The relationship between diversity, generalization and co-evolution is indeed intriguing. 
I have always been troubled with understanding the relationship between entropy and complexity. They are not the same thing, yet many formulations attempt to define one in terms of the other.
Complexity arises out of diversity, generalization and co-evolution. The latter two distinguish themselves from entropy in that both generate an implicit order.
Implicit order is not entropy. It only appears as entropy when the observer is ignorant of the underlying order. The undecideability in computation is not disorder, it is generated by order but we are unable to decipher this order.
There are at least two definitions of entropy, one is disorder as a consequence of diversity (an objective measure) and the other is a measure of ignorance (a subjective measure).
There are many more definitions of how to measure complexity. Some are related to entropy, others are distinct from entropy. There is always a need to conjure up a scalar to quantify an idea.
It is analogous to how we conjure up IQ as a scalar to quantify intelligence. But intelligence is a multi-faceted thing and one would expect the same for complexity.
Intuitively, there is a relationship between intelligence and complexity. There is also a relationship between what it means to understand and complexity. We can take a simple definition of intelligence as the 'capability of understanding'.
One cannot define intelligence without a definition of understanding. Coincidentally, entropy is a measure of the absence of understanding.
A lot of AI benchmarking involves the discovery of understanding (otherwise known as learning) of a narrow task. We must however seek a definition of understanding that involves a multitude of dimensions. This set is what we might characterize as 'general understanding'.
What are the characteristics of 'general understanding'. The first is that it is grounded in the innate umwelt (wirkwelt and merkwelt) of an individual.
The degree of competency within the umwelt (interpolation), the degree to which it can expand its umwelt (extrapolation) and the creativity of this expansion (ingenuity) defines 'general intelligence'.
These 3 capabilities of expanding an organism's umwelt is captured here:
AI has already shown how interpolation (i.e. induction), extrapolation (i.e. deduction) can be automated. What it yet has shown is how to do abduction.
It's actually more than this though. At a higher-level, human general intelligence involves symmetry breaking, analogy making and hypothesis generation:
That are driven by human intrinsic motivations that lead to learning:
Thus, it is in our intrinsic nature to seek creativity. It is our intrinsic nature to generate complexity.
Complexity, understanding, intelligence and motivation are thus unified under a single underlying theory. gum.co/empathy
@threadreaderapp unroll
• • •
Missing some Tweet in this thread? You can try to
force a refresh
