Towards a new computational model of mind and theory of the cerebral neocortex: Will Valentino help SAM to find a mind like ours for TARSAN ?

摘要

Building machines like us, machines with minds like ours, has been a goalof Artificial Intelligence (AI) since its inception. Unfortunately, the greaterpart of research in traditional AI has ignored work in neuroscience and thephilosophy of mind. At the same time, traditional AI systems havegenerally been brittle, difficult to engineer and, most importantly, havelacked the same nature and essential properties as human cognitivesystems. The objectives of this research are: 1) to justify and motivateresearch in computational neuroscience as an alternative path tounderstanding and engineering human cognitive systems, 2) to construct asoftware system for “wiring” and simulating brains at the functional (ornear-neural) level, and 3) to obtain a better understanding of highercognitive function based on a study of the neuroanatomy andneurophysiology, and current models and theories, of the cerebralneocortex. The thesis embraces a multi-disciplinary approach combiningwork in the philosophy of mind, neuroscience, and computer science.Although preferring to reverse engineer the brain (rather than engineering acognitive system independently, as generally done in AI), this work doesattempt to balance this analysis with the synthesis of systems and theories.Results include a resolution of the dispute between the classicalcognitivists and connectionists about the nature of mind which argues thatto build cognitive systems with minds like ours requires they be built at thefunctional (or near neural) level, a justification for the analysis of the brainto guide the construction of human-like cognitive systems, the constructionof a novel software system for “wiring” and simulating cognitive systemsat the functional (or near-neural) level, a review and analysis of thestructure and function of the cerebral neocortex, a review and analysis ofmodels and theories of, and systems inspired by, the cerebral neocortex,the preliminary aspects of a new architecture and learning strategy (whichrepresents preliminary steps towards a new theory) for the cerebralneocortex, and the construction of a powerful robot and environmentsimulation system. This thesis has taken initial steps to demonstrate that anappropriate simulation system, theory based on the human brain, and testenvironment can be constructed, that may indeed make it possible to buildcognitive systems with the same nature and essential properties as ours.An alternative AI may succeed.