indi.ca


Why Capitalism Keeps Falling Over

Left on its own, this computer program moved forward by just stacking itself really high and then falling over

When I was in middle school our woodshop teacher was unwittingly drafted into being our computer teacher as well. So he took us into this room with green monochrome computers to learn, among other things, touch typing. The way we learned this useful skill was a game. We played it so ‘well’ that Mr. Carter had to shut it down.

In this game you controlled a track-and-field runner, leaping over hurdles. To run faster you had to type faster, and if you typed wrong you fell. We tried and tried but we weren’t very good at touch typing, until someone figured out that actual skill didn’t matter at all.

‘E’ is by far the most common letter in the alphabet, used 11% of the time. Someone discovered that you could just hold down the ‘E’ key and run super fast. The character would stumble and fall but ultimately those ‘E’s would register and we all got high scores. Old Mr. Carter just grumbled. Woodshop was easier. The kids that tried to cut corners over a lathe just lost fingers.

Computer Programs That Fall Over

Even more than cunning middle-schoolers, computer programs are prone to taking shortcuts as well.

In a seminal work from 1994, Karl Sims evolved 3D virtual creatures that could discover walking, swimming, and jumping behaviors in simulated physical environments… However, when Sims initially attempted to evolve locomotion behaviors, things did not go smoothly
Figure 1. Exploiting potential energy to locomote. Evolution discovers that it is simpler to design tall creatures that fall strategically than it is to uncover active locomotion strategies. The left figure shows the creature at the start of a trial and the right figure shows snapshots of the figure over time falling and somersaulting to preserve forward momentum.

Sims hoped to evolve creatures that would walk, wriggle, or squirm, but the program—taking its goals quite literally—figured out that it was fastest to just get really tall and fall over.

The researcher Krcah found a similar effect when he tried to ‘evolve’ creatures that jumped. The program, taking him quite literally (as programs do), just built very tall structures that never leapt off the ground at all.

So Krcah changed the rules saying that height would be measured as how far the closest block to the ground got off the ground. But of course, as with any specific set of rules, this was quite specifically ‘hacked’. The program just evolved to get really tall and fall over again, flipping it’s lower parts in the air.

Falling over seems to be the default hack for a lot of algorithms. Including capitalism.

Figure 2. Exploiting potential energy to pole-vault. Evolution discovers that it is simpler to produce creatures that fall and invert than it is to craft a mechanism to actively jump.

Capitalism That Falls Over

Capitalism is really just a big computer program, an AI, and one of the inputs is the equivalent of dumb middle schoolers constantly trying to hack the thing. Investment bankers, stock traders, even computer programs sniffing out any loophole in the rules.

The 2008 Crash