Computational Dreaming

点击翻译按钮获取中文摘要

By focusing on a set of the physical-structural foundations of dreaming, this research will investigate: (1) new organizational principles for parallel computation and (2) why dreaming is critical to intelligence. Seemingly against all survival instincts, all intelligent beings must sleep and dream, even if they are under duress, even if it endangers their very lives because they are in a hostile environment. Sleep that includes dreaming is strongly related to efficient mental processes. The thesis of this work is that the need to dream can be inferred from the brain's most striking physical and behavioral characteristics. The computer architecture that will be investigated is the DALI (Dream Architecture for Lateral Intelligence) a true Multiple Instruction, Single Datastream (MISD) architecture in which multiple models process the same input stream in real-time. While awake, some lateral processors are observers while one or more others are active. A dream phase of computation resolves divergence between processors in a competitive feedback phase not dominated by a flow of logic. Reality contains multiple views of the same thing, between different individuals and within the same individual, with incongruence resolved over time. The goal of the DALI is to include multiple, lateral models that process what the system observes, and then to model the competitive process of model resolution during a dream phase so that the system may be more effective for the next day's real-time (awake) response. The initial problem which will be investigated is contextual partitioning for speech recognition in pervasive, portable computing devices.