Neuroscience-Inspired Artificial Intelligence

Although neuroscience and artificial intelligence have an intertwined history, capabilities of biological brains are beyond the reach of state-of-the-art machine-learning algorithms. Therefore, better understanding of brain function would play a key role in building truly intelligent systems. Regarding the historical interactions between neuroscience and artificial intelligence, we will review the ongoing research on learning algorithms inspired by the study of biological neural computation. Based on what is known in neuroscience and psychology, it appears that cognition is the needed functionality that should be built into systems in order to reduce human intervention in control loops. A cognitive system has five building blocks, namely, the perception–action cycle, memory, attention, intelligence, and language, where language is more relevant in the context of a network of cognitive agents. Deploying memory and attention in learning algorithms and integrating them in the perception-action cycle of engineered systems will pave the way for building effective learning machines that are capable of hierarchical decision making and strategic planning. The flow of information in the nervous system plays a critical role in sustaining vital activities. Similarly, the flow of information in engineered systems is of critical importance, regarding performance and robustness of the system. Therefore, controlling the flow of information in the perception-action cycle deserves special attention in designing a new generation of learning machines. Following this line of thinking, artificial intelligence and neuroscience can significantly benefit from each other

کلیدواژه ها:

Neuroscience ، Artificial intelligence ، Neural computation