Deriving Unsupervised Fluid Intelligence from Schema-Less Polymorphic Unstructured Data

Overview

This study investigates the fluid intelligent capabilities of a layered artificial intelligence in the feed of streaming polymorphic unstructured data. The model aims to simulate mathematically and statistically the intrinsically human ability of fluid intelligence using components including but not limited to temporal relationships and eventual consistency in memory regression. With a data independent design, we are able to find relations between ideas within certain defined topics and the relationship between those ideas. Our algorithm was back tested against the “United States Financial Markets” as a topic with apparent and easily quantifiable relationships. We also test it against other topics of seemingly easily quantifiable relationships such as music industries, food chains, and drug target hierarchies as a means to prove data independence. We present a method to simulate multi-layered memory as a means to augment temporal relationships between nodes, define properties of node types without a predefined schema, and find centrality and distribution overlaps between the context sensitive surroundings of each idea node. We then present an in depth analysis into the errors with such a fluid intelligent machines and the simulacra that facilitate its fluid intelligent capabilities.