I was experimenting this week.
So the idea was to use an AI to write a review without any need to listen to the album.
It was very easy. The results after a few prompts to modify the initial instructions proved fairly decent. I didn’t publish the results I hasten to add.
I happened to know the record chosen as a test and rough facts about it so could cross check.
Here’s the rub, though.
While perfectly decent – and I have read lots of much more poorly written published reviews in newspapers and magazines by contrast – the main issues are two fold.
The first is the reviews don’t actually say anything beyond the facts. There is no use of non standard language either unless you customise it to do this. We are not looking for our writers to be the next Sally Rooney or Colm Tóibín.
But we are looking for knowledge, taste and a sound grasp of what is decent and what isn’t. A few typos or crass grammatical grasp is allowable but think you know jazz and big up the likes of Dave Koz over Immanuel Wilkins? Then Houston, we have a problem. As a reader you need a decent guide.
But that and secondly is the scarily ingenious bit: see an inadequacy in a piece of writing? Specify what it is in the prompt and ask the AI to churn out a new version erasing that issue. Ask the AI to respond to the record as Leonard Feather would. As John Fordham. As Nate Chinen? Rewrite your favourite Eurojazz critic whose first language isn’t English. Iron out all the irritating balderdash pretentious snobbery that litters jazz crit. Do a bit of Ozempic like slimming down of all that tedious build up when the reviewer decides to review his journey to the gig on his trusty tandem sharing the journey with Fred from Wigan and fancy that the handlebar came off and they both had to hop on a bus and missed the first 20 minutes when by then the Intrepid Tootlers came on stage and they missed – according to some bloke holding a tankard – the IT crowd’s “moving” version of ‘How much is that doggie in the window?’ Got it, my friend.
The second is disclosure: going forward – especially when media are already using AI reporters as a brand thing you will just have to assume AI has been used as clearly few are ‘fessing up that they are using it.
But the trust issue applies. Just because AI has been used doesn’t mean the results are distrustworthy. The tabloid press and even the ever more sensationalist quality press has often been distrustworthy or just plain wrong long before the advent of AI and we have always had jazz journalists pre-AI whose opinions are not that valid, interesting or indeed trustworthy enough.
On this blog I am not going to suddenly use AI to generate reviews. If I did I could probably triple the amount of reviews on the site on a weekly basis without spending any more time if I did.
Real writers will however be less needed going forward because it is so much cheaper and quicker for publishers to use AI.
I think right now that I can make a good guess if something was written using AI. And there isn’t always time to run it through checker programs that will tell you it is or isn’t so guessing needs to be done.
But that hunch may not be possible in a year or two as the quality of the output continues to vastly improve. At that stage the jig is up.
Writing will then be the art of the prompt. It probably already is in some quarters. The much mocked preposterous ability to copy and paste looks golden agey naive and quite sweet by contrast.
Opinion in that dread scenario will not be to concentrate on honest subjectivity garnished with a decent bit of detail but instead a heist of algorithmic fabrication and an all day brunch of not always appropriate or contextualised facts at worst, the triangulation of desperately Clouseau like prompts cynically arrived at that betray the unknowable intent of the prompt writer at best.
Know your writers, trust them. The independent mind is more important than ever. Setting the bar lower is fine – offline, someone writing on the back of an envelope, scribbling out a review without cribbing things from the Internet I mean and pretending they are sage-like and completely omniscient – when technology is a barrier to the truth arrived at through the smoke and mirrors of slop. Can you polish a turd? It seems you can in this brave new world.
Recently on Marlbank: