RI: Small: Extracting and Representing Commonsense Knowledge Using Language Models

Enabling computers to perform commonsense reasoning is a long-standing, if somewhat elusive, goal of artificial intelligence. Due to recent advances in the availability of data, and more powerful computational models for learning from those data, accelerated progress in scalable, robust commonsense reasoning now seems within reach. This proposal is aimed at a key component of commonsense reasoning: the automatic acquisition of common sense knowledge.