Data doesn’t just mean line-ups, instruments and when the album was recorded
We live in an era when it’s never been easier for ANYONE to write a review of something.
Try it my friend.
Pick an album that has just come out. Open an AI. Prompt it to “write a review”.
Read what it says.
Add more prompts until you are happy.
Check it.
The better the prompts, the better the review. It’s extraordinary.
It will be properly punctuated, spelt correctly, formatted as you wish. Wonderful.
You can ask the AI to write it “in the style of…” change its tone, its register.
Ask it to write with 3 stars in mind as a judgement, 2 stars, no stars. Whatever you want, the obliging AI will help you.
AI vs Human
Is this ethical? It’s not illegal – but ethically very borderline. Best to say the review is written with AI but I don’t think this will happen at all much. Surely using AI for research is OK, to proof copy, examine grammar? Or is it?
Do you have to listen as well? To write the review with AI totally without listening, no!!! Because the AI isn’t listening, it’s generating text. This gets to the heart of the matter, what it is writing is ridiculous because it hasn’t listened.
How does the thing read? Not brilliantly although better than some hack writers.
How AI will actually improve the quality of reviewing
I guess as AI gets even better than it already is the quality of “the writing” will improve.
If an AI can “actively listen” then that will also enhance “the review” up to a point anyway.
It may well be able to analyse on a technical level in a way most reviewers who are not holders of doctorates in music or expert transcribers will be completely unable to.
So for instance if a review is able to add in elements of say transciption plus explanatory annotation for the non-musician to back up a point, basic analysis of the harmony used and a cross comparison with say the main influence – eg how a Coltranian saxophonist differs from Coltrane himself – then that is a massive improvement on what reviewers can do now.
An AI might be able to provide more facts and hard data then which is incredibly useful.
What it might also be able to do is to supplement this by perhaps summarising critical reaction in a few paragraphs which would be useful for readers as background. That’s more on the consumer side – review as aid for purchasing.
I suppose then the user of the AI, the jazz writer or journalist, can then scrutinise all this and add in their own emotional reaction and moderate their opinion in terms of their frame of reference built up via such criteria as they choose – the other albums they have reviewed over a year or two and work out via some sort of tabulation if the album they are listening to is better or worse than other work that can sanely be compared against.
At the moment this kind of deep analytical work on reviewing isn’t really done. Lists of albums of the year are the closest to this, but these lists are really just for fun and aren’t scientific even when magazines and other media award prizes or titles for their decisions.
An AI can’t make you love a record even when you know everything about it and can discern cool things about it.
The value of its reviewing might be of appeal to people however who want hard facts more and particular kinds of facts at that.
Feedback that’s far more useful than ever before is possible
Critical reaction is like feedback.
On a basic level, it’s whether what’s heard is any good or not. The opinion is valued because the critic has taste and is not a paid employee of the musician like a PR or manager or promoter.
Different people want to know different things.
If you are a festival programmer you want to know if the album gives enough clues to translate to pleasing a thousand people in a concert venue when music from it is played; if you are a record label you want a review to help you decide whether a certain kind of listener can be persuaded to buy the record in quantity.
At the moment most reviews won’t be able to let you know for certain if the reviewer is typical of the demographic the label boss is looking to recruit as a purchaser.
Again that may change: so if the review was able to explicity map out how and why – based on running analysis with a comparison – such and such release would have exact appeal for someone who say buys alto saxophone led records issued by Blue Note in the last 10 years that would be extremely useful. It might be boring for a reader though.
For the reader however I think it would be incredibly useful and interesting too if a reviewer profile is included with a review. I don’t mean biog – I mean what the reviewer is into, what their likes or dislikes are based on previous published reviews. A reader can then offset all this when consuming a review. It makes the reviewer work harder.
At the moment too many reviews are like diary entries, too brief, and too much like a laundry list of cliches and wannabe or surrogate literary or cultural criticism that shows up at the wrong address and reviews the music with parameters that don’t always square with either observed process or achieved outcome. AI might change that and reviews could end up as specified above a lot more valuable if writers respond to just what is possible to analyse and to work out just what a reader might want.
MORE FROM MARLBANK
