Neutrinowatch

By Martin Zaltz Austwick and Jeff Emtman

Companion episode: Gluteal Strategies

Our idea with Gluteal Strategies was to make an episode quickly. “Quick” is a relative term, of course, seeing as our first five episodes took nearly a year to develop. But in this case, we gave ourselves a weekend to get the core of the coding done…and then we proceed to turn the rest of the episode’s development into a months-long hydra of tweaks and bug fixes.  And so it goes.

Jeff Emtman:

The core challenge of this show is a really tricky one.  In traditional (static) podcasting, you attempt to make a single, good, listenable thing.  But in generative podcasting, you attempt to make a set of creative rules that dynamically create many good, listenable things.  And for me, I’ve really struggled with calling something decent “good enough” and something a lil’ bit cacophonous “listenable enough”.  This enough-ness is a challenge for me in all art forms, but it feels especially prescient in projects like this one.  Each episode is a little machine that poops out a new content every day, without consulting us first to check that it’s listenable/sensical/entertaining. 

The idea for this episode was simple enough: we’d each write an algorithm that could provide advice to cure artistic ruts (writer’s block etc.).  We wanted to write some code to help us (and listeners) unclog the creative pipes, so to speak. 

We each wrote our own algorithm to achieve this.  I took a branching narrative approach, and Martin did something a bit more advanced. 

Martin Austwick:

Brian Eno and Peter Schmidt’s Oblique Strategies has been helping creative people stuck in a rut get unstuck for over 40 years. Ultimately, utility is in the eye of the beholder. Let’s take an example: “Ask your body”. Provocations like this are invitations to consider what is meant by the strategy, how it relates to your current work, and then, how to implement it.

I may have used the original text from a well-known set of, um, “indirect tactics”, and swapped out nouns, adjectives and adverbs randomly chosen from a(n out-of-copyright) Great American Novel with a, um, fishy, vibe. The approach is partly inspired by Tim Clare’s 100 day writing challenge – one of the exercises (light spoilers) is to take the opening sentence from your favourite book and swap out the nouns, verbs, adjectives and adverbs. I chose the opening sentence of Ursula K LeGuin’s A Wizard of Earthsea:

The island of Gont, a single mountain that lifts its peak a mile above the storm-racked Northeast Sea, is a land famous for wizards.

Which became…

• The home of David, a small shack that sinks its lintel a foot into the soft Gobi desert, is a house welcoming to travelers;
• The town of Tonopah, a mining community that holds its Stealth Bomber heritage above its weathered heads, is a place forgotten by The West;
• The Land Rover of Ian, a scuffed vehicle that points its nose embarrassedly into the corner of the Woolworths parking bay, is a car rich in history;

And so on. Computers are even less good at this than I am, especially if the part-of-speech tagging (that tells you whether a word is a noun, verb, etc) is a bit punchy, so sometimes the strategies you get are a little hard to parse.

Jeff:

My own code took a simpler approach.  As a kid, I was an admirer of quips, proverbs, fortune cookies, factoids — pretty much any kind of info-nugget that felt quotable.  My algorithm here reflects that. 

The code works via branching narratives.  I’ll conceptualize it as if we were a squirrels in a forest.  There’s a couple dozen starting phrases that the code has access to, including: 

• “You should”
• “Society must not”
• “You will suddenly run across”
• “Seriously consider”
• …

Each of these can be thought of as a tree trunk.  When you (a squirrel) chooses to run up a particular tree trunk (let’s use “Seriously consider” as an example), it has access to a number of branches it can pick from, including: 

• "tomorrow's challenges"
• "maternal love"
• “gymnastics”
• "corrupting influences"
• …

Once our squirrel’s on a branch (let’s use “gymnastics” as our example), it could decide to stop, being happy with the current output: “Seriously consider gymnastics”.  Or it could venture out onto a twig full of “lest” conditions: 

• "lest the apocalypse draws closer"
• "lest the skies rain thunder"
• "lest the birds sing their last song"
• …

Once a squirrel’s reached this point, it can go no further, and it becomes satisfied with the output of “Seriously consider gymnastics, lest the skies rain thunder” and this is returns this phrase back to the main program. 

This by itself is pretty simple coding, namely achieved by a handful of random choices in Python pointed at some JSON files. But I did add a bit of complexity by allowing our a squirrel to “jump” between trees/branches/twigs at certain points, and also allowing it to occasionally circle back on itself.  This gives the possibility of outputs that name a cadre of characters pulled from the list, so that our metaphorical squirrel sometimes generates things like “Seriously consider the lessons of a second grade teacher, and a family of ducks and a sweaty person with a horrible secret.”  But that’s very silly and I’ve taken some other steps to minimize silliness, as it could easily distract from other elements of the piece that I like, like the music. 

Martin:

Mary Epworth’s score for the first season of Within the Wires was a touchstone for me. I was a massive pain about the music, I think I’m more of a perfectionist than I imagine, and I got very touchy when I thought the fretless bass solo wasn’t getting the pedestal it deserved. Jeff did a great job here, and was hugely patient with my bullshit about the mix for a track that would appear in sixty second chunks, highly mangled, under speech.

Jeff:

I found the music element of this tricky. Martin tasked me with making something vaguely new-agey, and then he recorded some bass and guitar to mix into it.  Without going into minutia, I’ll just say that I haven’t tried to mix a lot of music.  And in attempting to mix this track, I repeatedly got my ass handed to me.  While I may make a living as a podcast mixer/editor, I was not prepared for the different set of considerations that music has.  With a lot of soul searching (and transient shaping), Martin and I came to a good place with this super compressed, cassette-y sounding track that, for me, really evokes the sound of the music that’d play on my childhood audiobooks right when the tape was due for a flip. Perfect.  Well, perfect enough. 

Martin:   

While this process wasn’t quick… I do think the final episode is pretty fun.