Guiding Principles

This is an attempt to collect together the hidden guiding principles that I've been applying in trying to formulate my theories.

If it can operate without something, then it probably does

This is a special case of Occam's razor. On the face of it, a system that needs to model a particular behaviour should have a component that addresses that behaviour, or that behaviour may fail to emerge. But, in practice, it's often the case that a simple architecture that is generic enough can be bent in a way to produce unexpected behaviours.

This principle is based on many experiences in the scientific and even IT communities, that simpler solutions often lead to more powerful and adaptable outcomes. A possible explanation, in the context of AI, is that more complex systems can have more false starts, whereas a simpler system will need less training time.

Example:

We appear to experience the conclusions of decisions. Does that mean that we experience outputs, as well as inputs? However, it's also possible that the only experience of conclusions is via those conclusions circling back around as data that is passed into working-memory and then received as inputs. Based on the principle of 'if it can operate without something, then it probably does', it's likely that we do not directly experience any outputs.