal emails necessary.
As you can guess, this kind of generative AI has been the technological advance that launched a thousand think-pieces.
Some of the reasons are obvious, yes? If you’re a publisher looking to market a good old-fashioned locked-room murder, you’d be crazy not to be tempted by simply telling an AI to scan 10,000 earlier examples and produce a new one for you (traditionally, publishers have, in fact, been crazy – my old friend was – but distressingly few of them are anymore), instead of dealing with a needy, sloppy, whining flesh-and-blood author. You might eventually opt for that flesh-and-blood author, but you’ll be tempted (if nothing else, by the prospect of reaping 100% profits). Likewise, if you’re an avid mystery reader and you’re in the mood for a good old-fashioned locked-room murder, you might not care all that much whether or not your author has a pulse – especially if the price was right.
It feels like a revolution of some kind, but it’s difficult to tell what kind of revolution.