Autonomous Creation and Detection of Procedural Memory Scripts

摘要

Willingly or unwillingly, consciously or unconsciously, we all live scripted lives. This is not to say that one can look ahead into the next chapter and find out what one will have for dinner a month later. Life also brings surprises, but the scripts we are talking about are not the " scripts of life," but rather much smaller groupings of events that represent familiar routines and are stored as procedural memory in our brains. The general goal of the research presented here is to design, implement, test, evaluate, and improve a computational semantic model that will enable autonomous systems to evoke and use the correct procedural memory scripts similar to humans and thus approximate the human ability to comprehend and form procedures for learned tasks. The important step in this direction that is discussed here is to explore the constitutive properties of this artificial procedural memory model from the perspective of a broader artificial intelligence schema of information acquisition and retention and its computational semantic processing. The discussion will focus on the optimal ways for autonomous systems to acquire procedural memory scripts automatically and to activate them appropriately.