Why the world is simple: from evolution to algorithms
A talk by Dr Ard Louis, University of Oxford
In the famous trope of monkey's typing at random on keyboards, the probability of typing a specific sequence of length L on a keyboard of N keys, is simply 1/N^L. If, however, the monkeys are typing into a computer programme, they might accidentally type, with probability 1/N^20, the 20-character code "print "01" 500 times" which will output an ordered sequence of length 1000 of the form 0101010.... In the first scenario, all outputs of length L are equally likely, but in this second algorithmic scenario, certain outputs with short codes are much more likely to appear than others. This intuition has been formalised in the coding theorem of algorithmic information theory (which should be much more widely taught to physicists). Outputs with short codes are called "simple" ones, and the coding theorem suggests that many processes in nature may be highly biased towards simple outputs. Evolution proceeds by random genetic mutations which then affect developmental processes which generate biological outputs called phenotypes. So, evolution proceeds by the second algorithmic scenario above. In this talk we will show that the coding theorem predicts that random mutations in evolution are exponentially more likely to generate simpler, more compressible outputs. This algorithmic picture implies, for example, a bias towards higher symmetry, since more symmetric structures can be described in more compact ways by repeating a simple unit. Evidence for this strong bias towards simplicity and symmetry has been found in protein clusters, the structure of RNA, and in gene regulatory structures. Interestingly, these same principles predict an Occam's razor like bias in machine learning methods such as deep neural networks as well, suggesting an intriguing link between theories of learning and biological evolution.
About the speaker
Ard Louis is professor of Theoretical Physics at the University of Oxford, where he leads an interdisciplinary group of researchers. Recent interests include the physics of biological evolution, self-assembly, and theories of why deep learning works so well. He was born in the Netherlands, but grew up in Gabon, central Africa. He did his first degree in Utrecht, and his PhD in solid state physics at Cornell University under Neil Ashcroft. Before coming to Oxford, he was a Royal Society University Research Fellow in Cambridge. He recently made the 4-part documentary Why Are We Here with David Malone and appeared in The Story of God with Morgan Freeman, giving him an Erdős–Bacon number of 6. His hobbies include molecular gastronomy and wine tasting.