Rationalists Should Learn Lock Picking

To be clear, I’m not suggesting a widespread movement within rationalism to begin skilling up and stealing things to fund X-risk research. I promise.

What I am suggesting is that I found the first ~3-5 hours of learning to pick locks (which I did from sheer boredom) to be valuable in ways I hadn’t anticipated, mostly to do with forming models of things with very limited (and weird) information about them. It requires smushing your brain up against the territory in interesting ways; it’s motorcycle maintenance writ small.

Picking a lock is a pretty unique experience. A lock is made up of a cylinder inside another cylinder, with some ‘pins’ which are spring-loaded to enter the smaller cylinder and prevent it from rotating relative to the larger cylinder[1].

The pins are of different lengths, so the intended use is to open the lock with a key that pushes each of the pins to a height where they’re not blocking rotation, then apply tension to the lock so that it opens.

Because of a seemingly hard-to-fix fault in the process of making locks, the holes that the pins fall into are slightly different sizes. This means that if you apply rotational tension to an empty lock, only one of the pins is stopping the lock opening at any given time.

This also means that if, while still applying tension, you put some funny-shaped tool into a lock to push down gently on the pins one by one, the one which is currently blocking the lock feels subtly different because of the friction with the side of the hole. There’ll also be a subtle change[2] in the tensions (of the pins and of the cylinder) when that pin is in the right position; the cylinder will advance slightly. Rinse and repeat and voila, your lock is open.

Obviously it’s not quite this easy in (most) real locks. But it is surprisingly easy. Those already afraid of their house being broken into need not apply.


The real point, though, is that you’re trying to manipulate a machine which you only get information about through your hands, which you also have to use to manipulate the machine into doing what you want. It reminds me of the game Understand[3]; success is completely dependent on modelling the problem, actually noticing when (and how) your model is wrong, and trying to figure out why. It’s also very frustrating.

If you do feel inspired, you might want to try starting with a transparent training lock just to get a feel for the mechanism. Oh, and please don’t pick locks that are in use (I’m led to understand it might damage them in some way).

Let me know if you try it, and good luck!

  1. ^

    If you’re being picky, the pins have to be of two halves to make this system work (see diagram)

  2. ^

    And, if you’re in a movie, or very lucky, an audible click

  3. ^

    If anybody can find the LW post explaining & recommending the game please let me know. Believe it or not, words like “understand” “game” and “recommend” don’t make for a great search query