Language models (LMs) are sentence-completion engines trained on massive\ncorpora. LMs have emerged as a significant breakthrough in natural-language\nprocessing, providing capabilities that go far beyond sentence completion\nincluding question answering, summarization, and natural-language inference.\nWhile many of these capabilities have potential application to cognitive\nsystems, exploiting language models as a source of task knowledge, especially\nfor task learning, offers significant, near-term benefits. We introduce\nlanguage models and the various tasks to which they have been applied and then\nreview methods of knowledge extraction from language models. The resulting\nanalysis outlines both the challenges and opportunities for using language\nmodels as a new knowledge source for cognitive systems. It also identifies\npossible ways to improve knowledge extraction from language models using the\ncapabilities provided by cognitive systems. Central to success will be the\nability of a cognitive agent to itself learn an abstract model of the knowledge\nimplicit in the LM as well as methods to extract high-quality knowledge\neffectively and efficiently. To illustrate, we introduce a hypothetical robot\nagent and describe how language models could extend its task knowledge and\nimprove its performance and the kinds of knowledge and methods the agent can\nuse to exploit the knowledge within a language model.\n