Perceptlets: Key to Machine Learning

Current approaches to Artificial Intelligence and machine learning are limited by several intractable problems. Our research shows that eliminating the requirement to understand the data being processed will allow a different data representation created by relating sensor data elements to each other, spatially and temporally. In the course of developing this idea, a new and innovative data type, named “Perceptlets,” was discovered. Perceptlets have very unique properties that allow inexact matching of input and memory data as well as allowing for the creation of primitive procedures. Such procedures, after minimal training, let the machine learn without intervention. Mathematical models of the necessary components of such an indeterminate cognitive automaton were developed to illustrate plausibility.