Scientists Propose a New Law of Nature for Evolving Systems

You are walking through a museum of natural history, past the meteorites and the minerals and the dinosaur bones. You look at a star, a trilobite fossil, and a computer chip. They seem like separate kingdoms of existence. Stars fuse hydrogen. Trilobites crawled the seafloor. Computer chips process information.

But what if they are not separate at all? What if they are all expressions of a single, underlying rule of the universe, a law of nature that has been hiding in plain sight?

A team of scientists from the Carnegie Institution for Science, the University of Colorado Boulder, and the University of Arizona has proposed exactly that. In a paper published in the Proceedings of the National Academy of Sciences, Michael L. Wong, Carol E. Cleland, Daniel Arend, and Stuart Bartlett argue that there is a missing piece in our understanding of physical law (Wong et al., 2023). They call it the "law of increasing functional information." It is an attempt to codify something we all sense but have never been able to name: the relentless drive of complex systems to become more complex, more functional, more themselves.

This is not a metaphor. This is a proposed law of nature, as fundamental as gravity or thermodynamics. And if it holds, it changes how we think about everything from the origin of life to the future of artificial intelligence.

The Apple and the Moon Problem

To understand what Wong and his colleagues are doing, you have to understand what a law of nature actually does. A law is not a description of one thing. It is a description of the equivalence between many things. Newton did not just describe how apples fall. He described how apples fall and how the moon orbits and how the tides rise, all with the same equation. That was the breakthrough. He saw that different phenomena were, at a deep level, the same phenomenon.

The authors of this new paper are looking for that kind of equivalence in the living, evolving world. They noticed that three very different kinds of systems stars, minerals, and life all seem to follow the same basic pattern. All three start with a huge number of possible configurations. All three have processes that generate many of those configurations. And all three have some kind of selection that picks configurations that work.

A star, for instance, is a giant ball of plasma. The number of ways that plasma can arrange itself is astronomical. Gravity and nuclear fusion act as selection forces, favoring configurations that are stable and long-lived. The result is a star that burns for billions of years. A mineral is a crystal lattice. The number of ways atoms can stack is vast. Temperature and pressure select the stable arrangements. The result is a quartz crystal that persists for eons. A living organism is a collection of molecules. The number of possible molecular arrangements is mind-boggling. Natural selection favors the ones that replicate and survive. The result is a bacterium that colonizes a hydrothermal vent.

The authors argue that these are not just analogies. They are examples of the same underlying process. In each case, the system is accumulating what they call "functional information." That is, information that helps the system persist, replicate, or adapt.

What Is Functional Information, Exactly?

This is the key concept, and it is worth unpacking carefully. Wong and his colleagues define functional information as the information in a system that contributes to its ability to perform a function. That function could be anything from "staying stable" to "reproducing" to "processing information."

The critical insight is that functional information is not just any information. A random string of letters has information. A shuffled deck of cards has information. But that information is not functional. It does not help the system do anything. Functional information is information that has been selected for a purpose.

Think of it this way. A pile of sand on a beach has a certain amount of information about its grain sizes and positions. But that information is mostly random. A sandcastle, on the other hand, has functional information. The grains are arranged in a specific way that allows the structure to stand. The information is not random. It has been selected by the builder for the function of "not collapsing."

The authors propose that the universe has a tendency to increase the amount of this kind of information in evolving systems. They state it as a formal law: "The functional information of a system will increase (i.e., the system will evolve) if many different configurations of the system undergo selection for one or more functions" (Wong et al., 2023).

This is a big claim. It means that evolution is not just a biological phenomenon. It is a cosmic phenomenon. Stars evolve. Minerals evolve. Maybe even ideas evolve. And they all do it according to the same rule.

The Three Drivers of Selection

The authors identify three universal concepts of selection that underpin this process. They call them "static persistence," "dynamic persistence," and "novelty generation."

Static persistence is the ability to simply stay the same. A stable mineral, a rock, a dead star. These systems resist change. They have functional information that allows them to persist in a static state.

Dynamic persistence is the ability to maintain a stable pattern of activity. A living cell, a star fusing hydrogen, a hurricane. These systems are not static, but they maintain a dynamic equilibrium. They have functional information that allows them to keep doing what they are doing, even as they exchange matter and energy with their environment.

Novelty generation is the ability to produce new configurations that might be even better at persisting. This is where evolution really takes off. A mutation in a gene, a new mineral structure, a new star formation pathway. These novelties are tested by selection. Some fail. Some succeed. The ones that succeed increase the functional information of the system.

The authors argue that these three drivers are universal. They apply to any system that has many components, many possible configurations, and some form of selection. This is a powerful framework. It suggests that the evolution of life on Earth is not a freak accident. It is an inevitable consequence of the laws of physics.

How the Study Was Done

This is a theoretical paper, not an experimental one. The authors did not run a lab experiment with evolving systems. Instead, they did something more ambitious. They reviewed a vast range of phenomena from stellar evolution to mineral formation to biological evolution and looked for common patterns.

They examined the literature on information theory, thermodynamics, and evolutionary biology. They analyzed the mathematical properties of functional information. They built a conceptual framework that could unify these disparate fields.

The methodology is essentially comparative and synthetic. The authors asked: "What do all evolving systems have in common?" and "Can we state that commonality as a formal law?" Their answer is the law of increasing functional information.

This is a legitimate and powerful way to do science. Many of the most important advances in physics have come from this kind of theoretical synthesis. Einstein's theory of relativity was not based on new experiments. It was based on a new way of thinking about existing experiments. Wong and his colleagues are attempting something similar for the study of evolving systems.

What This Means for the Origin of Life

One of the most exciting implications of this work is for the origin of life. The origin of life is a hard problem. How do non-living molecules become living cells? The standard answer is "by chance." But the numbers are daunting. The probability of a self-replicating molecule forming by random assembly is astronomically small.

The law of increasing functional information offers a different perspective. It suggests that the origin of life was not a single, improbable event. It was a gradual, inevitable process of increasing functional information. A mineral surface might have functional information that allows it to catalyze a reaction. That reaction might produce a molecule that has functional information that allows it to form a primitive membrane. That membrane might have functional information that allows it to concentrate other molecules. Each step increases the functional information. Each step is selected for by the environment.

Under this view, life is not a miracle. It is a natural consequence of the law of increasing functional information. Given enough time and enough configurations, systems will inevitably evolve toward greater complexity and functionality.

The authors do not claim to have solved the origin of life. But they offer a new framework for thinking about it. They suggest that the question is not "How did life begin?" but "How did functional information accumulate to the point where a system could be called alive?"

What This Means for Artificial Intelligence

The implications for AI are just as profound. Current AI systems are not evolving in the same way that biological systems evolve. They are designed by humans. But that is changing. Researchers are now building AI systems that can evolve. They use genetic algorithms, neural evolution, and other techniques to allow AI systems to change and adapt over time.

The law of increasing functional information suggests that these evolving AI systems will naturally tend to become more complex and more functional. They will accumulate functional information. They will develop new capabilities. They will find ways to persist, replicate, and generate novelty.

This is not a prediction that AI will become conscious. It is a prediction that AI will become more functional. It will get better at doing what it does. And it will do so according to the same law that drives the evolution of stars and minerals and life.

The authors do not discuss AI explicitly in their paper. But the logic is clear. If the law of increasing functional information is universal, then it applies to any system that meets the three criteria: many components, many configurations, and selection for function. Evolving AI systems meet those criteria.

What This Does Not Prove

It is important to be clear about what this paper does not claim. The authors are not saying that the law of increasing functional information is a replacement for the second law of thermodynamics. The second law says that entropy, or disorder, tends to increase. The proposed law says that functional information, or order, tends to increase. These are not contradictory. The second law applies to closed systems. The proposed law applies to open systems that are exchanging information with their environment.

The authors are also not saying that evolution always leads to progress. Evolution does not have a goal. It does not aim for intelligence or consciousness or anything else. It simply tends to increase functional information. That increase can lead to dead ends. It can lead to systems that are very good at persisting but very bad at adapting. It can lead to extinction.

The paper is a proposal, not a proof. The authors are putting forward a hypothesis. They are saying, "Here is a pattern we see in nature. Here is a law that might explain it. Let us test it." The next step is for other scientists to try to falsify it. To find cases where functional information does not increase. To find systems that evolve but do not follow the pattern. That is how science works.

The Big Picture

What makes this paper so compelling is not just the specific claim. It is the way it reframes the entire project of science. For centuries, physics has focused on the simple and the universal. The laws of motion, the laws of thermodynamics. These laws describe the behavior of gases and planets and pendulums. They are beautiful and powerful.

But they do not describe the messy, complex, evolving world of life and mind and society. That world has seemed separate. It has seemed to require its own set of rules. Biology has its own laws. Economics has its own laws. Computer science has its own laws.

Wong and his colleagues are suggesting that this separation is an illusion. The same underlying law drives the evolution of a star, a cell, a corporation, perhaps even a galaxy. The law of increasing functional information is a candidate for a truly universal law of nature.

If it is correct, it means that the universe is not just a collection of particles obeying simple rules. It is a system that is constantly generating new forms of order. It is a system that is, in a very real sense, learning. It is discovering new ways to persist, new ways to replicate, new ways to be.

What This Actually Means

• ▸If you are studying the origin of life, stop looking for a single "magic" molecule. Start looking for the gradual accumulation of functional information in prebiotic systems. The question is not "how did life begin?" but "how did functional information cross a threshold?"

• ▸If you are building AI, consider that your systems may evolve in ways you cannot predict. The law of increasing functional information suggests that evolving AI will naturally become more capable. You are not just writing code. You are setting the initial conditions for an evolutionary process.

• ▸If you are thinking about the future of the planet, recognize that human civilization is an evolving system. It is accumulating functional information at an astonishing rate. That is a source of power and a source of risk. The same law that drives innovation also drives arms races and ecological collapse.

• ▸If you are a scientist in any field, ask yourself: "What are the many configurations in my system? What selects for function? How is functional information increasing?" The framework of Wong and his colleagues can be applied to economics, linguistics, geology, and cosmology.

• ▸If you are a curious person, the takeaway is simple and profound. The universe is not a clockwork. It is a garden. Things grow. They become more complex. They become more functional. And they do so because that is what the laws of nature demand.