Thursday, 27 July 2017

Clutter

So I'm picking through the factor 15 and the inappropriate shorts and the lengthy novels in translation, all the crap one has to take on holiday, and I think to myself, well, this is a time of departures and interventions, as good a time as any to check the bookmarks in my browser and throw out the stuff that I meant to find a use for but never, for whatever reason, did. 

I want that fresh, unencumbered Glade feeling, free trom things like Drip-Free Wine Bottle Makes Us More Grateful For Science Than Ever Before, one of those headlines which draws you up short at the same time as it leaves you with no inclination to find out more (a physicist at Brandeis University, as it happens, who's inserted a two-millimetre anti-drip groove in the neck of the bottle, who'd have thought?) while reminding you at the same time that this story fits somewhere into the larger pantheon of non-spill, non-drip, non-spatter, non-stain, non-marking, non-smear, non-streak products whose existence makes up at least a third of the internet. 'As yet,' concludes the HuffPo, one of several sites to cover the story, 'there's no news on whether the product will be adopted by wine makers'.

Similarly, How Women Are Changing Champagne sounds appealing, but adds pretty much nothing to the compendium of women-changing-the-face-of-wine stories which have been floating around for, what? A decade? A century? 'The rise of women winemakers will certainly change champagne,' apparently, 'though exactly how is yet to be seen.' Why did I think there was something worthwhile in this? I mean it's always good to celebrate the increasing importance that women play in industries traditionally dominated by men, but I can't help thinking that I allowed myself to be suckered in by some kind of human interest story that wasn't really there. Or worse, that I let myself be beguiled by the idea that something might be remarkable simply because it's done by a woman.

Rather as I allowed myself to be slurped into a micro-story involving some very rich guy called T. J. Rodgers (who also cropped up here, as well as lots of other places), the main eye-catcher being, in all honesty, the word billionaire, rather than the fancy tech he's using to make an impossibly perfect Pinot Noir. What do I really care that he has a mathematical formula to deal with every element of production, including 'root density, siphon run-offs, wine press effectiveness'? I don't even know what these terms mean. Out it goes, along with an Andrew Jefford threnody concerning the impossibility of wine writing ever positing a true equivalent to literary writing ('supportive intimacy' is a great phrase, though) and this chestnut - from the Evening Standard - about how drinking wine may be good for your brain. What's the ratio of wine-is-good-for-you stories to wine-is-bad? About two to one in favour of wine-is-bad? Someone, perhaps at Brandeis University, will know.

And so it goes on, until only two bookmarks remain; but these I keep. Both are cognate, in that they fool around with the idea of artificially-generated word formation - a little niche of AI which will eventually put all writers and journalists and, indeed, Sediment, out of business. The first is the legendary Brooklyn Bar Menu Generator, randomly creating on-trend menus for imaginary hipster eateries (Pan-seared water as a starter from your local Gerritson and Stockon, for instance). The second is this excellent blog from a young California scientist, who uses some rather deeper coding to create names from scratch. Paint colours is a particularly fruitful area (Sandbork, Flumfy Gray, Nungle, Shy Bather, Parp Green, Breedly Burf, to name but six) but she's also done action figures, heavy metal bands, bad recipe ideas and terrible Broadway musicals (my pick of the last? The Wither Bean, followed by The Burking Ding of 190 Bour Dadige, a comedy).

You can see where this is going. Let the neural network loose on wines; let AI come up with some really Twenty-First Century drinks. Dr Janelle Shane, the person responsible for the latter website, actually invites suggestions for her next spree - the only problem being (for me, at any rate) that she needs a plaintext dataset of 1000+ existing names for the AI to use as self-training material. But wine! It demands to be done! Does anyone happen to have a plaintext dataset of 1000+ existing names? How hard can it be to get one? Will it be waiting for me when I get back? 

CJ

http://amzn.to/2tlJ9dy







3 comments:

  1. I have an idea of how this data can be obtained. Will attempt and report back.

    ReplyDelete
  2. OK, that would have cost me £350. Will try something else.

    ReplyDelete
    Replies
    1. I know CJ will be impressed by your endeavours when he gets back.

      And Miquel Hudin Tweeted: 'Well, this is just a brilliant way to freshen up my July brooklynbarmenus.com '

      Glad you're all having fun…

      PK

      Delete