Tuesday, September 30, 2014

Links from September

I'm going to be starting to collect links and post them every month.  I've only been collecting these for a couple of weeks, so there'll probably be more next month.

Calvin and Muad'Dib, quotes from Dune as illustrated by Calvin and Hobbs.  It's pretty hilarious.

Guardians of the Galaxy as a short tabletop campaign.  I just discovered Max Gladstone's blog (thanks Brian) and I've been enjoying it.  He also wrote a book that was very good.  Oh, and apparently Gostbusters is the best comedy ever made about the limits of the Lovecraftian worldview.

I've always thought that augmented reality is a much cooler idea than virtual reality, and it looks like people are continuing to work on making the dream a reality.

India's Mar's mission arrived and is in a stable orbit.  This was impressive both because they managed to succeed on the first try unlike certain other space agencies and also because it cost of the movie Gravity.  I thought the Polar Satellite Launch Vehicle used to launch it was pretty nifty too.  For those of you who dont' play Kerbal Space Program, solid rockets are cheap and effective but once you light them they stay on at full power until they run out of fuel.  Liquid fuel rockets are trickier and more expensive, but you can throttle them and turn them off early.  Most space missions have used several liquid stages with maybe a few solid rocket boosters at the very beginning to get things going.  What the PSLV did, however, was to alternate solid and liquid stages, doing a lot of work with solid stages but still having the control of liquid stages in between to correct any problems with the trajectory that came up while they couldn't control the throttle.  A pretty clever idea.

They're making a Tetris movie?  And no, I don't mean the entertaining youtube Soviet history to the tune of Tetris.

Horribly Technical Links

Hopefully you already heard about this but there was a pretty bad bug in bash.  UNIX is very old and there's a lot of very impressive software that's been made for it, but this sort of thing makes you wonder if it would be better to just start over from scratch every few decades.

The Mill folks have new video out on their approach to pipelining loops.  It's very clever, as all their stuff is.

And speaking of computer architecture, here's pretty interesting paper on an idea for multipass execution as an alternative to the more expensive out of order execution we do nowadays to keep our computers running fast despite the continual risk they run of trying to grab some particular piece of memory and finding that they have to wait to fetch it all the way from RAM.

Monday, September 29, 2014

Red Plenty then a digression on motivation

I'd sort of been meaning to write about Red Plenty at some point, a book about the dream and reality of the Soviet economic planning system.  It was very well done and actually made me more sympathetic to the people who believed in Communism back in the day.   Well, it looks like Scott at Slate Star Codex has put together an excellent review that said everything I was going to say and more so just go read that.

One of the things I reconsidered after having read the book was the precise role of incentives in explaining the later problems with Soviet planning.

 In certain cases, Russians were very well-incentivized by things like “We will kill you unless you meet the production target”. Later, when the state became less murder-happy, the threat of death faded to threats of demotions, ruined careers, and transfer to backwater provinces. And there were equal incentives, in the form of promotion or transfer to a desirable location such as Moscow, for overperformance.

It wasn't that workers, managers, etc didn't have strong incentives to do their jobs - the problems were where those incentives were pointing.  If you want to know the details, well, you should go read Scott's review shouldn't you?

There's a certain danger economists sometimes fall prey too where they think of incentives purely in monetary terms.  There's a good reason for that, since many of the things they look at are more or less like that.  But there are circumstances where only looking at monetary incentives works and there are circumstances where it will leave you blind.

Everybody knows about Adam Smith's book The Wealth of Nations, but he also wrote The Theory of Moral Sentiments, and if he believed that "It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner," he certainly recognized that it is from the benevolence of our parents that we ate as children.  Forces more complex than exchange govern our relationships with our friends, our family, and to an extent our co-workers.

Few people would choose to do exactly the jobs they have if money wasn't a concern.  But given a job, a lot of what we do from day to day is governed by wanting to get along well with our coworkers and bosses rather than directly from threats of firing or such.  At least that's what life is like in engineering and my experience of the service industry; perhaps if I worked as a laborer where the boss didn't think of me as his equal and could see how many watzits I was stacking things would be different.

Arranging equivalent exchanges is challenging since everything has to be explicit and discovering what equivalent means in each circumstance can be taxing.  So when you're dealing with a small group it makes more sense and is maybe more satisfying to work together communally.  But as groups get bigger, organization by consensus gets harder and you need new methods.  As was once said on twitter "Market exchange is a pathetically inadequate substitute for love, but it scales better."

I can certainly see wanting to replace the impersonal forces of the market with something better.  Back in the day if you didn't like how your tribe was run you could just leave, and tribes were run by consensus without the hierarchy or impersonal forces of the modern world.  Wanting to get to that sort of communism in the modern world makes perfect sense to me as a goal even if I would despair of ever accomplishing it.

But in the end, Soviet communism needed hierarchical impersonal forces to organize everything just as capitalism does.  Their system just wasn't as good at it.

Saturday, September 20, 2014

Duck and cover, not so useless

I recently heard someone talk on Facebook about the old "duck and cover" drills that school children used to do in the cold war and how obviously that wouldn't protect you from a nuclear bomb.  I've actually heard that same thing several times, so I thought it would be good to chime in in support of the civil defense planners of yesteryear.  Duck and cover was actually a pretty reasonable way to reduce one's risk of dying in the event of a nuclear explosion.

To simplify a bit, there are basically four ways an atomic bomb can kill you.  When it goes off there's a flash of heat and light that can cause burns and fires.  There's a shock wave that can crush you directly, or which can collapse buildings and throw things into you.  There's a wave of ionizing radiation that can kill you through radiation poisoning.  And then there's any radioactive fallout that might kill you much later.

Someone who is standing directly under a big atomic explosion when it goes off is, obviously, going to be very dead.  But only the big ones.  Back in the day when Air Force was trying to convince people that nuclear anti-aircraft missiles were a good idea a group of five Air Force officer let them test one of those missiles directly over their heads.  It was just a relatively small 2 kiloton nuke, 1/8 the yield of the bomb dropped on Hiroshima, and pretty far above them so they didn't suffer anything worse than tinnitus.  According to Nukemap 2 kilotons means that the thermal effects were deadly out to .72 kilometers, the radiation out to .86 kilometers, and the blastwave could knock over buildings out to .89 kilometers, so since it exploded 5.6 kilometers over their heads they were relatively safe (provided nothing went wrong).

You can see that for the nuke those Air Force people survived the radiation killed at about the same distance as the thermals and shockwave did.  But that isn't true for all bombs.  For big bombs the range at which the immediate radiation is dangerous is tiny compared to the range at which you would be incinerated.  You can't really do anything to protect yourself from the radiation, so that's bad.  But for larger nukes the range of the shockwave grows much more quickly than the range of the radiation.  And unlike radiation, the blastwave only travels at the speed of sound so you have some time to take cover.  And there are actually useful ways you can take cover from an explosion.

Now, if you should survive the initial blast you still have to worry about the fallout.  Unlike a basic atomic bomb of the sort that was dropped on Japan or a hypothetical pure fusion weapon the Teller-Ulam devices that the US and Soviets were putting on their missiles release a lot of fallout.  Back when kids were being taught to duck and cover, however, the nuclear arsenals of the world hadn't risen to civilization destroying quantities yet.  Every sort of radioactive substance has a rate at which it decays, a half-life.  The shorter the half-life the more able that substance is to kill you by having lots of decay events that deliver lots of ionizing radiation to your body.  But the shorter the half-life the faster the radioactive substances go away.

Right after thermonuclear bomb goes off you do Not Want to be outside a fallout shelter.  But as the days or weeks pass the really nasty substances will go away, leaving merely the things with long half-lives which will drastically increase your chances of getting cancer but would you rather have a decade knocked off your lifespan or die right now?

So in summary, the idea that you should duck and cover after a nuclear attack wasn't crazy, at least when those videos were being made.  Nuclear war?  Totally crazy.  Duck and cove?  Not so much.

The Coming Interregnum after Moore's Law

An interregnum is a gap in governance, most commonly when a monarch dies without a child old enough to take over.  For decades the world has...