Georg, with whom I worked on co-writing The Station, asked me if I’d like to do another piece of co-writing with him, this time an opera libretto. I said yes (duh!). Last Friday, we had a three-way chat with Henry Vega, the composer, about Alan Turing, neural networks, science fiction, queer stuff, and all, for a sharp hour (Georg’s good like that with his one-hour meetings).

Today I spent a couple of hours (after some dipping of toes last night) in installing TensorFlow-Char-RNN, a “a character level language model using multilayer Recurrent Neural Network,” as made wildly lovable by Janelle Shane of Letting neural networks be weird. That involved installing TensorFlow. I went for the direct MacOS approach (after toying with either a Vagrant VM or Docker container) of the Virtualenv flavour. Plus Python 3. And pip. Dependencies. We have them.

A bit of faffing around, and out is spat a ‘Shakespeare’:

t ‘vkdwsa avf
neu irot rS
, mvuaeea giCsouo aed renat rs
;iiweszteseooiiWhe thrr l st !htt :hsre

I mean, I was expecting a single, long ‘aaaaaaaaaaaaaaaaaaaaaa’, so this was progress.

More faffing, fans to 6000rpm, CPU to 500%, and some short while later, ‘Shakespeare’!

Before we proceed any further,
Or each doth now foul branch with thy preser’d up
Young to devise me him;
But in my jewities rebeeve me to this,
Your soul than daggers and breeding
some abrother Arms
What will be pronound with a husband; he’s beauty much or a slaughter,
But I’ll wring my false find than how ill.

Nailed it.