Discovered, Not Designed
Concerned with the difference between complicated (causal, explicit and deterministically) versus complex (emergent, implicit, non-deterministic) designed objects and processes.
“Systems naturally evolve towards states with the greatest number of accessible microstates.” -> Analogous to over-parameterized deep neural networks (double descent phenomenon). The over-parameterization helps gradient descent find local minima.
Abstraction through emergent grouping through useful analogies. Complex systems are not to be deliberately designed, but discovered in an externally guided search process. Complex systems are “alive” and can adapt to changing environments. Complicated systems are brittle since they are rule based deterministic machines.
Knowledge expansion vs knowledge convergence: Science has drifted into the deep branches (details) of the fractal knowledge tree and lost focus to uncover the root of all that is (in nature). Uncover hidden abstractions to create truly novel understanding of nature.
Let the structure emerge, do not attempt to follow a structure to get there. Goals can be fine, but don’t get too caught up in those; even largely random behavior can lead to patterns. Knowledge is being able to validate what emerges.
The book is available here by Sean McClure