In an environment, can we predict the “averaged out” effects of what microscopic interactions will result in? This is a challenge in ecology, economics, and other complex systems.

The microscopic interactions of a neural network evolution are affected by the loss function. MAE, a simple one, is the average of absolute differences between the actual and the predicted value.

‣

Then, each neuron is updated via backprop according to the gradient of the loss function. Thus, the global “average” of the whole affects (goes back in feedback) to each individual part. Each individual is updated according to what the global system needs. The purpose of the global system is to optimize the loss function using each individual neuron, which in turn, will help each individual neuron “have a use”. This neural network system differs from a system of living beings because it does not care about optimizing the “health” (minimize suffering) of each individual being, which does not have a “health goal” of its own.

A system of living beings can each tune their beliefs, and thus decisions, in order to collectively have a “global objective function” which aims to minimize the suffering of each living being. However, in current systems on Earth, each living being seemingly does not update its beliefs in order to be in tune with this global objective function, but only their own immediate, local objective function. As shown in simulations such as Conway’s Game of Life, this produces emergent effects that result in “macroscopic collectives”. Each macroscopic collective (or component) can have its own objective function, which may be calculated based on its own composed components.

Systems such as a body have such a feedback system where the “macroscopic collectives” also help the more microscopic components. The brain makes decisions to help cells. But cells are also “sacrificed” to help the more macroscopic components. Societies also are built with these feedback systems in mind, and perhaps there may be solutions which do not require “sacrificing” microscopic components in a way that disregards their own local objectives. These systems are refined with “trial and error” selection through evolution, a messy process which does not always align its outcomes with what’s optimal for minimizing suffering, but propagates based on survival of the fittest.

Memes, like genes, may be another system of evolution that propagates on survival of the fittest. Memes spread through analogies. Thoughts may also spread through analogies, if we frame them as being composed of memes.

Also, remember that each component is how our own brains interpret (divide up and compress to an object) information according to its goals.