What drives complexity?
When the Universe began, it was the uniform ball of Big Bang fire. And yet, for reasons still not well explained, billions of years of natural, spontaneous evolution somehow led to emergence of very complex solar systems, living organisms, weather patterns, financial markets… All of them are highly organised, very different entities where complexity is understood as amount of information necessary to describe them. Why is it happening?
By now we have accumulated a lot of knowledge about how some of these objects function, but we still have not agreed on that “why”…
I suggest that the factor driving the complexity growth is information, more precisely — the Shannon information entropy. Before we dive into the proposed explanation of the ‘why’, we need to have a look at the already broadly agreed ‘how’. It comes from the tenets of the classical thermodynamics. Inevitably, when a complex system grows it requires a decrease in local entropy, which is at odds with the second thermodynamics law, requiring entropy global growth. That alone should effectively block formation of any complexity, and this creates a paradox, a contradiction between the well-established facts. There have been numerous attempts to remove this contradiction, e.g. Schrödinger famously introduced the concept of negative entropy (“negentropy”, as coined by Brillouin) as essential for living systems. This term lacks the precision of a strict scientific definition; however, it introduces an intuitive concept that resonates well with observations. Prigogine demonstrated how far-from-equilibrium systems can spontaneously organize, sustained by throughput and entropy export. However, he did not offer a single variational principle potentially driving it.
In 1961 Landauer linked the thermodynamic entropy with the information, Shannon, entropy. The Landauer limit defines the minimum energy necessary to erase one bit of information encoded in a physical microstate. The physical world of energy and the world of information have been linked. By the way, interestingly, only information erasing inevitably costs energy, its creation does not carry such a theoretical limit…
Let’s try a thought experiment — imagine a thermal, undersea vent full of various simple molecules. They undergo random chemical reactions using the heat provided by the hot environment. More complex structures form and break up continuously — until, by chance, a stable and efficient autocatalytic replicator system appears on the stage. In the language of information entropy, its creation is an event of ‘high surprisal’ (low probability), creating a structure carrying large self-information and costing free energy. However, its preservation (maintenance) requires surprisal minimization (homeostasis) which needs pruning, incurring the Landauer cost.
After a while, the system replicates exponentially and the copies export more entropy, allowing them to maintain the high level of internal complexity. Energy dissipation creates gradients that support information-processing structures. This stabilizes and amplifies rare fluctuations and may create a ratchet that biases the system towards accumulating information-rich structures. The resulting configurations enable faster pathways for entropy removal. In this sense, complexity emerges because information-rich structures are effective at dissipating energy — linking local order to global entropy increase in a way that is both physical and informational. A complex system is an entropy miner: it gathers the world’s gradients, shapes them into fleeting islands of order, and survives only by letting even more disorder spill behind it.
Get Krzysztof Maruszewski’s stories in your inbox
Join Medium for free to get updates from this writer.
Regions that achieve unusually high rates of entropy export therefore act as ‘complexity engines’: once they appear, they tend to amplify surrounding gradients and create further opportunities for structured, information-rich states to arise. In such conditions, the system not only maintains structure but can progressively build new layers of organization. And this seems to work for all kinds of complex systems (natural and man-made), across vastly different scales — from planets to living organisms to societies and their products. Is consciousness itself a side effect of it? And what of Artificial Intelligence? This process continues…
Fascinatingly, there is no new physics law behind it — rather inevitable consequences of how our Universe operates… What will happen when the Universe runs out of gradients to exploit?
More of my essays and how to reach me:
- https://orcid.org/0009-0002-7245-7993
- kmwikipediapl@gmail.com
Acknowledgment
The author acknowledges the assistance of ChatGPT (OpenAI), AI Claude (Anthropic), and Google AI systems in improving the manuscript.
Originally published on Zenodo: https://doi.org/10.5281/zenodo.18659415
