Sometime in Spring 2021
So much natural data follows a power law distribution, it should really catch our attention when data does not follow an exponential or power line decay. For example in the brain activity is largely bimodal not exponential between categories
Information processing mechanisms in the brain interplay is over in a spectra frequencies, from the slow drifting genetic and anatomical changes, to faster synaptic modification, to the very fast spike frequency spectrum of activations. Cross frequency phase correlations exist through the frequency product domain
Let the neural network grow automatically. Let out dynamically alter the computation graph. For example
Test the growing layer flat net before using it in a more complex system.
Machine learning makes a unidirectional flow of structure from hardware to software to application. Biological systems have a bi-directional flow of information from anatomy to physiology to cognition
By uniting system I and II type thinking under the same architecture and not extracting system II thinking into a totally separate symbolic reasoning process, the AI will inherently be able to dynamically select an appropriate thought process from a spectrum of thought from type-I to type-II and not just either extreme. Since the second thought system emerges from the former, it naturally should possess the ability to reflect on its own thinking and perform metacognition.
exhibit a noncategorical range of cognitive processes in between