On randomness

A while ago, I received the second issue of Holo Magazine - still my favourite publication on everything in between art and technology - and was happy to find the theme being something which has always fascinated me; mathematical randomness.

Holo 2, by CAN

Holo 2, by CAN

Now, you might say, what do I care about randomness? Numbers can already be pretty boring in themselves, and even more so if you present someone with numbers that have absolutely no pattern in them at all. Now, turns out - and now I'm quoting the magazine's main article - series of random numbers are vital for all sorts of modern science, and not in the least for encryption techniques. So yes, your online banking, your PayPal account, your professional mail; all of them need randomness to protect your data. It's simple; contemporary algorithms have the power to find even the slightest hint of a pattern in gigabytes of encrypted data, so the trick is to base that encryption on something that has no pattern in it at all: randomness. And that might seem simpler than it looks; in fact, it is very hard to generate true randomness a.k.a. series of numbers that, even after extensive statistical analysis, exhibit no larger structures. Most randomness is in fact pseudo-randomness; it looks arbitrary, but after long analysis, you (or a computer) might still find some order in it. In fact, true randomness is so important, that the US military decided, in the midst of the cold war arms' race, to commit a whole research department for 2 years to finding series of random numbers, leading to the publication of A Million Random Digits with 100,000 Normal Deviates in 1955. A random page (pun intended) of this book looks something like this;

A Million Random Digits with 100,000 Normal Deviates,  Rand Corporation, 1955

A Million Random Digits with 100,000 Normal Deviates, Rand Corporation, 1955

Now, my personal fascination for randomness comes through architecture. You see, most architects (at least in Flanders) hate randomness. They can't deal with it. They want structure, order, reasons. I have very troubling memories of being a student, and presenting a facade design to a jury, telling them I've put some randomness in the window sizing... Not good. Oh boy, not good at all. So, by the time I went to study computational design, I had firmly given up on randomness. Order, that was what made architecture great!

Then, one night, I was programming an algorithm that would design a building for me. It basically consisted of several hundreds of digital designer-programs, all taking decisions by a very well-balanced, very orderly, very reasonably structured set of rules. Pure logic, no creative subjectivity whatsoever! But it didn't work... Nothing happened. Decisions were being taken, but the sum of these decisions always ended up to no solution ever meeting my requirements. The design process simply didn't start. In view of the pending deadline, you may imagine I was slightly stressing out. So, just as a troubleshooting test, not thinking it would actually lead to something, I added a bit of numerical randomness to the decision-taking. Not much, let's say 5% if all the other design criteria were 100%. The effect of this step boggled my mind; it effectively unblocked the algorithm, and design proposals came flowing out of my processor... In the end, I wrote my thesis around the algorithm, and passed my degree.

Just 5% of randomness...

Just 5% of randomness...

Even now, my own thesis students, who work on similar algorithms, still come to the same conclusion; design procedures need a certain degree of randomness, and the earlier in the process, the more of it they need. It's like the creative spark that opens up the deadlock of logic and reason, the starting point that every good design needs.

Now, I guess if my old teachers were to read this, they would find even more reason to never ever trust computers (nor randomness). But to me, they can reason all they like; personally, I've come to find randomness far more fascinating than reason.