Before we dive into the inner workings of how I conceptualize energy systems in nature and their juxtaposition to transformer based models, we must first explore the unit of input and output to nature's model: Energy.

Energy is a core building block of the natural world. In its most simplistic definition, energy is motion. Einstein's famous equation elegantly depicts this. E = mc^2. Energy equals mass multiplied by speed. This includes all forms of movement within the natural world. Sound waves, electricity, and heat are all measures of movement at the atomic level and they are also measures of different forms of energy. These different forms of energy contain different degrees of entropy. Thermal energy is a form of random motion while sound waves move atoms in organized patterns.

The Large Energy Model (LEM)

The natural world is a closed system of components that are constantly transforming energy between different forms in an attempt to reach equilibrium. Every living thing and biological system is a component in the model. Each biological system is a "weight" in the larger model.

When you listen to someone speak, process the sounds in your brain, and respond with your own words, you are transforming one form of energy (the sound waves from that person's voice) into another (the sound waves from yours).

Your ears are also picking up all the sounds happening around you at the same time that they are listening to that person speak in front of you. You may rack your brain and organize your thoughts to communicate back concisely. You may store some of what that person said as memories.

You are performing key functions as a component of the energy model. Let's explore the workings of the "component" in more detail.

Entropy Metabolism System (EMS)

I refer to the "components" that process and transform energy throughout nature as Entropy Metabolism Systems (EMSs): combinations of deterministic and stochastic processes working together in predictable steps to process the entropy of nature's energy.

EMSs follow this pattern:

1. entropic input ->

2. entropic consolidation ->

3. input extraction ->

4. deterministic processing ->

5. outputs and byproducts

   • energy storage (low entropy)

   • controlled output (mild entropy)

   • waste output (high entropy)

Nested Nature

It's important to note that the EMS is a nested recursive system. The LEM itself is also an EMS made up of smaller EMS components.

When the earth takes in the rays of the sun and uses it to create all forms of life, it is transforming one form of movement (energetic rays from the sun) into another (all the living animals, plants and micro-organisms). The Earth is an Entropy Metabolism System.

Cyclical Nature

The system is a cycle because culminations of entropic outputs are used again as the inputs to other biological systems. This is how nature morphs and evolves over time. It is constantly reacting to its own inputs and outputs and continuing to adjust its systems based on those outputs. We have

Outputs and Byproducts

Low entropy states are how nature stores energy for later uses (e.g. action potentials in neurons). Mild-to-high entropy outputs are how nature disperses energy in controlled ways (e.g. fire or fecal matter).

The Purpose of Stochastic Steps

Stochastic systems are needed at the expansion and consolidation points in EMSs. Entropic consolidation steps in EMSs heavily rely on stochastic steps because the transformation is not 1:1.

For example evaporated water can be consolidated into clouds in our atmosphere. However, the formation of those clouds is heavily stochastic. No two clouds are the same and they do not follow a set pattern in forming their shape outside of the major formation patterns. Two nimbus clouds can follow a general pattern but have widely different shapes.

When rain falls, the consolidated cloud goes through an energy expansion. Rain falls at random rates and locations. There is a general pattern but the process is stochastic. Hence why rainy weather is predicted in percentages.

The Purpose of Determinism in Nature

Events, evolutions, and phenomenon can sometimes appear to happen at random. However, nature is also filled with deterministic systems. The human body is one of them; snowflakes are another. Nature likes order as much as it likes entropy and the two balance each other out.

Deterministic systems are needed at the intersection points of EMSs; when one form of energy gets transmuted to another. For example, combustion can cause uncontrollable fires. However the process itself is deterministic and follows a repeatable calculation. If you have the right ingredients and conditions, you can always start a fire.

Equilibrium & Destruction

Every EMS in nature is constantly adjusting in an attempt to reach equilibrium. In an EMSs perfect form, it has zero waste and stores zero extra energy. It becomes fully deterministic. The amount of energy in its controlled outputs are comparable to or equal to the amount of energy to its inputs.

This state is dangerous. It immediately leads to decay. This is because no EMS exists in isolation. As soon as it reaches equilibrium. The entropy being created outside of it's system is greater than that being exported. This is because its outputs exist in a single form, while it's inputs exist in an array of forms. As soon as an EMS reaches this state, the entropy outside of it will begin to decay the system. Leading to its destruction and fall back into entropy.

"Energy of the world is constant, entropy tends towards a maximum." - Rudolf Clausius (coiner of the term entropy)

However, the destruction of the old system and its energetic outputs are then used to create new systems. Starting the cycle to reach equilibrium over again.

Surviving Decay

In order for an EMS in equilibrium to survive, it must constantly grow and learn to process new inputs outside of its immediate environment to handle the onslaught of entropy it faces.

This is essentially what frontier models are experiencing now: model decay. Because the outputs of finished LLMs are beginning to outweigh the inputs of training them. AI labs have already ingested most of the internet and now a lot new internet content is created by those same LLMs that were created from training.

The training process and development of LLMs will stall and decay unless new inputs are found or new processes are attached to it to grow the system.