Download files
Abstract
This thesis examines bottom-up neurological methodology for creating biologically derived Artificial Intelligence (AI). Modern AI theory is strongly interdisciplinary, drawing on such diverse fields as the neurological and cognitive sciences, while also relying strongly on mechanical and software engineering. Throughout the philosophical examination of AI, these practical multidisciplinary aspects have not always been considered to the full extent necessary to understand and critique modern biologically derived AI.
If we are to take a functional replication approach to biological AI, then it is important that the theory of intelligence that we are functionally replicating is sound. Although there is a popular reliance on cognitive theory as a foundation for theories of biological intelligence, I show through a re-examination of the foundation of this idea and a closer critique of the Visual Representation debate that we cannot continue to assume that cognitive science has authority to dictate the structural account of cognitive and neurological relations. In order to establish a more useful framework in which to discuss the relationship between cognitive and neurological processes, I examine Glymour's reimagining of this relationship through his global weather analogy.
To demonstrate the success AI developers have had using neurological derived algorithms, I examine the work of Hawkins and the Numenta Platform for Intelligent Computing (NuPIC), and the latest work from Markram and the Brain Mind Institute, the Blue Brain Project. To demonstrate the strength of taking a bottom-up cognitive neurology approach to developing AI, I examine the holistic work of the DCN's ATR in Japan, and Milford and Wyeth's latest work on their RatSLAM algorithms.