Reading this article I realized that entropy is actively eliminating systems that are not efficient enough at compensating (various non-living chemical structures), making it much more profitable to be one of the systems that can compensate against it (living cells). By overcoming this threshold of entropy, life has gained resiliency. Without the constant destructive effect of entropy, probably there could be no life. Life is both created and destroyed by entropy.

On a parallel note, deep learning neural nets rely on randomness in connectivity and order of examples, and sometimes just consumes random numbers and turn them into coherent images or sounds. Entropy/randomness is essential in Reinforcement Learning in the exploration-exploitation dilemma - we need to discover the space of possibilities, and in order to do that we need randomness. Interesting how chaos is tangent with intelligence and life, it is practically creating a space for them.