Forecasts and delivery

A couple of months ago I made a few short videos. Two made it to Vimeo – I thought I’d uploaded a third, but it was all kinds of corrupted – and I sort of forgot about them.

They’re sketches of something I wanted to build; a script that would pull a given friend’s social media output for a week, stick it in a corpus with a recent MetOffice shipping forecast, and generate a Markov ‘forecast’.

It’s a spiritual companion to Russell’s Science Story Magic. I enjoy that as a vision of what something like Twitter might become – probably already is for people with screenreaders – but I look at the automation of services and I wonder if we might end up with human surplus leaking out in things like this instead. Humanising the algos.

A near future where we commonly apply emotion to artificial things seems more likely, to me, than one where we’re comfortable with the mechanisation of emotional things.

And my primary reference for that is the shipping forecast. It’s a machine generated output. Raw data, rendered emotional and poetical in the retelling.

Process wise, these were about as clumsy as you could get. The camera was set to record something for a few minutes, the background sounds slowed or removed entirely. I lobbed hastily cut-and-paste versions of several friends tweets and the day’s shipping forecast into Cheney, a browser-based Markov generator that Armitage threw together. I then toggled the settings until something useful was spat out. I used the dictation mic I have to record those as a voiceover and exported the results – badly.

Oh, and something half-remembered – I wish I had the photo to hand – but about five years ago I found a book in a phone box in New York that contained table after table of out-of-date statistical information about the price of various minerals or metals.

The book was thick and the title richly embossed, and picking it up it carried a Biblical weight. You could preach from it, and I did. The delivery changed utterly the words inside (see also the Lee-Morgan Method).

I left it, a month later, in a flat in Montreal.