A Mathematical Approach to Constraining Neural Abstraction and the Mechanisms Needed to Scale to Higher-Order Cognition
This addresses the challenge of achieving human-like intelligence in AI by providing a theoretical framework for neural abstraction, but it appears incremental as it builds on existing mathematical concepts without demonstrated empirical results.
The paper tackles the problem of scaling neural abstraction to higher-order cognition by proposing a mathematical approach using graph theory and spectral graph theory to constrain neural clusters based on eigen-relationships, with the result being a hierarchical hypothesis for scaling from small to large knowledge clusters to enable model building and reasoning.
Artificial intelligence has made great strides in the last decade but still falls short of the human brain, the best-known example of intelligence. Not much is known of the neural processes that allow the brain to make the leap to achieve so much from so little beyond its ability to create knowledge structures that can be flexibly and dynamically combined, recombined, and applied in new and novel ways. This paper proposes a mathematical approach using graph theory and spectral graph theory, to hypothesize how to constrain these neural clusters of information based on eigen-relationships. This same hypothesis is hierarchically applied to scale up from the smallest to the largest clusters of knowledge that eventually lead to model building and reasoning.