There are so many extraordinary things in life that we take for granted. Like how we can throw an object at a distant target without having to calculate the force, how our brains can focus on one voice after hearing our name called in a crowd of voices, or the mere existence of prime numbers. I can go on all day compiling a list of things that we take for granted, but by the end of it, the list would be as long as the line to buy tickets for a Justin Bieber concert in 2016. One thing I want to single out is randomness — it’s such a simple and amazing idea that I, for one, never truly considered up until recently.

Without the existence of randomness, there would be no human life because evolution requires that there are random mutations in the genetic code so that mutated species can out-compete for resources and be naturally selected to survive. No randomness means no evolution which means no humans; Earth without randomness would be desolate, the only thing present would be the most basic single-celled organisms. In biology, randomness is the most fundamental idea that allows for everything else to follow.

Randomness is ingrained in our reality. Over the years, computer scientists and mathematicians alike have worked tirelessly to come as close as possible to replicating perfect randomness; however, to perfectly replicate randomness on a modern computer is fundamentally impossible. John Von Neumann, the father of modern computing, put it simply:

“Anyone who considers arithmetical methods of producing random digits is, of course, in a state of sin.”

Anytime you use a random number generator it’s actually only producing pseudo-random numbers that follow a predictable pattern and this is because of the way computers are built [1] — they must follow deterministic paths. Even with our most advanced tools, we still cannot hope to replicate true randomness.

Moving to the artificial intelligence domain now, random search algorithms can compete with and, in some cases, outperform state-of-the-art reinforcement learning algorithms on designated tasks [2]. This means the best of the best algorithms, that some of the most gifted individuals have helped develop, can still be outperformed by something that is performing randomly.

Randomness is a simple yet extraordinary phenomenon that is integral to human life, has been around forever but is inimitable. It is the things like this that I marvel at because it makes me wonder how they can be programmed into our reality. Kind of like math being discovered and not created…

References:

[1] Skiena, S. , S. (2020). The Algorithm Design Manual. Springer.

[2] H. Mania, A. Guy, and B. Recht. Simple random search provides a competitive approach to reinforcement learning. arXiv preprint arXiv: arXiv:1803.07055, 2018.