The value of piles of scribbles

Hello, reader. Let's make a few assumptions about you. 

You keep a record of your life. It could be a long-standing Twitter account, notes from college classes, or a pile of fading restaurant receipts gradually accruing underneath your desk chair.

You're willing to take your own advice. You know your past experiences have value and you're going to learn from them, even if it means having to slog through a minefield of memories. You think it's hyperbolic to describe memories as a minefield but you understand the metaphor.

You don't mind keeping records and memories of your life around as long as they don't create friction. You don't need anything slowing you down. Life is prone to enough inertia already.

Why not take a walk into the past with me? Why not dig through some old notebooks and see what we find? There could be a wealth of forgotten advice in there.

Here's something courtesy of notes I took on a career-focused lecture last October:

My scribbles aren't making any high-level claims about true purposes or how to manage your time. It's quite a bit more specific: if you reply to a job post listing a set of desired skills, address those skills in your cover letter. It may seem obvious to do so but it's easy to dismiss a cover letter as something that no one will ever read in detail. In some cases, failure to specifically mention a full set of required skills in your application materials will get you filtered out of the applicant pool. This advice was courtesy of a Virginia state employee who knows the hiring process well. 

More material once I find it. It won't be difficult to locate. Some of it may be more useful than basic career advice.

Wikipedia-based poetry for June 12, 2015, plus neural networks

i retired from my downward lascerations on my
wrists to die." computers do most
of the altar. the soundtrack album
isbn 0-8247-7487-6.
basic form, the
test assumes that there
are no parameters to
be
over the
next series of
boats sylphe. on the money, high-powered
money,
jeff became close to the
player will
encounter
a
interaction that
tends to
change
the motion of an object. in

I'm thinking of upgrading the wikipedia-poetry script from a set of randomized Markov chains to a recurrent neural network. I'd prefer to see output that toes the line between interesting and coherent but the current approach leans too far toward the former. This blog post should be helpful - see the comments for a fun example using generated NSF grant abstracts.

Of course, the most obvious use of an RNN here would be to generate wikipedia entries. They could still be poetic if we cut and paste bits of them together at meaningful intervals. Perhaps the RNN could be taught to recognize the emotional value of certain words (i.e., the words die or money in the above example have some emotional impact, while jeff might offer comparatively little). 

Overthinking and underscanning

Here's a brief example of how overthinking can be counter-productive.

For no particularly useful reason, I read this StackExchange question today. The poster wanted to know how to produce an effect mimicking badly scanned medieval documents. It isn't a difficult thing to do with simple graphic editing software but there's a thin line between mimicry and the real deal. This, however, was my favorite answer:

Design your mark, then take it down to Kinkos. Find the crappiest photocopier, and make a copy. Then make a copy of the copy. Then maybe crumple/uncrumple the copy and make a copy of that. Continue until it looks the way you like it. 

It's a bit like that likely-apocryphal story about cosmonauts and writing implements. Sometimes, even in everyday life, a complicated solution might not be worth it, especially if the simple solution is practically trivial to implement.

The complicated solution, of course, may be easier to automate. There are always tradeoffs.

Spoiled for choice

If you're one of the handful of people who have read this blog in the past year or so, you may recall my attempts to listen to randomly-chosen genres on Spotify. Attempt probably isn't the correct word, of course: it's difficult to fail at listening to music. The script I was using to choose target media wasn't really cutting it, though. I needed more genres! I needed pre-defined playlists! More than anything else, I needed more experience with setting up simple GUIs.

I ended up with an exceedingly basic, TKinter-based interface for generating Spotify playlists within Echo Nest genres. Echo Nest is part of Spotify now, so their annotations are fairly comprehensive on Spotify tracks and artists. Spotify also discontinued apps last year but external access doesn't require much effort.

It looks like this:

Is there any other kind of post-rock?

The next step will be moving over to a fancier GUI framework. Pyjs, maybe. 

The program is beside the point, though. What is it telling me to listen to? Ignoring the test suggestion of Atmospheric Post-Rock, it told me to listen to:

Bluegrass.

I'm going to start with The Seldom Scene and go from there.

A related project on everynoise.com: the Spotify New Release Sorting Hat.

Yeast, drugs, and copies of copies

Read a Nature piece this morning about yeast-based synthesis of opiates.* It won't be long before morphine may be predominantly produced using yeast rather than the opium poppy. This is a significant feat for synthetic biology on its own, but it will be very interesting to see what the long-term effects are.

The authors discuss a few security concerns. No one really wants to see more illicit opiate trade in the world, especially if it's funding organized crime and the like. I became a bit worried at the suggestion that opiate-producing yeast strains should bear DNA watermarks in their genome. As whole-genome sequencing gets less expensive, it should be perfectly feasible to screen for multiple watermarks even if they differ slightly in sequence, but the easiest way to find said watermarks would be through an automated approach. It could look something like the Counterfeit Deterrence System present in Photoshop or other software designed to detect patterns hidden in banknotes. Sequencing isn't counterfeiting, though. I'm worried that a home-use sequencer could detect a DNA sequence specific to opiate-producing yeast strains, surreptitiously report the result to an enforcement agency of choice, and contribute to a new wave of privacy concerns at the very least. 

OK, so worried may be a strong word here. The War on Drugs just doesn't have a great history as it is. I hope that new developments in synthetic biology don't get added to its list of victims.

*I saw some complaints on Twitter about how the title of this article (or more problematically, the title of this PBS piece) implies that home brewing of opiates is currently possible. No one has shown this can be done yet. It's certainly feasible, but yeast strains allowing for efficient home production of morphine just don't exist yet.